Home

Towards Hoarding Content in M-Learning

Contents

1. Figure 33 Words usage according to the word type Usage of additional material and notes taking Users that were preparing themselves for the exam of bilingualism were almost always using additionally a study book and a dictionary Partially this is due to the fact that in the current version not all word en tries of the texts are developed Moreover these users were also taking paper notes As we were not giving instructions on how the user could write and save the answers on the device and synchro 111 CHAPTER 5 CONTEXTUALIZATION AND OUTCOMES nize it later most of the users were not doing it but later they mentioned that this option will be very useful On the other hand some said that they study better if they take paper notes so they would do it anyway Users that were not preparing for the exam generally were not taking paper notes and were less using an ad ditional dictionary they were trying to guess the word meaning from the text context Target language Another difference was observed between the behaviour of users that were preparing for the exam and the ones that were just studying the target language As mentioned the texts are both in German and in Italian For the bilingual exam the student should read a text in one language and answer to ques tions in the other language This is done for both languages Texts have two difficulty levels AB and C Users who were pre
2. CHAPTER 7 CONCLUSIONS AND FUTURE WORK Butler M H 2001 Current Technologies for Device Independence HP Labs Tech Report HPL 2001 83 Chan T Sharples M Vavuola G Lonsdale P 2004 Educational Metadata for Mobile Learning In International Workshop on Wireless and Mobile Technologies in Education 2003 March 23 25 2004 Jhongli Taiwan Chang H Goldstein M Zeitoun A 2000 WebScrooge A Hoarding Agent for the World Wide Web Final Project Report available online at http www cecs umich edu azeitoun docs final_report582 ps on 24 10 2005 Chang C amp Sheu J 2002 Design and Implementation of Ad Hoc Classroom and eSchoolbag Systems for Ubiquitous Learning IEEE In ternational Workshop on Wireless and Mobile Technologies in Educa tion 2002 V xj Sweden pp 8 14 Chen M Zhang D Zhou L 2005 Providing web services to mobile users the architecture design of an m service portal Int J Mobile Communications Vol 3 No 1 pp 1 18 Cheverst K amp Davies N amp Mitchell K amp Friday A amp Efstratiou C 2000 Developing a Context aware Electronic Tourist Guide Some Is sues and Experiences CHI 2000 Netherlands Chisalita I amp Shahmehri N 2001 Issues in image utilization within mobile e services Tenth IEEE International Workshops on Enabling Technologies Infrastructure for Collaborative Enterprises 2001 Mas sachusetts p
3. Nevertheless they re port a small number of SMSs sent and received daily For Italy less then 5 SMS a day are sent by 67 of the students and re ceived by 65 For Bulgaria these percentages are even bigger 3 1 3 Ways of usage and attitude to e learning platforms Mobile learning is often considered the next step of e learning provoked by the new mobile technologies It was noticed that there are some differences in the ways students use e learning and one of the presumptions that we wanted to test is if those differ ences extend to their attitude to m learning In this section we only report the similarities and differences we have discovered and in a later section we discuss how we think the students ex perience with e learning influences their feelings about m learning We have found that the ways of using e learning differ slightly in the answers of Italian and Bulgarian students The rea sons might be partly because of the different offering of their own university s e learning platforms For example the percentage of students that do not use e learning is about 36 for the University of Trento and almost 10 less 27 for University of Rousse Differences in the ways e learning is used were found also ac cording to the studied subjects in both universities at the Uni versity of Trento non technical specialties utilize more the uni versity platform then technical ones which is on the contrary in Bulgaria This led us t
4. Hoarding step Figure 40 Hoarding presumable with Critical Set and after LO prioritizing Note that the Critical Set statistics might be done after cluster ing and classifying the users into groups For doing such experi ment and extracting meaningful results though we would need more tracking data than available currently 5 4 Other outcomes from the mobile learning system Some other outcomes not related to the hoarding problem were obtained by observing the participants and from interviews and 124 questionnaires with the users of Mobile ELDIT There are both positive and negative outcomes Most of the problems we found up to now are of a formal nature and are not directly related to our research work Therefore the positive outcomes reported at the beginning of this section are much more important for us and en courage us in our approach Nevertheless we find it important to report them as they would be common for all research in m learning domain and might help for the future development of mobile learning applications 5 4 1 Positive Outcomes v The users of Mobile ELDIT found the system very easy to use Even those that have never used similar devices started using the system almost without problems after a 10 minutes introduction The users liked the browser interface a lot as they felt familiar with this way of interaction v One thing that almost all users mention they liked in having a mobile
5. Note The procedure should be done only when the new user will start using the system not when the content packages are changed For 2 When setting up the NetFront3 browser see step 2b on page 169 you should point as a homepage the page following http www mobileeldit com User_Change instead of http www mobileELDIT com TextsList This will make the system show the following screen every time the NetFront3 browser is 165 APPENDIX A started Thus the user will choose from 9 possibilities This option is proper if the users are very responsible HOW TO SWITCH ON THE PROXY The system uses a local on the PDA caching proxy called FoxyProxy that should be located in the following directory On the PDA Program Files ewe a 4x 10 16 amp Dispositivo Program Files 14 02 05 14 8K FoxyProxy 14 02 05 14 8K The proxy should be switched on For doing this you should do the fol lowing steps RJ Foxy Proxy 1 Executing the FoxyProxy program pean iint irait der by clicking on it ean Soar n 4 Storage for backup CF Card 2 Starting the proxy by pressing the gt Zip File OPENED Start Proxy button Proxy Server gt Waiting for rem 3 You should see written on the screen what is shown on the picture on the right it might take about 1 min 4 Press Start Menu gt Today Note that you should NOT press the OK button in the top right corner but directly open t
6. connection charges will progressively reduce in the com ing years Though we also agree that this will sooner or later become true the current situation is not as we would like it to be The problem of mobile devices being often offline exists First we can not as sume that learners will equip themselves with the top technolo gies The devices have really become mobile in the sense of light small and powerful for impressively short periods of time and though there are quite a lot of technological ways to connect to the Internet through WAP GPRS Wi Fi etc still users have long periods of disconnection Second the always growing need of more space can be seen also with desktop PCs Once more space is available the user starts using it and will soon need more As it is true for the compression technologies and for web content caching that they will be always needed it will be the same with the mobile devices and hoarding Once we can put on the devices memory all the text data we will want to put video also once we can put video we will want higher quality that needs even more space etc All these led us to the conclusion that hoarding should be considered whenever we want to develop an efficient real world mobile learning system 1 3 Contributions of the Thesis The main goal of this thesis is to address the hoarding problem which has been previously weakly explored but is a particularly important issue in the mob
7. 39 7 To collaborate with other students 38 4 8 To fill in tests and questionnaires for self assessment 31 8 To access educational content off line 29 10 To receive supporting educational information via SMS MMS always 23 7 39 CHAPTER 3 RESEARCH CONTEXT Students preferences for mobile services Figure 4 Students preferences for mobile services As one can see more positive weight is given to supportive ser vices A few people mention that they just can not imagine the di dactic material viewable with mobile phone but are very optimis tic for the rest Some mention the fact that they expect that these services will be free of charge for the regular students and teach ers of the University based on authentication A frequent vision is of a system utilizing the university wireless network some times supposing a high bandwidth connection An interesting supposition very often found in the answers is that mobile learning should be the medium to facilitate the communication and collaboration between student student and student teacher probably because of the students perceiving the cell phone mainly as a device for conversations Another repeat edly given hint is that a nice university supporting m learning system should be strongly integrated with the university e learning and available there services As imagined by students an m learning system is often quite complicated and shou
8. 50 83 50 83 67 17 17 50 17 U41 67 50 83 100 33 100 33 100 67 83 83 0 0 33 0 U12 50 33 67 83 50 83 50 83 50 67 83 17 17 17 17 13 33 50 17 0 67 0 67 0 33 17 0 17 100 67 100 U14 33 50 17 0 67 0 67 0 33 17 0 17 100 67 100 U15 33 83 50 33 67 33 67 33 67 50 33 17 67 67 67 Uis 33 50 17 0 67 0 67 0 33 17 0 17 100 100 67 A similarity measure might be defined also directly based on the words used in every text as the grouping is done In other words similar will be users that have more cases of known unknown words that are the same In the most cases the similarity measured with the two methods overlaps well On the table below Table 6 we show an example with one of the users Table 6 Users Similarity comparison U Us Ua Us Us Uz Us Us Uio Urs Ure Ura U14 Uis Ure U based on 50 67 83 50 83 17 83 83 67 83 67 17 17 50 17 groups U based on 79 88 94 79 85 59 88 85 71 82 76 79 74 82 62 text2 We have performed experiments on clustering the users based on other parameters like the usage times both for single words and 120
9. AvantGo Palm Web Clipping and client based XHTML CSS A com prehensive survey of current technologies for device independ 13 CHAPTER 2 STATE OF THE ART ence and device independence activities can be found at 12 and on www w3 org detailed reference 108 Adapting the content through transcoding servers is one of the often used techniques The web content is retrieved from the Internet by the server and is converted into a form suitable for the device Different transcoding techniques are used for simply translating from one presentation language to another e g WAP HTML WAP for reducing the content size 54 for satisfying bandwidth or screen capabilities of the devices 8 9 to adapt the structure of the content in more appropriate logical fragments 86 41 113 or to present the content in some symbolic way 35 Some solutions also face the problems of connection speed and processing capabilities of the devices for delivering streaming media 91 All these approaches though suppose online access to the content Only some of the transcoding proxies e g www AvantGo com take care also of caching web pages for off line usage Caching and synchronization are two of the main problems of mobile applications in any domain Mobile devices are often disconnected because of the lack of access in certain places but also because of the high prices in most of the cases Two different situations arise when the device is disc
10. From the File Menu choose Home this should open the starting page of the sample Mobile ELDIT package with the list of available texts see the screenshots Note To exit the browser you should use the File menu Exit HOW TO CHANGE THE DATA PACKAGE To change the content data package the cache with new texts and related words you need to m Switch off the proxy and the browser Delete the old package Copy the new package N 3 4 Switch on the proxy again For step one the easiest way is to clean the memory from ALL programs including the proxy and the browser Please see the relevant HowTo on page 171 Step two might be performed on the device using the File Ex plorer StartMenu gt Programmi gt Esplora file Go to Dispositivo Program Files ewe mEldit By click and hold on the Cache item you will get a fall down menu from which choose Delete Elimina Note The step could be also performed from the desktop PC 168 For step three you need either to connect the device to the PC where you have the new content package or you have to use a memory card with the package copied to it previously In both cases you need to put the new package in Device Dispositivo Program Files ewe mEldit where the old package stood If necessary rename it so its name is Cache Step four follow the How to switch on the proxy instructions HOW TO GET AND SEND THE TRACKING DATA This procedure might
11. Klopfer E amp Squire K amp Jenkins H 2002 Environmental Detec tives PDAs as a Window into a Virtual Simulated World IEEE Inter national Workshop on Wireless and Mobile Technologies in Education 2002 V xj Sweden pp 95 98 Knutsson B amp Lu H amp Mogul J 2002 Architecture and pragmatics of server directed transcoding 7th International Workshop on Web Content Caching and Distribution 2002 Boulder Colorado USA Korkea aho M 2000 Context Aware Applications Survey avail able online at http www hut fi mkorkeaa doc context aware html Last accessed September 1 2003 Krashen S D 1981 Principles and Practice in Second Language Ac quisition English Language Teaching series London Prentice Hall In ternational UK Ltd Kuenning G H Popek G J 1997 Automated Hoarding for Mobile Computers Proc 16th ACM Symposium on Operating Systems Princi ples St Malo France Oct 1997 Kuenning G H Reiher P Ma W Popek G J 2002 Simplifying Automated Hoarding Methods 5 International Workshop on Model ing Analysis and Simulation of Wireless and Mobile Systems WiM 02 September 2002 Atlanta Georgia USA 149 59 60 61 62 63 64 65 66 67 68 69 70 CHAPTER 7 CONCLUSIONS AND FUTURE WORK Kurbel K Hilker J 2002 Requirements for a mobile e Learning Plat form IASTED 2
12. USA Quinn C 2000 Mobile Wireless In Your Pocket Learning LiNE Zine Learning in the new economy available online last viewed 12 11 2004 www linezine com 2 1 features cqmmwiyp htm Rich S amp Brebner G 2003 Storing and Accessing User Context 4th International Conference on Mobile Data Management 2003 Mel bourne Australia pp 1 8 Ryan Bob 2001 Dynabook Revisited with Alan Kay Byte Vol 16 2 February 1991 151 83 84 85 86 87 88 89 90 91 92 93 CHAPTER 7 CONCLUSIONS AND FUTURE WORK Sariola J amp Sampson J P amp Vuorinen R amp Kyn slahti H 2001 Promoting mLearning by the UniWap Project Within Higher Educa tion International Conference on Technology and Education 2001 Sariola J 2001 What are the limits of academic teaching In search of the opportunities of mobile learning TeleLearning 2001 Conference Vancouver Canada Sazawal V amp Want R amp Borriello G 2002 The Unigesture Ap proach One Handed Text Entry for Small Devices 4th International Symposium on Mobile Human Computer Interaction 2002 Pisa Italy pp 256 270 Schilit B N amp Trevor J amp Hilbert D M amp Koh T K 2001 m links An infrastructure for very small Internet devices 7th Annual In ternational Conference on Mobile Computing and Networking 2001 Rome Italy pp 122 13
13. You can get more information at Merano Bolzano Piazza della Rena 10 Via Cappuccini 28 Palazzo Esplanade I piano entra 39100 Bolzano ta piazzetta interna tel 0471 300789 fax 0471 303406 39012 Merano centromultilingue provincia bz it Tel 0473 252264 63 Fax 0473 252265 meranolingue provincia bz it 174 Who should be contacted if further information or help on Mobile ELDIT usage is needed For any further information help or comments please con tact the author in one of the following ways Work Address via Sommarive 14 C A P 38050 Povo TN Italy E mail trifonova science unitn it Office Phone 39 0461 88 2076 www Www science unitn it foxy Fax 39 0461 882093 175 Appendix B Mobile Learning can be viewed from various angles and research can be performed in different directions One very important and innovative one is on providing context aware services to the learner At the beginning of the thesis some work was done on exploring location dependent services in mobile learning domain As this work is not in the main focus of the thesis but is still re lated to the thesis we find suitable to report it as appendix Context dependent services in an m learning environment the printing case 1 Introduction and Related Work m learning is one of the successful buzzwords of the beginning of the millennium It combines the promises of two very promis ing fields e learning and mobi
14. like slow connection via mo dem are mentioned more that once as a problem which makes e learning not so comfortable and pleasant to use The connection speed proves to be a major consideration as about 40 of the Ital ian and 17 of Bulgarian students use modem connection at home Some participants mention that the computer distracts dur ing the study other say that e learning is too complicated or that a long usage of the computer is tiring especially for the eyes but also the cost of the computer itself considering that for using e learning you are obliged to have a PC or a laptop and the cost for the Internet connection appear to be a problem When asked about the connection between e learning and the quality of University instruction most of the students think there is such and it is a positive one About 75 of the Italian and 85 of Bulgarian students share the opinion that e learning in creases the quality of the university studies This is valid even for more than 50 of those that have never used e learning We have also asked the students if they would like to have access via Internet to video recorded lectures of the courses they follow A bit less then 85 of all participants were positive and more than 60 think that this possibility will not decrease the face to face attendance of the lectures For this scenario we did not find meaningful gender age specialty or other influencing factors 34 3 1 4 What about
15. years the quick growth of mobile technologies is promising a new revolution that might be comparable with the Web The forecasts for 2004 95 were that about 63 millions handhelds will be sold world wide and that approximately 38 of them will be smart phones integrating PDA functionality with features for commu nication And the forecasts are already getting true since accord ing to DoCoMo 87 more than 37 of Japanese population owns Internet capable phones More and more mobile devices with improved capabilities are appearing on the market In fact according to Canalys 99 in the beginning of 2006 13 per cent of all mobile devices will be smart phones which will number 16 9 million Again according to 99 by 2008 more than 130 million smart phones will be selling worldwide each year and Yankee Group predicts that there will be more than 300 million smart phones in circulation by 2009 Though the numbers vary in different sources there is a clear tendency for fast growth in the number of mobile devices Lots of mobile clients already support Java J2ME making it easier and less costly to develop portable applications Mobile learning m learning is a field which combines mobile computing and e learning Will e learning undergo a revo lution as it happened with the Web We do not know but we must try to answer the question by trying to imagine how mobile devices can enhance e learning or modify it Many people are
16. 100 it v mollare 1 lemma gt __ it n gancio 1 derivati 4 7 4 it v mollare 1 lemma eT 3 100 it v stirare 1 lemma gt itmgancio tdervati 3 7 3 4 100 it v rendere 1 lemma gt it n gancio 1 derivati 3 7 3 5 100 it v stirare 1 lemma gt it n gancio 1 derivati 4 7 4 When association rules are acquired after clustering the users as described in the previous section we can find even more mean ingful or stronger rules Examples are shown in Table 10 in this case more rules are found and the support of the extracted rules is 123 CHAPTER 5 CONTEXTUALIZATION AND OUTCOMES 35 50 With the previous settings in one of the two clusters only 2 rules are found while in the other cluster more than 100 tules The clustering described so far and association rules dis covering should serve for further decreasing the hoarding set It would be done by prioritizing the LO and setting certain limits for what to include exclude It leads to the situation shown in Figure 40 The green dotted line shows the presumable size of the hoard ing set if all mentioned techniques are used How high low the line will be depends on the parameters limits set for the prun ing Hoarding 100 90 Used Items Simple Hoarding 80 l With Critical Set 70 After Grouping and Association Rules Hoard Size percentage form the full material set
17. 2 STATE OF THE ART via the Internet but were working in separate locations and so on After few weeks of experiment the team found out that the students were using the devices mainly to read information from a digital medical handbook not to retrieve it via Internet as it was expected and as a communication device to discuss problems with colleagues but mainly for sending SMS messages and to or ganize social events after hours The research found that the rea son for this was that whiles the medical students were eager to test the PDAs and investigate how they can be useful in learning they still had some technical difficulties With proper guidelines and education the students might overcome these problems They concluded that the PDAs should not be regarded as Personal Digi tal Assistants but rather as gateways in complicated webs of in terdependent technical and social networks Research on new forms tools for collaboration has been go ing on in different institutions schools and Universities In differ ent projects people are experimenting on collaborative conceptual mapping and notes taking systems 49 71 An example of such collaborative work is a project that took place at MIT 53 The team used PDAs to simulate the real environment in the form of map and to use simulation for a game played by kids They use PDAs equipped with GPS extensions The idea is that the virtual world simulated on the PDA which has the same geog
18. 343 347 Judith Knapp and Anna Trifonova 2005 Mobile ELDIT Language Learn ing on the Go Academia EURAC s Science Magazine Vol 37 March 2005 Edited by Sigrid Hechensteiner amp Stefania Coluccia pp 21 22 Trifonova A Ronchetti M 2004 A General Architecture for M Learning International Journal of Digital Contents Vol 2 No 1 Special issue on Digital Learning Teaching Environments and Contents Proc of the II In ternational Conference on Multimedia and Information and Communication Technologies in Education mICTE2003 Badajoz Spain December 3 6 2003 Edited by Antonio M ndez Vilas and J A Mesa Gonzalez pp 31 36 Trifonova A Ronchetti M 2004 A General Architecture to Support Mobil ity in Learning Proc of the 4th IEEE International Conference on Advanced Learning Technologies ICALT 2004 Crafting Learning within Context August 30 September 1 2004 Joensuu Finland pp 26 30 Trifonova A Knapp J Ronchetti M Gamper J 2004 Mobile ELDIT Transition from an e Learning to an m Learning System Proc of the World Conference on Educational Multimedia Hypermedia and Telecommunica tions ED MEDIA 04 June 21 26 2004 Lugano Switzerland pp 188 193 Ian G Kennedy Sanaz Fallahkhair Veronica Rossano Anna Trifonova An tonella Grasso Sabine Graf Jean Claude Ziswiler Ricardo Fraser 2004 Simple Web based Adaptive Learning Technology Learning Technology newsletter Vol 6 Issue 4 Octob
19. 5 80 48 7 70 61 9 60 50 No opinion High m Normal m Low 40 T 40 6 30 20 10 T BG M BG M BG M BG TT BG PC Cell Notebook PDA Smart phone phone 0 Figure 2 Opinion about devices prices When talking about prices of services we can see on Figure 3 below that the usage of cell phone is still considered costly by 66 4 of Italian and 72 of Bulgarian students Less then 2 of the participants consider these prices low About 30 of Italians and 21 of Bulgarians think that the prices are normal The situation differs little when talking about prices for us ing the Internet also shown on Figure 3 below The opinion that the Internet usage has a normal price is considerably higher in 30 confrontation to cell phone services 38 in Italy and 49 in Bulgaria However the opinion that the prices are high is still strong about 50 of all participants No opinion High E Normal m Low Internet Usage Cell Phone Usage Figure 3 Opinion about services prices Usage We have tried to discover how our participants use the devices and services available to them to see if we can anticipate its influence on mobile learning Focusing on the connection type that is used by the students at home we found out that for Italians a relatively l
20. Apache Group Tomcat 4 1 webapps cocoon mount mEldit word dtd gt oun index 1 language italiano gt n famig lt lemma id famiglia 1 lemma gt famiglia lt lemma gt lt morphology id it n famiglia 1 morphology gt lt w type content gt la lt jv gt lt w type content gt famiglia lt jv gt lt w type content gt lt w gt lt w type content gt le lt w gt lt w type content gt famiglie lt jv gt lt w type conten gt lt jw gt lt morphology gt lt sense id it n fami ja 1 sense0 gt n famiglia 1 sense0 subsense0 gt iglia 1 sense0 subsense0 def0 gt N type content gt Comunita lt jw gt ontent gt di lt jw gt type content gt persone lt w gt type content gt lt jw gt ersona lexref lexref SSS ctag comporre lexref it v comporre 1 lemma ctag type content gt composta lt w gt da lexref it ition 1 ctag S typ content gt da lt w gt SSS ctag PO arial moglie lexref it n moglie 1 lemma cta w type content gt lt jw gt jio 1 lemma ctag N type content gt figli lt jv gt S type content gt e lt jw gt x Figure 15 Low granulated raw data XML file For Mobile ELDIT we have decided to use server side adaptation namely XSLT transformations Figure 16 of the XML data Figure 15 on a Cocoon server corresponding to the Presenta tion Adaptation module on Figure 5 in
21. CHAPTER 2 STATE OF THE ART come into play in an important way Activity theory theories of adult informal learning lifelong learning and etc are at the basis of lots of experiments using mobile devices Let us start with the simplest interactions Although simple learning related applications may benefit from the messaging ca pabilities of mobile systems only relatively few different educa tional bodies have made experiments in this area At Kingston University UK an experiment was under taken to research the effectiveness of a two way SMS campaign in the university environment 96 97 The team has developed a system that sends SMS to students registered to the service The content of messages is about their schedule changes in it exami nations dates and places student s marks etc After registering the students were automatically separated in 5 different groups One group was receiving announcements via e mail other 3 groups via SMS but different interaction was necessary in every group and the last via web The conclusions of the experiment were that the students in certain scenarios where a certain type of response is required preferred SMS as a medium to e mail or web based announces They feel the data is more personal and they like this SMS could be efficiently used in education m learning as a complementary media As the technology improves i e EMS and MMS the potential for more user friendly inter faces t
22. Further we will use the term single session to indicate the first definition above while we shall use session to indicate the hoarding related meaning We will also speak of daily session to mean all the activity that has taken place in a calendar day 4 3 Hoarding on the first access to the system Earlier we mentioned that the hoarding process differs on the first access of a user to the mobile system This happens because we do not know this concrete user and his her particularities Nev ertheless most of the steps of the hoarding should exist although they will be a little changed We still have to predict the starting point to generate a candidate set and to try to sort the objects in this set but in this first access of the user the hoarding sub system should calculate and use some default values extracted by analys ing the behaviour of all previous users of the system If the mo bile application is an addition to an online e platform it is possible 83 CHAPTER 4 SOLUTION OUTLINE and even desired to see if some knowledge could be extracted about the user from the e learning system and use it instead Let us start with the learner s entry point Often learning material is created by the educator with a certain sequence in in which the students browse dur mind Thus based on the additional knowledge about the learning material structure the system can be aware of the most probable star
23. Note that this evaluation criterion can be used only on real use of a system and its hoard part It is also strongly related to the hoarding size Another possible measurement is the miss free hoard size defined as the minimum amount of disc space that a particular hoarding system would require to allow a complete dis connection period to take place without any misses The two important measurements that can be used by the hoarding for improving its work on every step are the hit rate and the miss rate A low hit rate means that the hoarding was somehow ineffective because much unneeded stuff has been cached The user is never directly aware of a low hit rate but s he is strongly affected by a high miss rate since it measures the sys tem s failure to respond to the user s requests Of course the two measures are somehow interrelated wrong priorities might lead to include some unneeded stuff in place of some useful one therefore adversely affecting both measures Set of LO selected by the hoarding algorithm Set of LO used by the student in one session Figure 19 The ideal hoarding set The goal of the algorithm is to maximize the hit rate and at the same time to minimize the miss rate The ideal situation is to achieve hit_rate 100 and miss_rate 0 which would mean than the hoarding set contains all and only the items that the user needs during her his studying session as shown in Figure 19 above O
24. Results Proc of World Con ference on Educational Multimedia Hypermedia and Telecommunications ED Media 2005 June 27 July 2 2005 Montreal Canada pp 4751 4758 e Trifonova A Ronchetti M 2005 Hoarding Content in an M Learning Sys tem Proc of World Conference on Educational Multimedia Hypermedia and Telecommunications ED Media 2005 June 27 July 2 2005 Montreal Can ada pp 4786 4794 e Trifonova A Ronchetti M 2005 User Behaviour Observations for Support ing Offline Delivering of Learning Materials in a Mobile System Proc of World Conference on Educational Multimedia Hypermedia and Telecommu nications ED Media 2005 June 27 July 2 2005 Montreal Canada pp 1520 1527 e Trifonova A Ronchetti M 2005 Hoarding Content in M Learning Context Proc of PerEL 2005 Workshop on Pervasive eLearning held in conjunction with the Third IEEE International Conference on Pervasive Computing and Communications PerCom 05 March 8 12 2005 Kauai Island Hawaii IEEE Computer Society Press 2005 pp 327 331 e Trifonova A Ronchetti M 2005 Prepare for Bilingualism Exam with a PDA in your hands Proc of the International Conference on Methods and 187 APPENDIX C LIST OF PUBLICATIONS Technologies for Learning ICMTL 05 March 9 11 2005 Palermo Italy WIT Transactions on Information and Communication Technologies vol 34 Edited by G Chiazzese M Allegra A chifari amp S Ottaviano pp
25. S amp Wright K 2002 Nokia and Midwest Wire less Establish Model Wireless Campus at Minnesota State University Nokia press release 2002 Last accessed Sept 1 03 available online at http press nokia com PR 200009 790728 _5 html 108 W3C Device Independence supportive initiatives and technologies a Device Independence Activity Access to a Unified Web from Any Device in Any Context by Anyone http Avww w3 org 2001 di b Mobile Web Initiative hitp Awww w3 org Mobile c XML http www w3 org XML d Style Activity CSS XHTML SVG SMIL XSL and etc http www w3 org Style e Composite Capabilities Preference Profiles CC PP www w3 org Mobile CCPP 109 Wang J 1999 A Survey of Web Caching Schemes for the Internet ACM Computer Communication Review 25 9 pp 36 46 110 Want R amp Hopper A amp Falcao V amp Gibbson J 1992 The Active Badge Location System ACM Journal of Transactions on Information Systems 10 1 pp 91 102 111 Ward A amp Jones A amp Hopper A 1997 A New Location Technique for the Active Office IEEE Journal Personal Communications 4 5 pp 42 47 112 WIPS 2000 WIPS Technical Documentation available online http 2g1319 ssvl kth se 2000 group12 technical html Royal Institute of Technology Sweden Last accessed September 1 03 113 Yang S amp Lee H amp Chung K amp Kim H 2002 A Content Pro
26. Section 3 1 5 View XML2HTMLword xsl alol xi File Edt View Help lt xml version 1 0 gt lt xsl stylesheet xmins xsl http avw w3 org 1 999 XSL Transform version 1 0 gt lt xsl output method html indent yes gt lt xsl output encoding 1S0 8859 1 gt lt xsl preserve space elements gt lt xsl template match word gt lt xsl apply templates select child nodef gt lt xsl template gt lt xsl template match structural noun adjective verb gt lt xsl variable name href_base select gt lt HTML gt lt HEAD gt lt META HTTP EQUIV Content Type CONTENT text html ISO 8859 1 Y gt lt TITLE gt lt xsl value of select lemma gt lemma lt TITLE gt lt HEAD gt lt BODY bgcolor FFCC99 text 000000 lt center gt lt H1 gt lt xsl value of select lt xslichoose gt lt xsi when test node contains zwsp gt lt xsl apply templates select morphology gt lt ixsi when gt lt xslotherwise gt lt H3 gt lt xsl apply templates select morphology gt lt H3 gt lt ixskotherwise gt lt ixsl choose gt link 000000 viink 000000 alink 333333 gt lemma gt lt H1 gt lt jcenter gt lt HR gt lt xshtextVai a lt xsl text gt Figure 16 XSLT for word entries 73 CHAPTER 3 RESEARCH CONTEXT Our decision was motivated by two facts 1 on one hand our data was already in XML format as shown on Figure 15 which allows an easy creation
27. We analysed different ways to apply mobile devices for educational purposes This led us to classifying services that are specific and should be provided by a general m learning platform and later we concentrate on one of these services as a concrete problem to solve during the thesis Namely this is the hoarding of content for offline usage 1 2 The Problem and the Motivation The problem we focus on is the one of supporting the access to web based learning content from a PDA device during its periods of disconnection Such offline periods may appear for different reasons intentional e g the available connection is too expen sive for the user or unintentional e g lack of infrastructure at a given time and location Such offline periods are frequent nowa days and our expectations are that the situation will not change much in the next years During offline periods the user can only access materials located on the device s local memory Mobile systems typically have a relatively small amount of memory which is often not enough to store all available study material In such a case a decision should be taken on which part of the mate rial will be needed and has to be cached Often we can not count on the user s own judgment of what he she will need and pre fetch it Rather in our opinion some sort of automatic pre fetching would be desirable The process of automatic selection 2 and caching of material to be used during offli
28. a basic dictionary might provide only translation of those words available in ELDIT Add multimedia material We mentioned before that ELDIT con tent is continuously growing This is true for enrichment of the dictionary with words meanings examples etc and also for add ing different multimedia material While in ELDIT most of the content is still in text form Mobile ELDIT in its current version contains only text In this context it is a necessary next step for improving the existing system to include also adaptation and transcoding of the multimedia content for the mobile users At present the ELDIT provides certain explanations and examples in picture form and pronunciations as sound However video formats should be also anticipated and support should be provided in Mo bile ELDIT Add collaboration functionality Mobile ELDIT is a limited ver sion of an online language learning system One part that we did not include was the collaboration between the learners The so called tandem module of ELDIT allows users with different mother tongues Italian and German to collaborate by playing the role of the teacher for the other person in a couple This means that an Italian native speaking person will check the answers to the comprehension questions written in Italian by a German na tive speaking person This very useful functionality was also re quested by some of the Mobile ELDIT users A possible im provement would be to connect
29. activities In collaboration with L Istituto Svizzero di Pedagogia per la Formazione Professionale a class of 10 students used the system for the period of 4 weeks Both teachers and pupils report that the system integrates well with their study process and are eager to continue using it in the future CHAPTER 1 INTRODUCTION e Last but not least though not directly connected to the hoard ing problem we should also mention the outcomes of a survey aiming to determine the readiness of University students for mobile learning that was carried out involving more than 600 Italian and 200 Bulgarian participants Section 3 1 The survey led to very interesting and important deductions about parameters that influence students attitude to mobile learning in general and their preferences for prospective services that should be supported at their university 1 4 Thesis Organization The thesis organization basically reflects the process of our work over time As illustrated above the thesis is directed towards solv ing the hoarding problem In order to research possible solutions the full context around hoarding had to be reproduced Thus the thesis treats a number of questions and looks at different aspects of the whole process of designing and developing an innovative mobile learning system The work started from scratch and ad vanced step by step requiring decisions to be taken at every phase This leads to the logical structure of the t
30. aggregated data for every text As explained earlier section 5 3 2 our users reported longer review times if their goal was to take the examination Our supposition was that such study might help for prediction of students goals more specifically about the goal to take the bilingualism exam or not X Links_Count Num E Y Total_time Num Colour Cluster Nom E Select Instance J Plot QueryResult_clustered Figure 39 Clustering of users based on requests number and spent time However due to the small number of participants and in particular small number of user who aimed at the exam we were unable to obtain meaningful data for the hoarding results Still clustering according to usage time as the one shown on Figure 39 will cer tainly be useful as a step further after the separation based on concrete requested words An important question to answer for a real word system would be what should the number of clusters be and what does this number depend on In our experiments we used two differ ent methods 1 leave the algorithm automatically to discover the best number of clusters and 2 force the creation of to 2 3 etc number of clusters It proved that in certain cases the automatic separation is not possible again because of the small data set we experimented with On the other hand in Figure 39 we show re 121 CHAPTER 5 CONTEXTUALIZATION AND OUTCOMES sults when the cl
31. always connected to Internet through WAP GPRS UMTS Blue tooth etc Pure mobility is when no connection is available and so all the data the applications need should be uploaded on the device and used offline The first op tion gives strong impact on context dependent applica tions while the second approach needs research on data management e Adaptation to the surrounding context in a mobile envi ronment is also a very interesting and promising area Finally we note that whenever a new technology comes it takes a while until its real potential is deployed because we continue thinking according to old paradigms We probably have not yet found the new paradigm for fully deploying e learning and yet another variation comes It is up to our ingenuity to free ourselves from the old thinking and unleash the power of our fantasy to allow a new revolution to happen 25 Chapter 3 3 The Research Context This section outlines our research context The first sub section presents a survey performed in a cooperation between the Univer sity of Trento Italy and the University of Rouse Bulgaria The aim of the survey was to see the readiness of the university stu dents for mobile learning Afterwards we present the general mo bile learning architecture with elaborate analysis of how it was designed and what each module is responsible for The third sec tion is dedicated to the Mobile ELDIT system developed accord ing
32. amp Y NetFront v3 1 4 dx 13 51 amp probable reason When the system have shown you a ay context menu you click l and held on the touch screen you unintentionally pressed the New Window choice solution Click and hold somewhere in the blank field and from the menu that will appear choose Close Window KNOWN PROBLEMS There are few known bugs that appear in the current version of the system that should be listed here When using the NetFront3 browser version 3 1 there is a small problem displaying the special German and Italian let ters German umlauds A 6 amp Italian LLO After such letter a space is displayed which in practice should not exist The problem is due to the NetFront3 implementation where this letters are not correctly supported As the developers are aware of the problem it is expected to be corrected in later versions With some PDA devices namely iPaq 3800 but probably can happen on other models a sudden crash of the proxy was no ticed that happens if the device is often switched off and on without exiting the started applications Our suggestion is that a memory optimization is done automatically by the OS 172 which is out of our control In such case the proxy needs to be restarted The problem will be easily noticed as on the first request done by the user the browser will give a message that no response was recei
33. an important issue in the real world and as will be discussed throughout the thesis it appeared to be not a trivial research issue However as the prob lem was often ignored we think it is essential to stress its exis tence e The main goal of the thesis was to explore the hoarding prob lem In this context we propose a general approach for hoarding Section 4 that explores the integration of a couple of 5 CHAPTER 1 INTRODUCTION different techniques into the hoarding algorithm Our bottom up approach to hoarding starts from the special case of a real world system and is based on a set of general principles described in section 3 1 5 Our ultimate goal was however to provide a general strategy that can be also used in different mobile learning systems with a relatively big learning materials base and that need to ac cess it during offline periods We describe ideas of the possible approaches what algorithms are appropriate and in what cases We analyse how different parameters that emerge from our work should be tuned by researchers and developers of mobile learning solutions All discussed would help to those who also want to automatically decide what part of the learning content the user will need in the next offline period and do the caching e As mentioned earlier the domain of mobile learning is quite young and researchers are wandering between one problem and another and between different technological solutions One of the
34. and has two main data streams words corpus learner s dictionary and texts corpus with comprehension questions The text corpus about 800 texts split into thematic groups and two difficulty levels both for Italian and German languages has been collected by the Goethe Institut Milano 1 For ELDIT an XML version of the texts has been created For major details about ELDIT see 46 ELDIT is designed according to the principle of separation between data and their presentation The data are XML formatted see 108 c and the learning content is very low granulated The text corpus is the main part that is later adapted to be used via mobile devices Every text is made of about 150 words and addi tional comprehension questions that the user should answer as required for the bilingualism exam in the other language Words currently nouns verbs and adjectives are connected to their en try in the dictionary On the other hand each word entry contains explanations translations and examples on different senses It also provides additional information like idiomatic expressions derivations from the word etc The online system contains more than 600MB of raw data Moreover such data are continuously growing as the ELDIT system evolves and the data are being en riched over time 3 3 2 Why Motivations for the Mobile ELDIT First of all the field that we have selected for our experimenta tions in mobile learning is the one of l
35. be done at any time but the most suitable two options would be either when new package is uploaded or when the PDA will not be used anymore by the same user The preferred option is the first one on package change The tracking data is automatically recorded in log files and also backups are often done on the external memory The files that are needed are 1 From the memory card all the files with name Backup_ log Note The files might be quite a lot as number not in size 2 Two file from the folder Device Dispositivo Program Files ewe mEldit which have the following names History log Feedback log The star means that there is a varying text on this place I would prefer that from any device these files are copied sepa rately into a folder with the name of the user on the desktop PC and zipped are send to me by mail 169 APPENDIX A Note It is not that important not to mix up the files as their names contain the information we need to split them but some post processing time could be saved When you are sure that these files are safely copied on the PC please delete them from the device both from the local and from the external memory HOWTO CLEAN UP THE MEMORY From the Start Menu choose Impostazioni In the System tab press the Memory button and go to the third tab which is Programmi in esecuzione A screen should be as shown on the figure below All unnecessary programs should
36. contributions during this thesis to the research community was the survey of the state of the art and ongoing projects partially Sec tion 2 which was done at a very early stage of this work It pro vides guidelines for successful development of mobile learn ing applications and directions for further contributions in the field see 2 3 The overview published in one of the biggest e learning conferences E Learn 03 was further cited and used as a reference point by multiple researchers e Based on the above mentioned survey we proposed a gen eral architecture for mobile e learning Section 3 1 5 It ex plores the possibilities to extend e learning system so as to pro vide services to mobile devices These services range from distribution of didactic material to support of location aware ser vices to mobile users equipped with variety of devices The pro posed architecture is general and would be able to provide all possible services from an e learning platform plus additional ser vices only for mobile users At the same time it is extensible for the new generation of devices e Based on the proposed general architecture and general ap proach to hoarding we developed a working proof of concept system called Mobile ELDIT The system was designed after analysing various suggestions from researchers in the field of mobile learning described in the state of the art section and comprising with all the findings so that it shou
37. cost of the books sometimes printouts are even omitted by using lecturers slides or other digital materials available via the e learning platforms the less need to be physically presented at the university thus sometimes skipping costly and time expensive journeys the ease to compensate the loss of a missed or skipped lecture etc On the other hand a reason frequently given against e learning is the lack of personal contact between teachers and stu dents Students often prefer the traditional educational approaches blackboard and chalk reading books in the library and everyday contacts at the faculty for University of Trento these are about 80 of the negative and neutral opinions It should be mentioned that even some of those students that have positive attitude to e learning as a useful medium to fast and time independent access to learning materials mention that it is a cold environment and 33 CHAPTER 3 RESEARCH CONTEXT that there is a lack of personal contact For many students e learning is not necessary or even quite useless at this stage be cause unfortunately material is not provided for a given spe cialty subject or e learning is not supported by enough courses they participate or because it is not well supported materials are not updated often enough An interesting opinion of a student is that e learning is yet not enough mature well developed to be useful Also technological obstacles
38. courses and SMS notification systems were pub lished by different universities in the last couples of years A few examples are HyWeb 45 at Griffith University Gold Coast mid 2000 107 at Minnesota State University and the NAIT http www nait ab ca MobileLearning m learning project in Canada An m learning project that focuses on the testing of the use of WAP technology in higher education is the UniWap project 83 84 88 The team tries to explore the process of creating an operating environment for studying and teaching through smart phones and WAP phones The Virtual University needs to support the mobility of the participants of the learning process the stu dents and also the teachers One phase of the project was to cre ate some working prototypes courses modules and to investigate the problems and the value of such courses The positive results they encountered easy to develop willingly accepted and widely used modules encourage them to continue investigating the new coming technologies digital imaging with mobile devices 3G etc From E learning to M Learning is a long time project see http learning ericsson net leonardo thebook book html that aims to create a learning environment for wireless technologies by developing course materials for range of mobile devices A discussion about the characteristics of the devices that are proper 17 CHAPTER 2 STATE OF THE ART for learning is made when taking
39. details of our approach to analysing the data gathered from the Mobile ELDIT we should once again men tion that the common ways to determine the user characteristics and knowledge generally in e learning systems but also in m learning are assessment through questionnaires quizzes and tests and letting the user manually set his her own preferences Our re 98 search interest though falls on the automatic discovery of these attributes and for this reason we were collecting the same tracking data as a normal standard proxy would collect It is obvious that the user knowledge of a concept determined by assessing him her and checking the tests results especially if a human teacher is in volved can commonly give some quite precise quantified meas ure of the learner understanding and advances in the chosen sub ject On the other hand by analysing only the user s interaction with certain system might give less precise approximation but might sometimes makes life easier As one can see on Figure 26 above with the help of differ ent algorithms for knowledge extraction we expect to get two types of data on one side are the different typical usage patterns that we need to extract out of all available data set and on the other side is the understanding and categorization of every con crete user It should be possible to automatically extract knowl edge for all three groups of user modelling parameters discussed previously in Section 4 8
40. development so lution that should satisfy the need of a given project v As our main experiments were carried out on Windows based devices we gathered some experience on the specific prob 126 lems that appear with them A significant problem is that the battery of Windows based devices discharges quite fast When a device is frequently used it discharges in 1 2 days but the main problem comes from the fact that even when not used the battery discharges in about a week time The dis charged device forgets the software installed by the user administrator and all user s data This leads to the neces sity to do backups of all important data on an external mem ory quite often It is also very inconvenient and even irritating as all the programs that were installed should be re installed Another particularity of the Windows CE based PDA devices is that once a program is run it remains in the device memory until specifically closed from special menu command The misunderstanding comes as the programs generally are not closed when the x button in the upper right corner is pressed while the users think it will As lots of users were not previ ously familiar with such devices they were often clicking and starting by chance and unwanted different software Once re alising the mistake they were closing it as they would in a desktop PC by clicking the x button In this manner very of ten lots of programs remain open and occupy device s m
41. equipment etc 110 112 111 78 40 39 These different systems ad dress different problems and so the location sensing in each of them has different parameters properties and accuracy Some of them are suitable only for finding the position of the device when outdoors Global Positioning System GPS while other only work indoors Additional infrastructure and or equipment is nec essary for most of the location determining systems i e Active Badge and Active Bat systems require special tags and ba sis stations in the GPS case the infrastructure is already in place so that it can be given for granted but the user is required to have additional hardware on the client machine a GPS receiver In general more appropriate would be a system that does not require additional hardware or infrastructure In our system we use the IEEE 811b network that is already in place so it re quires only adding a software layer A small module on the mo bile device connects sequentially to three or more access points in the wireless local network and measures the signal strength the wireless network card acts as a sensor Note that in a conceivably wireless networked city such method would work indoors and outdoors The results of the measurements are used to determine the position There are different ways of doing this With one of the methods called regression the position could be returned in 183 APPENDIX B physical coordinates
42. for the discovery of context one for mobile content manage ment and adaptation and one for packaging and synchronization of the content for supporting offline delivery of learning material With our implementation we show the viability of our arguments 135 CHAPTER 7 CONCLUSIONS AND FUTURE WORK We have successfully designed and implemented a real world mobile learning system called Mobile ELDIT Behind it sits an innovative language learning e learning system called ELDIT Mobile ELDIT helped us also in gathering good practical experience from the work with real m learning users Based on the users feedback we concluded once again that hoarding is very important and should be considered in developing m learning platforms Users suggestions and advices also helped in under standing real students needs for further system improvements Considering hoarding we have outlined the general solution i e provided a theoretical plan for action with possible techniques to be used We discussed step by step in details the proposed strategy leaving it abstract enough to be general and applicable to different mobile learning systems We point into some particulari ties of the mobile scenario that would influence on hoarding like showing the importance of a new definition of user session for this scenario and the utilization of the measurements over it in hoarding We compared different approaches measures for pre senting the successful
43. frequency of access and the active periods of the file usage The algorithm also considers upper space limit of memory The re ported effectiveness of their filtering algorithm is more than 84 115 132 Facing the hoarding problem for mobile computing discon nected operation an interesting solution has been proposed in SEER 57 The authors were also inspired by the work on Coda system but go in different direction They defined a new measure semantic distance between individual files by observing the user activities and propose an algorithm for automatic hoarding of projects for mobile computers With semantic distance the au thors try to quantify the user s intuition about the relationship be tween files in the same project For this different measuring crite ria are used temporal semantic distance sequence based semantic distance lifetime semantic distance directory mem bership filename conventions and hot links These criteria are combined to assign weights to documents and take decisions for hoarding them in an automatic way automatic periodic hoard ing The approach met some unpredictable behaviour in the real world system which appeared because of the way the operating systems and some often used programs work like the find op eration under Unix Recent experimentations with the same sys tem 58 showed surprising finding the complex clustering methods that are used in the
44. from where you learn it increases in dependence of your wish to learn The modern student is delocalized and the educa tional institutions must encourage his her global thinking and per formance and not to restrict him her in terms of time and space The students can participate in interesting lectures which are not in the frame of their educational profile Q Which services must mobile learning provide Describe how you imagine a mobile learning system For this question the students were first given a list of possible services and had to check which seemed useful for them After wards they were supposed to describe what they imagine will be offered by a mobile learning system and what are the services they consider valuable A large number of students answers dis cuss as possible services all or part of those mentioned in the list given by us Others suppose m learning should provide the same functionalities as e learning whenever possible Summing up all students responses students expect that most helpful and used will be the services in the following order 1 To access supporting educational information e g schedulers exams results via WWW 79 4 2 To communicate with teachers 65 3 To access educational content online 54 4 4 To communicate with other students 53 7 5 To receive supporting educational information via SMS MMS on demand request 50 5 6 To fill in tests and questionnaires for exams
45. gt T ae EET o be closed by selecting each of Memoria them clicking on the name in the Programmi in esecuzione list and then clicking on the Close Termina button Esplora file NetFront v3 1 Note For Mobile ELDIT system to work calendario properly only the FoxyProxy and mms Termine the NetFront3 browser are neces sary If the FoxyProxy is not avail able or is by chance switched off Eresrammiin esecuzione Ls M 3 Disinst appl per liberare spazio in memoria please switch it on Trova File grandi che usano molta memoria WHAT TO DO IF o THE SYSTEM OR THE PDA ITSELF WORKS VERY SLOW probable reason Too many programs are started simultane ously and occupy big amount of memory solution 1 Clean up the memory see above e THE SYSTEM DOES NOT WORK 170 symptom Browser returns the follow Browser Warning fok ing message A Cannot read page probable reason the proxy is switched Steege TSi ah off Might be because of an accidental Type TCP connect action from the side of the user or an overload of the memory solution 1 Clean up the memory 2 Start up the proxy e NO INFORMATION IS SHOWN tFront v3 1 gh 4x 14 07 Q symptom A strange web page is fitcom de v zahien Lidiomexer_ ty a shown saying something in Eng e Eld lish as the one on the picture reason The word you have re quested is not available at the moment and
46. imaginary picture in free text The questions were split into thematic groups like Avail ability of devices E Learning Usage Opinion about prices etc Later on we performed grouping of the users according to their answers to specific questions and did the analysis and de ductions A complete report of the students answers full statisti cal data and comparative graphics can be found in a separate re port 104 while here we give a summary and some of the most interesting and important results for this thesis 3 1 1 General Information About 600 Italian students participated from the University of Trento They were mainly from the Science and Engineering dis ciplines respectively 30 6 and 57 6 but also from more hu manistic faculties like psychology economics languages and other in total 11 4 Considering gender 71 1 of the partici pants were males which is due to the fact that big part of the par ticipants more than 50 was from the engineering disciplines where in general the percentage of male students is noticeably greater than the one of females The students were evenly distrib uted between the different years of their study both bachelor and master and mostly less than 25 years old 87 3 About 95 of the Italian participants were of Italian nationality Bulgarian students were about 200 and were with more smooth distribution across University faculties Nevertheless also here student
47. is certainly possible although it requires digging in OS dependent technical details The Client Device Server a sae Figure 44 The printing process We implemented a little less convenient but more immediate so lution in which we perform the following steps the user generates a postscript file as we described it earlier then he she contacts explicitly through HTTP to the server using a web browser The server provides a form where the user points to the file and sends it to the server An active component e g a servlet than opens a 182 socket to the client and collects the context information On this basis the server finds the nearest printer prints the document and informs the user of the choice together with explanation where the printer is located The information is returned again in the browser via HTML page Of course we have here implied that the mobile system is able to provide a service via socket to pass the context dependent info As an alternative one could pass all this info through HTTP Figure 1 describes the whole process As far as the positioning system is concerned we note that to implement a location aware system we need a proper posi tioning system and there are many possible solutions provided in literature Different technologies are developed for determining the user s location Lots of research had been done and systems had been made for automatically locating people
48. learning system is the availability because of the fact that the device is light and small one can put it in his her pocket or purse and have it with him her all the time v Some users of Mobile ELDIT were familiar with ELDIT the online desktop system They reported after getting used with the mobile system to have started to use the two systems in different ways As the mobile one was more comfortable for using it in any moment they started using it for systematic studying especially on the road However often the mobile system was utilised also at home even when a PC was avail able On the other hand they started to utilize the online sys tem more often for searching and controlling the meaning of the arbitrary words mainly at work v The users liked a lot the freedom that the mobile device gives them Some of them often used the system in the train while travelling others at home or in the office To the question Js there a place where you preferred using the system Why one user responded On the coach Because it is comfortable 125 CHAPTER 5 CONTEXTUALIZATION AND OUTCOMES Another user answered None As I could and did use it in any environment v Two of the persons who used a mobile device as an addi tional tool for the exam preparation have passed the exam and said that the system has helped them a lot Though this is not a real measurement for an increased learning effectiveness it is a hint fo
49. m learning First of all we should mention that very little of the students knew what mobile learning is and have ever used it only 4 6 of the Italian students and 2 5 of the Bulgarian participants All of them use also e learning and were mainly boys 74 of the Ital ian and all Bulgarian participants It should be underlined that an m learning platform is not offered by any of the two universities for the studied subjects In this section we give a complete analy sis of the students answers on three main questions namely 1 would they like to use m learning and why 2 will in their opinion m learning increase the quality of edu cation and why 3 what services should mobile learning provide Q Would you like to use m learning Why As most of the students were not familiar with mobile learning or had never even heard of it a short description definition was given Afterwards they were asked if they would like to use m learning and why In free text they described their feelings and expectations rather than real impressions or knowledge Almost 60 of Italians said they would like to try m learning The most often given reasons for positive attitude to m learning is the students curiosity and willingness to use new technologies and innovations about 35 out of all positive an swers Answers like I love technology or I like anything that has to do with technology or just Why not were not rare O
50. machine and NetFront browser are provided To this date these are all Pocket PC Pocket PC 2002 and Pocket PC 2003 devices Please let me know if you try it on other than earlier men tioned mobile devices models so I can update the information here How DOES MOBILE ELDIT WORK Start the NetFront browser by clicking on its icon in the Start Menu Programs or in the recent programs NetFront3 Automatically a page with the list of texts that are currently available is displayed on the screen as shown on Figure 1 This list contains in table form the names of the texts For the Demo package the texts are in both languages German on top and Ital ian following the German texts and both difficulty levels AB advanced on the left and C intermediate on the right Clicking on a text name will lead the user to a page similar to the one shown on Figure 2 In the upper left corner there is a link which will lead to the list of texts though the user can always use the back button of the browser the blue arrow next to the Tools menu on the figures 159 APPENDIX A fig netrront 3 1 E 4 19 08 amp http www mobileeldit com T w a Mobile ELDIT BF NetFront v3 1 ef 419 11 X http http www mobileeldit com it http www mobileeldit com it comjit v a Wai a Lista dei testi La musica musica Liste Texte Lista dei testi 7 p z Non esiste musica di imma o di seconda cl
51. most frequently mentioned reason for not using m Learning is the lack of financial resources high prices of mobile communi cations and devices limited or no access to mobile networks the students mobile devices don t support new mobile technologies as GPRS EDGE etc Some students do not feel any need to use it or find m Learning is unsuitable format to present information Many feel they are unfamiliar with this technology mainly stu dents from faculties different from engineering and others think that the quality of education will decrease Only few answers are totally negative It is unnecessary I don t find any advantages and applications of m Learning and I don t like this education Q In your opinion will m learning increase the quality of in struction Why Though lots of students are curious and would like to try m learning some 57 of Italian and 27 of Bulgarian students have major doubts that the quality of instruction might increase by using small mobile devices in university education Moreover it seems students do not connect the quality of instruction with the addition of supporting services via mobile devices One of the reasons often given at the University of Trento for m learning not augmenting the quality of instruction although it might be inter esting and useful is that during lectures mobile devices generally distract people instead of helping them concentrate The students that think m learn
52. oas E tube dese a Seas geeks 112 Typical pattern for a user not preparing for the bilingualism ORAL oieves Soi coh dees te ssscs Deen scnded coves EE 112 Overlapping in users requests cecceeseeeceeeeeteeeeeeneeenes 114 Critical Set Average hoard overhead in respect to the Satisfied TeqUEStS ceceecseesceesceesceeeceeceeeeeseceeceeenseenseenaes 115 Critical Set Hoard size and error rates ce eeeeeeeeeeeee 116 Clustering of users based on requests number and spent time I EE E E ETT E tans wee Serres Abela ee deste 121 Hoarding presumable with Critical Set and after LO Prioritizing 2 cos eiiie iseer enc aE EEE 124 Transfer between desktop PC and PDA device 128 Load Times for Zipped packages cccceccessesseeteeeteeees 128 Response Time depending on the package size 129 The printing Process ccccececsseesseesceeseeeeeeeeeneenseenseeneenee 182 viii List of Tables Table 1 Table 2 Table 3 Table 4 Table 5 Table 6 Table 7 Table 8 Table 9 Table 10 Links Between LO ce edisted eat aeien coheed oes 88 Example of sessions and requested LO ecceseeeseeeeeees 92 Associations found for clusters of sesSions cee 93 Result of the automatic grouping of users based on the pequested words cenae Bei sein Bie teh ies 119 Users Similari yaen cnc E ein a aes 120 Users Similarity comparison ccceeceeseeet
53. of occurrences in the log files The idea is to create a Critical Set for every text which will contain a number of the most requested words Every time a text is prepared to be hoarded the words of its Critical Set will be included into the hoarding set independently of the fact if it is in the user s knowl edge base and is prepared for pruning In other words even if up to this moment the system believes the user knows some of these words they will be made available during the offline period An important thing to decide is how many words should be included into the Critical Set At the same time it is important to 114 keep an eye on the overhead that the Critical Set brings to the hoard By overhead we mean what fraction of the whole set of words is included into this addition to the hoard i e how bigger the final hoarding set becomes because of the inclusion of the Critical Set In other words if we decide that 10 of all words that are the most frequently used ones will be included into the Critical Set it will bring a maximum of 10 increase to the hoard In the Figure 37 below we show the number of satisfied requests as a function of the hoard overhead Words Usage Stats for the Critical Set 100 ee 90 80 70 60 50 40 30 20 10 0 Satisfied requests sx essex IL
54. of the adaptation rules by using XSLT example shown on Figure 16 2 the adaptation on the server side is a much better solution in the mobile context as the adapta tion process consumes quite a lot of computational power and does not fit well on a mobile device as the devices are limited in CPU speed operational memory and battery Figure 17a shows a screenshot of a word entry of the EL DIT system displayed in a desktop PC browser One can see that it is made out of three frames Meaning descriptive information about the selected word is shown in the left hand frame and addi tional information in the right hand frame The frame on top is dedicated to thg isa functionalities of the system famiglia la famiglia le ot am b tacsis Maissa estrineo domes sapere goetinche arg fete Persinisher mix weirraish Seccesso Lecise di sokara is questions iy BeAseTeN wis aisat suspentaveed Seapile peed pt ere slate marynaty y Areare A _Pannt npo tm caee anere a tunigta me Fane paravan varo ho seagtate noe arp de bere me doer A bet einer Faniih aetralinn farebbe deto rma di famiglia lt nants wit Vn Mitet der Fasive compliment MAGEE 1 serve perfioo de sai BehzeTet wiri De N aeteeree Figure 17 M ELDIT Content Adaptation a on the left browser view of ELDIT word P with three frames i b right top m ELDIT additional information idiomatic expressions for a word en
55. organi zation of explicit or tacit knowledge from data sources e g Web e mails chats etc Application of KM to e learning can be of vital importance in companies while in university context where most of the knowledge to be acquired by the students is explicit and formalized it can be a useful but less relevant addition Tools to support learners and tutors in managing their learning resources some systems allow different users to have their own workspace and to upload personal resources links documents notes etc or to markup learning material Common services Support of different actors students teachers tutors adminis trator and guests and integration with the company s univer 53 CHAPTER 3 RESEARCH CONTEXT sity s information systems different users typically have dif ferent levels of permissions Unregistered users guests can have some typically very limited level of access to the plat form Collaboration tools synchronous chat rooms shared applica tions whiteboards web cast audio or video conference role games simulations and asynchronous FAQ forums wikis blogs message news boards e mail mailing lists usually a few different services are offered for communication between users of the system learners lecturers tutors mentors Some of these tools are mainly meant to support cooperative work while others aim at sharing and accessing important or topical informa
56. platforms while about half of the engineers use also other platforms As the situation at the University of Rousse differs i e 44 engineering specialties use e learning much more often and a big number of non technical specialties students use more than one platform we think that the reason should be searched in the quan tity and the quality of the material offered by specific courses and programs Differences according to the owned devices PDAs As men tioned earlier we were expecting that the attitude to m learning will be strongly related with the devices owned by the provi sioned users We had looked at the answers of people who posses a PDA device and it can be noticed that more than 20 of them for Italy have tried m learning Though this statistic is based on very few participants it is obvious that the percentage is much higher than the one of people without PDAs that have used m learning From Bulgarian side no student that owns a PDA have ever used m learning The attitude to all of them though is no ticeably more positive Nevertheless their expectations to what functionality to be provided are very similar to all other students We shall mention that we were not able to study well if dif ferences can be found according to the age of the students as our participants were much concentrated in one age group namely less then 25 years old 3 1 6 Related Work Almost the same survey was done at the University of Rousse
57. reference see http ieeeltsc org Support for Learning Metadata repositories for metadata can help to catalogue learning objects and facilitate search and re use Quizzes and questions lecturers can create a pool of ques tions and answers to be used both for automatic formal exami 52 nation summative assessment or self assessment of the stu dents E learning specific services Content management services most e learning system has the notion of Course and Lecture A course can be composed by collection of resources syllabus one or many lectures a structure for describing lecture sequence forum board etc A lecture is usually composed by many resources presentation section exercise section additional material section All these components should be organized and accessed through a proper engine There could be searchable directories of courses programs etc Assessment one of the main advantages of computer supported learning is the automation of some important proc esses Self assessment is one example The pool of ques tions answers and a suitable engine allow automatic generation of different versions of tests and quizzes and also automatic checking of the results evaluation of performance and com parison with others results Knowledge management KM today most e learning sys tems do not really support knowledge management services KM in general aims at extraction summarization and
58. show that students are the top consumers of mobile content thus the best audience for the coming mobile applications and this fact should be used by the Universities At the same time practical experiences reported in 101 show that very accurate planning should be done for both devices to be used and software applications that will be needed and util ized by the students for achieving success in an m learning sys tem The students interest might be quite higher if they are sup posed to use their own devices Instead if the device is borrowed they are not that eager to invest their time in learning how to use the new media Anyway in this experiment the students were us ing PDA devices mainly for organizing their time and rarely for more closely related to the study process tools Studies on the success of real mobile learning applications in practice show a big success For example 100 shows that more than 70 of teachers that used Palms in K 12 classes feel their positive contribution and only 5 fully disagree with the statement that handhelds will help improving the quality of class room learning activities Other sources like 5 and many others described in the State of the art section of this thesis also show the good success of once implemented for a specific audience so lutions 48 3 1 7 Conclusions Here we presented the outcomes of a parallel survey made throughout Italian and Bulgarian university students tog
59. such de vices 64 3 3 Mobile ELDIT A real world system For experimenting and doing tests in the field of m learning and more concretely for studying the hoarding problem we have de veloped a real mobile learning system called Mobile ELDIT It is based on a system called ELDIT 46 whose aim is to support language learning Our system became also a proof that the gen eral architecture described earlier is a viable model as Mobile ELDIT was developed according to the principles described there Next we give more details on why we chose to develop mobile version exactly of ELDIT what part of the online system were developed for mobile and why together with more particularities and facts on both systems 3 3 1 What Description of ELDIT ELDIT is an innovative electronic language learning system es pecially designed for the needs of the population of the bilingual region South Tyrol in Italy http www eurca edu ELDIT The system can be used by anybody interested to study the Italian or German languages though its mobile version is mainly helpful for preparation for the exams in bilingualism in the mentioned area This exam must be passed by everybody who wants to work in public administration Source 46 Figure 7 Core modules of the ELDIT vocabulary acquisition system 65 CHAPTER 3 RESEARCH CONTEXT The original ELDIT consists of different modules including query engine and an intelligent tutor see Figure 7
60. the price Males are also more convinced that m learning would en hance the quality of instruction both for Italy and Bulgaria with about 10 difference comparing to female answers This might be also explained with girls feeling more comfortable with more traditional tools and media Differences according to the nationality There are noticeable dif ferences between Italian and Bulgarian students opinions only for a few parameters In some cases the differences in the answers might be due to the much lower general income of Bulgarian stu dents Though we did not directly ask the students about their in comes some deductions can be made based on their answers For example there are almost no Bulgarians that consider any of the 43 CHAPTER 3 RESEARCH CONTEXT prices low This percentage is quite often small also for Italians but the difference is sometimes up to 10 e g prices of PCs and cell phones The balance changes only when talking about the price of Internet connection where 10 more Bulgarians consider it normal This is probably due to the fact that in Bulgaria there are wired network providers and cable TV operators that often provide also quite cheap Internet Though these seem national dif ferences in our opinion the origins should be searched elsewhere Of highest importance for the students is the cost both for acquir ing the devices needed to use a certain system and the price to be paid to access its services O
61. the high prices in most of the cases thus our intention is to sup port both online and offline access to data A problem similar to the one we face off line access to data is treated in the offline browsing of web content A review of the available offline browser utilities like www avantgo com www httrack com www webstripper net etc shows that gener ally during the online periods the user selects sites that should be uploaded for later off line usage and entire sites are dumped to the local storage or the user specifies the depth of the links to be cached In situations where mobile devices are considered the memory limitations make such an approach often unfeasible as the data set might be bigger than the available space The caching problem has been studied in the general case for the Internet Wang 109 presents a survey of the state of the art techniques and elements of Web caching systems These tech 131 CHAPTER 6 RELATED WORK niques include Prediction by Partial Matching analyses of users access patterns provided by the servers prediction of the user s future Web accesses by analysing his or her past Web accesses etc Although some of these techniques are useful for predicting the content needed also in m learning domain still they aim at a different goal reduction of bandwidth consumption of access latency server workload and etc They explore the case of the Web where the search space is much bigger the
62. the mobile system to the tandem module of ELDIT and to research of the added value It would be interesting also to experiment to mix the mobile and desktop us ers 7 3 Other Research Issues Automatic Extraction of User Preferences An interesting further direction of current work would be to ana lyse if and how possible is to extract automatically user prefer ences in similar manner as extracting information about user be 142 haviour and knowledge For doing this it will be necessary to adapt the system so that the tracking data that is collected should be different i e more extended 143 Bibliography 1 2 3 4 5 6 7 8 9 10 11 Abel A Cavagnoli S amp Putzer O 1998 Zweisprachigkeit Bilin guismo Deutsche Italienische Texte Testi tedeschi italiani Arcadia Edition Abowd G D et al 1997 Cyberguide A mobile context aware tour guide ACM Wireless Networks 3 421 433 Aleahmad T amp Scotta J 2002 Integrating Handheld Technology and Web based Science Activities New Educational Opportunities World Conference on Educational Multimedia Hypermedia and Tele communications 2002 Denver CO USA Andreou A S Leonidou C Chrysostomou C Pitsillides A Samaras G Schizas C N Mavromous S M 2005 Key issues for the design and development of mobile commerce services and applications Int J Mobile C
63. two years ago 33 Interesting to mention are the changes in the an swers of the students with the passing of time On first place the percentages of devices owned by students has not changed nota bly Also students perception about the prices of Internet usage stayed almost the same Considering prices of cell phone usage the opinion that they are normal decreased while the number of answers high increased Also noticeable is the opinion that PCs and cell phones prices tilts more to normal than to high Though these are important to notice more interesting are the changes about the real usage of e learning and the attitude to m 45 CHAPTER 3 RESEARCH CONTEXT learning For both the percentages increased significantly for the e learning usage by 15 and for willingness to use m learning by about 7 Our explanation for the first change is that at the time the initial survey was done end of 2003 the Univer sity e learning platform was not loaded with learning material Only some courses were offering material online thus the system was not used a lot With more courses having published materials and updates the usage increased appreciably One can notice also the difference in the places from which students access Internet There was increased access from home probably because as men tioned before in the last few years cable TV operators started to widely provide comparatively cheap Internet Also the dra
64. which contains also the usage combinations of the word together with other words also appropriate translation and example sentences are given Another possibility is to see the words which derive from the current word or composed ones as shown on Figure 6 161 APPENDIX A On the figures below are shown the texts list available in the Demo package of Mobile ELDIT for both languages Italian on Figure 7 and German on Figure 8 One can see the group to which they belong in bold font and the name of few texts listed with bullets AJ NetFront v3 1 E q 19 07 amp http www mobileeldit com T ty amp Lingua del testo Italiano al Famiglia bambini giovent educazione Confitti in famiglia Padri e figli cuore di mamma La festa della L universit mamma dei bambini Societ Societ politica sociale politica sociale e M File View Toos 2 6 2 a Ej Fig 7 Available texts Italian E q 19 08 http www mobileeldit com T v ty a FPR Die meisten Ich lebe allein Va ter zahlen zu Kinder und wenig Karriere Deutschlands Chico Kinder geben Milliarden aus Kinder an die Macht File View Tools 2 amp S za Ej Fig 8 Available texts German INSTALLATION AND SET UP INSTRUCTIONS WHAT SHOULD BE DOWNLOADED To use Mobile ELDIT you have to download and install 1 The sample m ELDIT package from the following address http www scie
65. 002 Int Conference on Communications Internet and Information Technology US Virgin Islands Kwong Yuen Lai Zahir Tari Peter Bertok 2005 Improving Data Ac cessibility for Mobile Clients through Cooperative Hoarding In Proc of 21st International Conference on Data Engineering ICDE 05 Lamming M amp Eldridge M amp Flynn M amp Jones C amp Pendlebury D 2000 Satchel Providing Access to Any Document Any Time Any where ACM Journal of Transactions of Computer Human Interactions 7 9 pp 322 352 Lemlouma T Layaida N 2003 Adapted content delivery for differ ent contexts Proceedings of Symposium on Applications and the Inter net 2003 Liang H Xue Y Byrd T A 2003 PDA usage in healthcare profes sionals testing an extended technology acceptance model Int J Mobile Communications Vol 1 No 4 pp 372 389 Lin B Vassar J A 2004 Mobile healthcare computing devices for enterprise wide patient data delivery Int J Mobile Communications Vol 2 No 4 pp 343 353 Liu T amp Wang H amp Liang J amp Chan T amp Yang J 2002 Apply ing Wireless Technologies to Build a Highly Interactive Learning Envi ronment International Workshop on Wireless and Mobile Technologies in Education 2002 V xj Sweden pp 63 70 Liukkunen K Tolonen P Laru J 2004 Developing new mobile ser vices for the Universities University students conceptio
66. 1 Scott D J 2002 Japan s Widespread Use of Cellular Telephones to Access the Internet Implications for Educational Telecommunications 14th World Conf on Educational Multimedia Hypermedia and Tele communications 2002 Denver USA Sepp l P amp Sariola J amp Kynaslahti H 2002 Mobile Learning in Personnel Training of University Teachers JEEE International Work shop on Wireless and Mobile Technologies in Education 2002 V xj Sweden pp 23 30 Sepp l P 2002 Mobile learning and Mobility in Teacher Training IEEE International Workshop on Wireless and Mobile Technologies in Education 2002 V xj Sweden pp 130 135 Sharma S K Kitchens F L 2004 Web Services Architecture for M Learning Electronic Journal on e Learning Volume 2 Issue 1 Febru ary 2004 pp 203 216 Shen B amp Sung Ju Lee amp Basu S 2003 Performance Evaluation of Transcoding Enabled Streaming Media Caching System 4th Interna tional Conference on Mobile Data Management 2003 Melbourne Aus tralia pp 363 368 SLL 2001 Mobile Learning Explorations at the Stanford Learning Lab A newsletter for Stanford academic community Speaking of Computers Issue 55 available online last accessed September 1 03 http acomp stanford edu acpubs SOC Back_Issues SOC55 3 Slotta J D amp Clark D B amp Cheng B 2002 Integrating Palm Tech nology into WISE Inquiry Curric
67. 1 Accept application vnd wap xhtml xml application xhtml xml profile http www wapforum org xhtml text vnd wap wml image vnd wap wbmp A OS Windows CE POCKET PC Version 3 0 color color16 pixels 240x320 A CPU ARM SA1110 A Voice FALSE A Language JavaScript Accept Encoding gzip deflate User Agent Mozilla 2 0 compatible MSIE 3 02 Windows CE PPC 240x320 Host science unitn it Proxy Connection Keep Alive U U U U U U Figure 14 HTTP request from a mobile device iPAQ Pocket PC 71 CHAPTER 3 RESEARCH CONTEXT The HTTP request see the figure above contains what we need i e what kind of device is used e g Windows CE device what kind of screen it has e g 240x320 the colour resolution color16 the browser available Mozilla 2 0 etc In a more ad vanced version of Mobile ELDIT it will be possible to use other context discovery methods There are quite a lot of technological solutions nowadays for example the device independence initia tive www w3 org 2001 di In another scenario the user might receive context dependant e g location dependant language learning material like for example the system presented in 47 For such scenarios additional equipment and other methods would be necessary but it is out of the scope of this work As said before in order to keep the Mobile ELDIT users experiences close to the experiences with the online ELDIT sys tem we chose t
68. 2002 Birmingham UK pp 15 18 Garzotto F Paolini P Speroni M Proll B Retschitzegger W Schwinger W 2004 Ubiquitous access to cultural tourism portals Proceedings 15th International Workshop on Database and Expert Sys tems Applications 30 Aug 3 Sept 2004 pp 67 72 Georgiev T Georgieva E 2004 Preconditions for Using m Learning at The University of Rousse Proc of the Int e Learning Conf Brus sels Belgium 7 8 Sept 2004 pp 3 4 1 3 4 13 Goh T amp Kinshuk 2004 Getting Ready For Mobile Learning Pro ceedings of ED MEDIA 2004 World Conference on Educational Mul 147 35 36 37 38 39 40 41 42 43 44 45 46 CHAPTER 7 CONCLUSIONS AND FUTURE WORK timedia Hypermedia amp Telecommunications June 21 26 2004 Lugano Switzerland pp 56 63 Gonzalez Castafio F J amp Anido Rifon L amp Costa Montenegro E 2002 A New Transcoding Technique for PDA Browsers Based on Content Hierarchy 4th International Symposium on Mobile Human Computer Interaction 2002 Pisa Italy pp 69 80 Hand D J Mannila H Smyth P 2001 Principles of Data Mining Massachusetts Institute of Technologies 2001 ISBN 0 262 08290 X Heidemann J and Shah D 2000 Location Aware Scheduling With Minimal Infrastructure In USENIX Conference Proceedings pp 131 138 San Diego CA USC Information Sciences Institute US
69. 5 at http www mobiledia com news 35474 html Molton J 2002 Literature review in Languages Technology and Learning Nesta Fururelab Series Report 1 Murphy J amp Pathak P 2003 The changing role of the Universities in supporting E Learning amp mobility in Higher Education Proc of Ed Media 04 June 27 July 2 2005 Montreal Canada pp 1416 1419 Nord J Synnes K Parnes P 2002 An Architecture for Location Aware Applications Proc 35th Annual Hawaii International Confer ence on System Sciences HICSS 02 January 07 10 2002 Big Island Hawaii Volume 9 p 293 Oquist G amp Goldstein M 2002 Towards an Improved Readability on Mobile Devices Evaluating Adaptive Rapid Serial Visual Presenta tion 4th International Symposium on Mobile Human Computer Interac tion 2002 Pisa Italy pp 225 240 Pospischil G amp Umlauft M amp Michlmayr E 2002 Designing LoL a Mobile Tourist Guide for UMTS 4th International Sympo sium on Mobile Human Computer Interaction 2002 Pisa Italy pp 140 154 Priyantha N B Chakraborty A Balakrishnan H 2000 The Cricket Location Support System Proc 6 Annual Internat Conf on Mobile Computing and Networking Aug 6 11 2000 Boston Priyantha N B amp Chakraborty A amp Balakrishnan H 2000 The Cricket Location Support System 6th Annual Internat Conf on Mobile Computing and Networking 2000 Boston
70. AI xc srsererxrererxr sek E SE SH OMDOH OH OH OH CH OH BH OHND runnnnontrvy OH CORN DODAGAHAS 2 Critical Set size Hoard Overhead Figure 37 Critical Set Average hoard overhead in respect to the satisfied requests It is obvious that the bigger the Critical Set is the bigger the number of satisfied requests will be thus the smaller the miss _rate value will be But on the other hand this will lead to in creased size of the hoard and thus lower hit rate Nevertheless one can see that the improvement is steeper at the first few per cents of overhead because of the unequal distribution of the re quests On Figure 37 one can see that in Mobile ELDIT with only about 5 of the overhead in the hoard almost 35 of all requests will be satisfied and a 10 limit of the Critical Set will lead to satisfying about 50 of the students requests Though the initial idea for adding the Critical Set was for increasing the accuracy in terms of decreased error_rate it might 115 CHAPTER 5 CONTEXTUALIZATION AND OUTCOMES be used also for increasing the speed of the hoard size decrease It is very possible at least it turned to be the case for Mobile ELDIT that part of the words in every text is never accessed On Figure 37 above it is about 50 but it is because of the small number of users we had during the current experimentations Nevertheless we would expect to have a certain percentage of items that
71. Activity Figure 6 Context Discovery module For example for finding location different positioning systems can be used in one case the user will be outside and can use a GPS system and in another will be inside the building and will use the local network signal for that A possible solution is the in troduction of a conversion server which translates data from the format used by the device GPS WLAN etc into format proper for the service that requests the context information It is not nec essary that the system detects all possible context data at the first user request for service Some context data might be detected and provided when needed on demand Mobile Content Management and Presentation Adapta tion Currently the main service provided by e learning systems is the presentation of content The presentation of learning materials is 59 CHAPTER 3 RESEARCH CONTEXT an important issue and should be carefully designed If for exam ple the content will be accessed through a nowadays standard web browser on the PDA then it should not contain incompatible elements like scripts Adapting e learning material for a mobile scenario might imply something more than a simple reshaping of material or translating from one presentation language into an other It should be more precise and could involve different pres entation logic than in e learning Mobile Content Management The presentatio
72. E REQUIREMENTS S oara a GAEE RER 159 How DOES MOBILE ELDIT WORK cccccccccceceseeneeseee 159 INSTALLATION AND SET UP INSTRUCTIONS 162 WHAT SHOULD BE DOWNLOADED Wueeccscccccccccsssessecescceees 162 HOWTO INSTALL THE EWE VIRTUAL MACHINE 00 163 HOW TO PUT M ELDIT PACKAGE ON THE PDA 164 HOW TO SET UP USERNAMES scsesscececcccccsssssseececcceees 165 HOW TO SWITCH ON THE PROXY o oo 166 HOW TO INSTALL AND SETUP THE NETFRONT BROWSER 166 HOW TO CHANGE THE DATA PACKAGE sssececceseeeeceees 168 HOW TO GET AND SEND THE TRACKING DATA 0000000 169 HOW TO CLEAN UP THE MEMORY 170 WHAT TO DO IF 170 KNOWN PROBLEMS 172 FAQ 173 157 APPENDIX A GENERAL INFORMATION ABOUT MOBILE ELDIT Mobile ELDIT or M ELDIT is a system for studying German and or Italian languages with PDA devices It allows offline ac cess from windows based PDAs to the learning materials of ELDIT http www eurac edu ELDIT an adaptable lan guage learning platform The main design goal in the development of M ELDIT was to be used as an additional tool for the preparation for the exams in bilingualism in the South Tyrol region Though Mobile ELDIT targets mainly the users preparing for the mentioned exam it can be also used by people interested in practicing German and or Italian languages M ELDIT is a limited mobile version of the ELDIT system It allows access to the texts and associated words in both German and I
73. ENIX June 2000 Helal A Khushraj A Zhang J 2002 Incremental Hoarding and Re integration in Mobile Environments Proc Symposium on Applications and the Internet SAINT 2002 pp 8 12 Hightower J Boriello G 2001 Location Systems for Ubiquitous Computing IEEE Computer Aug 2001 57 66 available on line at http www intel research net Publications Seattle 062 120021154 45 pdf Hightower J amp Boriello G amp Want R 2000 SpotON An Indoor 3D Location Sensing Technology based on RF Signal Strength Techni cal Report 2000 02 02 2000 University of Washington Hoi K K amp Lee D L amp Xu J 2003 Document Visualization on Small Displays 4th International Conference on Mobile Data Manage ment 2003 Melbourne Australia pp 262 278 Inagaki T Kobayashi Y amp Nakagawa H 2004 Attitude Survey for pupils about Using Cellular Phones in Classrooms Proc of Ed Media 04 June 27 July 2 2005 Montreal Canada pp 1059 1065 ITU MIC Workshop on Shaping the Future Mobile Information Soci ety Seoul 4 5 March 2004 available on September 14 2005 at www itu int osg spu ni futuremobile presentations Jiang Way Lin Su Cheong Mac 2004 An Offline Scheme for Learn ing with Mobile Devices Proc of M Learn 2004 Conference Rome July 2004 Jones V amp Jo J H amp Cranitch G 2002 HyWeb A Holistic Ap proach to Technology based Tertiary Ed
74. Figure 9 Figure 10 Figure 11 Figure 12 Figure 13 Figure 14 Figure 15 Figure 16 Figure 17 Figure 18 Figure 19 Figure 20 Figure 21 Figure 22 Figure 23 Figure 24 Figure 25 Figure 26 Figure 27 Figure 28 Figure 29 Figure 30 Figure 31 Availability of devices 00 cceccceeceesceeeeeeeeeteeeeceeeenseenseenseenes 29 Opinion about devices prices 2 0 ceececseesseeseeesceseeeteeeseenees 30 Opinion about Services prices eceeceeeceeeeeseeeeceteeeeenees 31 Students preferences for mobile services 40 General M Learning Architecture cceeceeseeeeeseeeteenees 56 Context Discovery Module cccceccesseeseeeeeeseeeeesseeseensees 59 Core modules of the ELDIT vocabulary acquisition system EEE EE E E E E EEE EEE 65 Do you find ELDIT useful for preparing the bilingual exam E E R R E A AA E E E E E eee 67 What is considered mainly for understanding words MEANING 2 aa aane a a s e i N e eo ees 68 How often do you use ELDIT sssssseeesseeessseeressrsersresereee 68 For how long do you use ELDIT cece eceeseeeeeteeeteenees 68 Do you use the texts or the dictionary of ELDIT 69 Architecture of Mobile ELDIT cece ceeeeseeeteeseeneeees 70 HTTP request from a mobile device iPAQ Pocket PC 71 Low granulated raw data XML file s es 73 XSET for word entries moensis nie bea seas 73 M ELDIT Content Adaptation cceeceeceeseeeeeeteeeteeees 74 Mobi
75. INTRODUCTION wesisvesscsssdcesscesvetcisesvsstecuteescssasedstsessesiesvoossssedetvosetses 1 1 2 THE PROBLEM AND THE MOTIVATION cccecceceessececseeeeeenseeeeeees 2 1 3 CONTRIBUTIONS OF THE THESIS ccccsssscecseceeceessececseeeeeenseeeeees 5 1 4 THESIS ORGANIZATION ccccecscceessececesssececseececeesaeeecsesaeeeeneaeeenses 8 2 MOBILE LEARNING STATE OF THE ART sccccssseeees 11 2 1 DEFINING M LEARNING ccccssscessescecesssececssscececsueeeceesaeeecseaeeens 12 2 2 M LEARNING RESEARCH cccsssceesssceceessececseseececseeeeesesaeeeeseaaeens 13 2 2 1 Infrastructural Research 13 2 2 2 ACCESSING Contento eii ian e n E Bae estes 16 2 2 3 Communicating and Interacting with People 000 19 2 3 GUIDELINES FOR M LEARNING APPLICATIONS ss cccesseceeeereeees 23 3 THE RESEARCH CONTEXT eesseoeessecesesecssocecssecesssccssosecsseceese 27 3 1 SURVEY ON THE READINESS FOR MOBILE LEARNING 000ee000 27 3 1 1 General Information sisi iisirssiinone niii 28 3 1 2 Availability of devices their usage and attitude to prices 28 3 1 3 Ways of usage and attitude to e learning platforms 32 3 1 4 What about m learning neeesser 35 SDS Deductions enan costes e a costes aea a dl EE sesh 42 IKO Related Work oare e a aa Roses e a 45 Sih TF CONCLUSIONS 28 sisena A aa a a a a aea 49 3 2 GENERAL MOBILE LEARNING ARCHITECTURE s ccccsseseeeeeeeeeeees 51 SZ E learnin priera a ia s
76. MS and the mobile Learning Management System mLMS whenever possible This will allow to deliver content and other services from the eLMS to the small devices by giving the possi bility to reuse what is already available At the same time the mLMS should take account of the unique properties of the mobile learners and the mobile technology In this section we give a description of what e learning is and the services generally offered by e learning platforms We also give a description of functionalities of m learning and the problems in the transition from e to m learning which lead to the proposed general mobile learning architecture Certainly various scenarios exist in which the learning material is different from what is discussed here or a different pedagogy approach is ap plied However specific scenarios and related to them services like for example mobile communication or informal teamwork should be additional and will not be discussed here In some of those cases the proposed architecture will not apply while in oth ers those specific services should take advantage of the modules described here At the end of this section also the related work is presented 51 CHAPTER 3 RESEARCH CONTEXT 3 2 1 E learning E learning has two main facets the first is relative to using tech nology to support distance learning the second is concerned to enhance the learning experience with the help of information technology In the first ca
77. Nevertheless this simple rule can not be applied in some cases like for the last page in the session as there is no next request In such a case we were using the calculated average time of the requests in this ses sion However other rules could be used like average of the times of all requests of the current user or the averages of all requests As the different strategies did not impact much the values calcu lated for Mobile ELDIT content and this factor did not seem to have great importance up to the current stage of our experiments we did not test in depth different possible options Calculating accumulative values In order to facilitate further ex ploration we included into the database also some additional in formation extracted from the log files like for example the time spent on every single text total number of words requested from each text number of texts viewed in every session total time of the session number of interruptions of a daily session and etc Some of these values were later used in the experiments of auto matically grouping the users It should be mentioned that this step was very sensitive to the way previous steps were done In practice the real experimentations with Mobile ELDIT started in June 2004 with three mobile devices an iPaq H3800 and two Acer n10 All devices are Windows CE based Up to now we have observed 12 users for longer period of time and about 16 non Italian speaking persons partici
78. OF THE ART difficult in most cases to use complex and multimedial con tent although devices of the same size are used for entertain ment with great commercial success It should be possible to use an m learning system without reading a user manual and the experience of studying with the help of such devices should be interesting and engaging Area Domain specific content delivered just in time place The mobility should bring the ability to guideline and support students and teachers in new learning situations when and where it is necessary The dependency of the content can be relative to location context i e the system knows the location where the learner resides and adjusts to it temporal context i e the system is aware of time dependent data behavioural context i e the system monitors the activities performed by the learner and responds to them adjusting its behaviour and interest specific context i e the system modifies its behav iour according to the user s preferences It should be mentioned that as far as we are aware at the time when these guidelines were presented in the beginning of 2003 no other work in this direction was available More lately a pro found work on gathering guidelines for m learning was made in the context of MobiLearn project and presented in 106 As far as future directions are concerned we think the main re search topics could be the following e Pedagogical research is immediat
79. PhD Dissertation International Doctorate School in Information and Communication Technologies DIT University of Trento MOBILE LEARNING WIRELESS AND MOBILE TECHNOLOGIES IN EDUCATION TOWARDS HOARDING CONTENT IN M LEARNING CONTEXT Anna Trifonova Advisor Thesis Committee Prof Marco Ronchetti Prof Marco Ronchetti Universita degli Studi di Trento Universita degli Studi di Trento IT Prof Mike Sharples University of Nottingham UK Prof Kinshuk Massey University NZ Prof Giancarlo Succi Free University of Bolzano Bozen IT March 2006 Abstract M learning is a very new but rapidly expanding domain Pro voked by the fast advances of mobile technologies different appli cations and systems are developed continuously Many new re search topics are emerging in various areas including technological issues pedagogical and methodological ones problems related to content context user interfaces adaptation etc The main goal of this thesis is to address the hoarding problem which has been previously weakly explored but is a particularly important issue in the mobile domain and whose solution should be included in every system with a large quantity of data For e learning systems which are being converted adapted for access ing the content through mobile devices it is generally the case the learning material is often of big size especially compared with the locally available memory of the device Hoa
80. TER 5 CONTEXTUALIZATION AND OUTCOMES when they were familiar enough with the system to evaluate the system in terms of ease of use usage preferences etc Users could propose further improvements and discuss disturbing fac tors Some of the described problems difficulties opinions and suggestions were gathered by face to face discussions during the experiments Interesting and important points are reported further in the chapter 5 2 Automatic extraction of knowledge about the user The process for retrieving automatically knowledge about the user is shown on Figure 26 It consists of two main steps 1 preparation of the data for analysing and 2 applying different al gorithms for automatically extracting interesting knowledge User Profiles Preprocessing Transaction Automatic Knowledge Data Extraction Algorithms mie amp Integration Usage Metadat aqai Patterns Figure 26 The process of extracting knowledge about the student By data we mean the log files where the user interaction with the system is saved plus any additional data about the learning material itself about expected user behaviour known logical grouping of the users and etc In our particular case the data con sisted of the tracking data gathered by the on device proxy and the raw XML data files from which we extracted the structure of the learning material Before describing
81. This led to the fact that one of the most interesting for the mobile learning 139 CHAPTER 7 CONCLUSIONS AND FUTURE WORK community direction was left outside the scope of the current work It is the connection of the hoarding and its solutions to dif ferent pedagogical approaches and theories One example can be given with the idea to improve the study outcomes by combining hoarding with personalization In the current work depending on the experimentation phases we ei ther tried not to limit the browsing preferences of the users in any way or we were strongly limiting the possible actions In the first case we were trying to predict and satisfy every user s request leaving all possible links available and considering a hoarding miss a negative measure for our system On the other hand in certain cases it might be better from the pedagogical point of view to just not give the user the opportunity to reach to the point where a hoarding miss will appear For example the system should never show to a beginning user the information learning material that is commonly for expert In our system we were mainly leaving those judgements to the user itself For checking the real outcomes of the learning process not in terms of hoard ing more consideration should be paid on examining the acquired knowledge and comparing it with the learner s knowledge at the beginning of the mobile system usage based on feedback tests quizzes and e
82. We experiment with such requirements addressing a simple task that is often necessary in all e learning environments and that is in general taken for granted the ability to print a document This quick simple task if performed in a non mobile environment becomes less trivial when performed from a mobile device like a palmtop laptop equipped with a wireless LAN card We show that this apparently trivial task contains elements that constitute a template for other problems that can be experienced when ap proaching e learning from the mobile side 2 The Problem and Possible Solutions In few words the problem we experiment with can be expressed this way Printing on the nearest suitable printer Let s compare the non mobile and mobile case of the printing process In common one should be able to print from any application which actually uses the APIs of the operating system OS There is usually a default printer local if exists but more often the printer is on the network and the OS is managing the network interface and passes the job to the printer when needed The OS has to be able to talk the particular printer s language in order to do the task correct i e a specific printer s driver should be installed If the computer is mobile some more conditions should be added location time dependant data should be considered Based on this data and possibly to some other preferences like user s access rights o
83. ads to almost the same result even if the ses sions separation was made precisely The log file might some times be incomplete and miss the reference point for some of the links While in the first case there are no problems during the post processing and knowledge extraction phases in this second case some obvious inconsistencies often appear For example if the first request of the day is for a word it is obvious inconsis tency as the words in Mobile ELDIT are always accessed by clicking inside a text By using some heuristics the appropriate text entry should be added to the log file Identifying separate user sessions This is a very important as explained previously in Section 4 2 and turned to be not a so triv ial step It was necessary to apply some heuristics like to choose the single session time limit or the inactivity period after which the session is considered over Important to mention is that PDA devices in contrast to desktop PCs and laptops are used in quite a different manner For example they are never switched off and users generally leave all applications open thus often even after one day interruption the user was starting to work from the same point where he she suddenly was interrupted 103 CHAPTER 5 CONTEXTUALIZATION AND OUTCOMES Calculating times for which every requested URL was viewed In most of the cases this was the time difference between the time of the request and the time of the next request
84. al_Time_Seconds BIGINT 20 Q Start_Time TIME mn Number_Single_Sessions BIGINT 20 Total_Time_Seconds BIGINT 20 oser o Q Number_Texts_Per_Session BIGINT 20 a rx users ID BIGINT 20 Figure 28 Database containing the content data description and user tracking information The pre processing of the content of Mobile ELDIT consisted of parsing the XML files of each text and word earlier created by the linguists of ELDIT and from there we extracted all possible links Every file text word and link was assigned an id and the 102 tables were linked with foreign keys used afterwards for easy querying On the other hand the pre processing of the log files consisted of Generating missing history values Identifying separate user sessions Calculating times Calculating accumulative values Generating missing history values As explained earlier we used as interface to access the learning material of ELDIT a standard web browser on the PDA device which requests were captured by the local proxy and were written into the log files A particu larity of the browser is that it keeps cache of the pages already re quested in the same day It leads to the fact that in the log file the history is not full for example the pressing of the back button by the user and thus viewing once again a page is never written in the log A particularity of the usage of PDA devices explained in the next paragraph le
85. and to 0 if the user did not review it In such a way the grouping of the users over one text was independent of the grouping for the other texts We applied k means clustering details about the algorithm are available in 36 and forced 2 clusters to be produced The opti mal number of clusters might be discovered automatically how ever we preferred to force only two clusters to be created in order to be able to use the information of users words usage in our hoarding algorithm In Table 4 below we show the results of the grouping The cluster to which each user was assigned is written in each cell What is noticeable is that the classification is gener ally stable i e the users are classified quite steadily in the same cluster Note that half of the users the dark grey shaded ones shown in the first column are classified always i e for all the six texts in the same cluster Table 4 Result of the automatic grouping of users based on the re quested words User Text 1 Text 2 Text 3 Text 4 Text 5 Text 6 1 2 1 2 1 1 1 2 3 5 7 9 10 12 15 njal nins aana ninini nj ixn ninini nj ln n v ninini nj in in n v nj lninjin ivn n ninini nj In injn We can use the information extracted with the clustering algo rithm for measuring the similarity bet
86. anguage learning as lan guage learning fits well in the boundaries that we placed after analysing the work done in the mobile learning domain see 2 3 Our starting point is the ELDIT system details in the sec tion 3 3 1 developed at EURAC European Academy Bozen Bolzano Our contact with EURAC and its geographic nearness with the University of Trento allowed close collabora tion and simplified cooperation during the design of the mobile system its development and also during tracking data collection in the experimental phase and analysis of results 66 As mentioned earlier the online ELDIT contains a large quantity of learning material which is an order of magnitude lar ger than the typically available PDA memory Therefore it is ob vious that the mobile version will strongly need a hoarding sub system thus ELDIT gave us a good opportunity for our final research goal attacking the hoarding problem However we still had to limit the Mobile ELDIT to only fraction of the ELDIT sys tem as some of the functionalities of ELDIT were out of the scope of our current work mainly due to time and resource limits For example ELDIT allows the user to perform a free search for any arbitrary word in the dictionary which is unpredictable and would force uploading on the device the entire dictionary An online questionnaire has been made available in the Web based ELDIT version and some of its outcomes were very useful for deciding what pa
87. antages in the use of digital technologies in education As mentioned in our practically developed mobile learning system we used only text representations though for the hoarding we treated the material parts independently from the media An important and interesting issue is to include also dif ferent multimedia elements in such a system The main obstacle for such experiments in the current work was that preparing of multimedia learning materials is quite time consuming task and requires special artistic and or pedagogical skills Nevertheless different multimedia pictures and sounds is partially presented in the online ELDIT and its quantity is continuously growing It will be interesting to see up to what extend the existing multime dia materials could be presented to mobile users Of course im portant and motivating question to research is if adding different media will influence on the hoarding In such case we will have to add also strong rules considering material size Possible solutions might include alternative ways to present the same materials and other issues widely discussed whenever research on personaliza tion is performed It should be noted though that the addition of multimedia if not especially designed and annotated for our purposes might add additional technical problems discussed also in 14 like the one of discovering embedded objects Pedagogical Issues This thesis has in most parts a technical orientation
88. arch like Microsoft Marlin project http research microsoft com research sv Marlin Mobile Ac cess to Resources Living In NET Context is observed in variety of fields in everyday and business life that profit from the usage of mobile devices Solu tions for various scenarios are proposed like personal context storage system 81 support systems 17 77 and location aware shopping assistant 10 A survey of context aware computing and applications can be found in 23 and in 55 Nevertheless some authors like 67 and 13 discuss that a special support is needed for the mobile learning domain in order to carry its specif ics Context information for m learning and the ways it is sup ported is discussed in more details in Section 3 2 where we talk about the architecture to support m learning Usability of different mobile devices through different ac tivities is also an important issue in multiple domains Work and experiments have been done for improving input usability of the small devices 68 69 85 towards improving readability 76 displaying multimedial data 105 18 etc Location discovery can be performed with various tech niques Some systems use the Global Positioning System GPS but they work only outdoors Some indoor positioning systems offer context aware services the Active Badge System 110 and WIPS 112 Wireless Indoor Positioning System use infrared beacons Active Bat 111 uses ultrasonics the C
89. arching in the full tracking data set it is ex pected that not a lot of such associations will be found as the common scenario is to have big variety of LO and also big diver sity of students knowledge interests and learning preferences The rules extracted in this way will be of the following type LO gt LO conf 0 99 sup gt 0 5 which we can read as Almost every time when the LO was viewed by some user also LO was viewed in the same session An example can be that LO is a problem given to by the educator the students to practice the comprehen sion of certain material studied and LO the solution given also by the lecturer and linked at the end of the lecture Table 2 Example of sessions and requested LO LO1 LO2 LO3 LO4 LOS LO6 Session1 Session2 Session3 Session5 Session6 0 1 0 Session4 0 1 1 1 a O Koay ko 0 1 0 1 1 1 0 0 0 ee oC Ee FS O l ale le ole FS O l ale le Ole Session7 For the example we can pre process our tracking data the user s clicks recorded on the mobile device into the data shown in the Table 2 above Every row represents a single session not taking into account to which particular user it belongs In every cell 1 means that LO was viewed during Session not taking care of the sequencing For this data association rules algorithm will discover with confidence 1 the following rel
90. ards technological approaches in their per 42 sonal study habits and strategies Nevertheless when talking about the services that should be provided by a provisioned m learning system no insignificant preferences can be found for grouping based on this criterion Gender difference Some gender differences were noticed throughout the questionnaire The differences might be qualified mainly as a slightly more positive attitude and interest of male participants to technology in general thus to new things and ex perimentations It was noticed a 10 gender difference in the type of available Internet connection Italian students have at home Modem F 46 6 vs M 35 9 ADSL F 45 9 vs M 56 5 Our interpretation is that often males tend to acquire the newest and fastest technological solution More female partici pants have the feeling that prices of PC and Notebooks are high but on the other hand they consider more often the prices of cell phones and services as normal or low Male participants are more aware of the prices of PDAs and Smart phones In our opinion these findings might be an important factor when choosing what an m learning system should provide In the cases when a new device should be acquired we should expect more males to be in terested at initial stages while if the medium of providing the fu ture m learning services is a well known one e g cell phone and SMS females are more eager to explore it thinking less about
91. are never used even with big number of learners Figure 38 below shows how both the hoarding size and the error_rate will decrease with the help of the Critical set One can see that still the hoard size is much bigger than the real learner s usage At the same time though much less than in the first trial still errors appear Hoarding with Critical Set 100 e Used Items Error Rate With Critical Set Simple Hoarding 60 a oO x Hoard Size percentage form the full material set R wo Q dD N Q dD Lia 3 i oe ot e e iiss cio 7 z 0 5 10 15 20 25 30 35 40 45 50 Hoarding step Figure 38 Critical Set Hoard size and error rates The first effect the hoard does not become smaller than 20 is due to the way we were doing pruning In fact for this experiment we were excluding two things first as in the first experiment previously described in 5 3 1 the items that were shown to the user but he she decided not to use e g what we consider the user knowledge set contains and second we were excluding the items words of the current text that were never used by other users This percentage in the texts that we used was never bigger than 50 We would expect that it will be even smaller if bigger number of users were using the system However we c
92. arge share falls to ADSL connection 53 6 fol lowed by modem connection 38 8 and in total 93 8 of Ital ian participants use the Internet from home About 90 use the Internet also from University or work Quite a lot use access from public libraries 22 On the other hand in Bulgaria 37 3 of the students do not have Internet access at home ADSL is almost not utilized at all The Internet is used also often from work or the University 72 7 public libraries 17 4 and in this case from Internet cafes 14 3 Students report that they use Internet in wide variety of ways for searching for information 97 for studying 82 for entertainment 74 for online shopping 30 and other ac tivities Slightly more often it is used during work days and the major part of the Italian students do it for less then two hours a 31 CHAPTER 3 RESEARCH CONTEXT day while a big part of Bulgarian students report also usage of 3 7 hours a day In both cases it was noticed that more boys use Internet for longer periods A relatively small number of participants report to access the Internet from their cell phones 7 2 in IT and 14 3 in BG More than the half of the all students gave as reasons for having a cell phone that it gives them the freedom to communicate and that it makes their life easier For only 2 the cell phone is fashion Almost everybody is using the phone for conversations about 98 and to send and receive SMS 93
93. ariety of research and development issues Different parts of this multidis ciplinary process of design and implementation of a concrete mo bile learning application are discussed thoroughly We have started with a deep analysis of the mobile learning field and with the lessons learned previously in different projects We gathered the experiences of the others in a very early stage of the m learning domain and we have included them into guidelines for developing a successful mobile learning application We re ported these guidelines and used them in our further work We then checked their applicability in our own environment perform ing a survey on Italian and Bulgarian university students in order to verify the students expectations and readiness to use mobile devices for certain learning activities Interesting observations and find outs also stimulated our decisions We have also deeply analysed the similarities and differ ences between current e learning platforms and the functionalities they offer to the users and the services that should be accessible in mobile learning Based on this we proposed a general mobile learning architecture which should be able to transform all possi ble functionalities from the e learning platform and to add the new functionalities that come with the introduction of the mobile device We argued that such a general mobile LMS should sit on top of eLMS and should have three main modules one responsi ble
94. arliest initiatives in the m learning domain is the one of University of Birmingham the HandLeR project http www eee bham ac uk handler default asp The project tried to explore the lifelong learning The stress is on communica tion and on human centred systems design The main concepts they investigate are concept mapping and knowledge sharing lifelong learning wearable learning technologies and conversa tion between mobile learners Similar in some concepts to HandLeR is the project under taken at the Tampere University of Technology Finland 48 where PDAs are used for improvement of the mathematical skills of children The study content is presented in the form of a game again the idea of human centred education is explored where the pupils can communicate and help each others and the electronic device is used to measure the average students knowledge level and to adopt the speed of presenting new material to the learners To support Problem Based Learning was the aim of KNOWMOBILE project in Norway 94 where PDAs and smart phones were used for experiment in medical education of students from the School of Medicine at University of Oslo The students were put in different environments and were given different de vices some of the students were living together and had PDAs with the possibility of peer to peer connections to each other in another group students were able to connect between each other 21 CHAPTER
95. as dev Acetate ANE A A E a R e 52 3 2 2 MEL CAINS cies oiingan a a a a i a e 54 3 23 he Architecture F ea i dk cosas aae aa E a Ea i 55 S32 4 Related WOK iss a a a Eaa ea aie nE bee 60 3 3 MOBILE ELDIT A REAL WORLD SYSTEM cssscccesseseeeseeeeeeees 65 3 3 1 What Description of ELDIT ossessione 65 3 3 2 Why Motivations for the Mobile ELDIT 100008 66 3 3 3 How Details on Mobile ELDIT development 69 4 HOARDING OUTLINE OF THE SOLUTION ccsseeeees 79 4 1 MEASURING THE QUALITY c cccecesseccesssececseecececsueeecsesaeeeessaaeens 80 4 2 DEFINITION OF SESSION IN THE MOBILE LEARNING CONTEXT 83 4 3 HOARDING ON THE FIRST ACCESS TO THE SYSTEM scccceeseeees 83 vV 4 4 PREDICT THE STARTING POINT ccccsessscecececeesssssceeececeeserssaeeeeees 85 4 5 GENERATE CANDIDATE SET ccesssscecessseeecssececeessececsesaeeeenneeeenes 87 AO PRUNING ERA E EAEE inte ofits bes lice teed cattle coz 89 47 PRIORITIZING 200 2 28 secs E beovecssss tects A EA EET 91 4 8 USER MODELING e e a a a aeara eann Eaa E ETE 94 5 CONTEXTUALIZATION OF THE SOLUTION AND EXPERIMENTAL OUTCOMES cssscssssccssccessccessccessccssccessceees 97 5 1 METHODOLOGY FOR LOOKING AT THE OUTCOMES s scccceeeeees 97 5 2 AUTOMATIC EXTRACTION OF KNOWLEDGE ABOUT THE USER 98 5 2 1 Approach 1 using the online desktop System 100 5 2 2 Approach 2 using the mobile Sys
96. asse non _ possibile distinguere Sprache des Textes Deutsch con con dei limiti precisi tipi diversi di musica Eun unico mondo meraviglioso al quale bisogna avvicinarsi con curiosita_ dice il maestro Riccardo Muti che apprezza anche la musica leggera Durante un intervista egli critica chi non aiuta i giovani a sviluppare il loro talento chi non si preoccupa di creare nuove scuole e non pensa al futuro Lui che dirige gratuitamente alcune orchestre File View Took Z 6 Zag aj Fig 2 Sample Text Fie View Took 2 O ZIE m Fig 1 List of available texts The underlined words in the text are clickable and when pressed will generally lead the user to the basic word entry which con tains explanation of the word meaning the translation to the other language and one or more examples of the word usage Links to additional information are also provided at the begin ning of the sadat page as shown on Figure 3 S NetFront 3 1 19 16 xX E htt viv mobieekit com it j NetFront 3 1 http www mobieeldit com it TETEE http www mobileeldit com it com it v E Tt musica File View Tools 2 Oz a a Fig 3 Sample word entry Sense 1 File View Tools 2 Ss Ej Fig 4 Sample word entry Sense 2 160 In many cases the chosen word has more than one meaning ex ample shown on Figure 3 and Figure 4 two meanings of the word musica In such case all
97. ate from about 5 on the first step to grow up to 25 Even though we 106 have used a rather simple rule for pruning what we consider that the user knows we see that in 75 of the cases we have no misses which is quite good correctness Nevertheless at the next step we should refine the algorithm and to improve it in the fol lowing two directions 1 To make the hoard decrease faster and 2 To assure that the algorithm works more precise To make the hoard decrease faster i e in fewer steps and with bigger values we should combine the knowledge gathered from other users usage data This means that we have to analyse the similarity between the users For similar users for example simi lar in their proficiency on the studied subject we can guess that a user knows certain word based on our awareness that the other similar user or users in a group knows it In contrast with this first experiment where we considered a word familiar for the learner only when he already had the possibility to see it in a fur ther trial we will try to guess in advance On the other hand one can see from the figure that in some cases we have a big sometimes 100 miss rate We have men tioned earlier that for the mobile learning scenario the accuracy is very important This was also proven by a questionnaire that our first users filled in where almost 100 mention hoarding misses as the most disturbing problem of the mobile system A miss in t
98. ate with people and understand them better and how to build and save my self esteem She became a great friend To Prof Fausto Giunchiglia I m particularly grateful for helping me clear the initial ideas behind the current work and especially for encouraging and supporting my efforts at the beginning of the thesis Afterwards I was supervised by Prof Marco Ronchetti to whom I thank for supporting me and for helping me become an independent researcher I am thankful for the serious work of all the users of mobile ELDIT which helped me obtain the results discussed Some of my dearest friends had also participated which I appreciate highly Big 10x to Alex who was often my technical and programming tutor and help Grazie to Angela who gave me some very useful hints in the very beginning and never stopped smiling During these four years I have found lots of friends that filled in my days in and outside University of Trento I should mention the closest of them my Bulgarian friends Galia Vanya and Nelly my ICT colleagues Arianna Csaba Navrati the Bielorussian iii cluster with which I had great fun Andrey Roma Ilia and many others like Manuel Mher and Lilit Sara and Cesco I do hope that even when our roads split we will still remain in touch And a special thanks to Mitashki Dolly and Dani for making my life bright and giving me strength by being next to me March 2006 Trento Italy Table of Contents 1
99. ation continues in the m learning domain As previously one can discuss the adaptation of the content for the concrete user but in m learning the adaptation is needed also towards the device that is used and to the surrounded environment It should be men tioned that the adaptation should be ensured on an architectural level so apart of some references given here more details in this direction will be given in section 3 2 4 on page 60 In 34 a sys tem is presented that utilizes a multi dimensional framework to support adaptation The adaptation both to the user and to the de vice is discussed in detail in 50 and shows that modelling of the user is a very important step and in most cases rules should be mapped to the full list combinations of parameters describing 18 the users learning styles A unified approach to educational con tent adaptation for mobile device is proposed in 103 These and many other sources like 51 116 etc also suggest that adapta tion in mobile learning is very important and opens many techno logical and pedagogical issues For supporting adaptation additional information might be needed specific for the mobile learning context For this reason a possible need would be an analogue of e learning metadata as proposed in 13 for extension of the metadata standard for the needs of the mobile scenarios It will comprise not only the data about the learning material itself but also about the learne
100. ations LO gt LO LO gt LO k LO gt LO LO gt LO LO gt LO 92 Association rules can be discovered also in more limited number of sessions not all at a time For example one might search for correlated objects only in the sessions of users that were classified in the same group of content interest or field expertise Consider ing again the example data in Table 2 if we apply a clustering al gorithm like k means see again 36 the algorithm will pro duce 2 clusters We marked the rows that would be in different clusters with different shades intensity in Table 2 clusterg in grey and cluster in white and we represent them separately in the first two columns of Table 3 Applying association rules only to the sessions in the same cluster we get some additional associa tions The clusters and discovered associations are as follows Table 3 Associations found for clusters of sessions Cluster Instances Additional Associations Cluster y Session LO gt LO Session LO gt LO co LO gt LOs ession T LO gt LOs Cluster Session LO gt LO Sessions LO gt LO Sessions The above associations like LO LOs show that if LO is to be selected for the hoarding set there is big probability that the user will also be accessing the object LO during the same ses sion Moreover associations of the type LO s 1 amp LOg 1 gt LO 0 can also be discovered showing tha
101. be removed from the device After finishing with the installation of the virtual machine go to the Programs folder Ewe Launcher VM Version 1 43 on your device and open the EWE folder Michael L Brereton Click on the ewe icon to activate the VM You see a screen similar to the one shown on the right www ewesoft com Freeware Version The launcher is loading For more detailed instructions and support see EweSoft web site http www ewesoft com 163 APPENDIX A How TO PUT M ELDIT PACKAGE ON THE PDA The MobileELDIT zip file you have downloaded contains all the data of the demo package of Mobile ELDIT system 1 Unpack MobileELDIT zip in some temporary directory on your PC e g C Temp From the zip one directory called ewe no quotes should be extracted 2 On the mobile device using File Explorer put the folder ewe no quotes in the folder Program Files The result is On the PDA Program Files ewe 3 Now on the mobile device you should have ewe folder that contains a folder m ELDIT and a file FoxyProxy ewe like shown below aloj CEC 01 x fe Eat yew Favores Tol tib CEEA EEE GQsearch GyFolders Back gt Qsearch Gyrolders ladies fe 3 Program Files ewe bz E eo t adress fa Program Files ewe mEidt z O60 7 19MB Winzip File 469 byte File E Storage_Name txt Sbyte Text Document If you have an external memory on your PDA it is preferable
102. bed previously in the section 3 2 General mobile learning architecture i e the separate modular support for the three func tionalities that are important for a mobile learning system Con text Discovery Specific Adaptation and Packaging and Syn chronization We wanted to keep the user experiences during the use of the mobile version as close as possible to the online ver 69 CHAPTER 3 RESEARCH CONTEXT sion so we used a web browser on the mobile device as interface to the Mobile ELDIT Web browsing is already very familiar to almost every user therefore it is not necessary to learn yet another user interface This makes the system very easy for the user to get used to and after few clicks the user feels already familiar with it Web Interface Eldit Server aS Server Logs User model Text corpus Word entries Redesigned Content Tracking Data User Interface Web Browser Figure 13 Architecture of Mobile ELDIT Mobile ELDIT consists of two main parts 1 server side which we call the m ELDIT server and 2 client side a proxy that serves to respond to browser requests during disconnected periods by providing the pages that are already in the cache and collecting the tracking information into log files see Figure 13 Basically the log file is a list of user s requests for learning materials to gether with time information These log files are the main source of infor
103. ble when needed When the seminar finishes and the user gets to a zone with a local net work connection his notes are uploaded to the server by the Packaging and Synchronization module together with some tracking information The tracking data will be used in the future for user modelling and decisions what materials are needed dur ing offline periods to this and other users On the other hand the notes are made available for further access online Printing at home is done easily by accessing the University s e learning plat form On the next day when the student wants to print the notes at the Faculty building the Context Discovery is once again trig 58 gered and the student s location is discovered The system pro poses the closest printer where the student is allowed to print based on his access rights and is given instructions on how to get there from his current position Context Discovery This module adds an abstraction that can hide the details about the different physical methods of context discovery By context as shown on Figure 6 we mean identity spatial information i e physical location temporal information environmental informa tion e g noise level availability of resources i e battery dis play network and bandwidth etc Context Discovery User identity Location Temporal Information Environmental Information Nearby Resources Availability of Resources Infrastructure
104. can not be displayed solution Nothing to be done Just click on the back button and continue your study I hope you will not meet this situation often e A FALL DOWN MENU APPEARS symptom You can not follow the links because a menu ap pears every time you try to as shown on the picture probable reason when you click and hold on touch screen the sys tem shows the context menu as a right mouse button click on a PC _ computer Pat e solution Click somewhere outside _ pattiessa des To einen Teil the menu so that it disappears v Show Images a Die Summe die der Val v Enable Animations je nach seinem Einkor For following thig links you need preh z to do a short click not to hold Wrap Content down and wait for few seconds Erauen die finanzielle Proererr for the page to be displayed While the browser shows a ro Mra ted NetFront v3 1 ef dx 13 24 amp Ser http www mobipeldit cor id Wai a Lista dei testi New Window Die meisten V ful screen a File View Tok 2 S O ZIE Ej 171 http www mobileeldit com d v ala APPENDIX A tating globe next to the address field see the figure below it downloads the page so you have to be patient e ADDITIONAL TABS IN THE BROWSER APPEAR symptom You see multiple tabs in the browser window and you do not find the learning content anymore as on the picture E
105. coding servers or proxies are often used for ad aptation of content see e g 62 which is retrieved by the server together with the client preferences and constraints Negotiation is done between the client and the server about the needed adapta tions Finally the converted content is delivered Different transcoding techniques can be used for translating from one pres entation language to another e g WAP HTML WAP for reduc ing the contents size for satisfying bandwidth or screen capabili ties for adapting the structure of the content etc What is missing here is that generally only online access to the content is consid ered Only some of the transcoding proxies take care for caching web pages for offline usage e g AvantGo Another point to con sider is that in the learning scenario the content that is to be deliv ered could be sometimes quite large We think that delivering content for offline usage is an important issue as still mobile de vices are often disconnected because of the lack of access in cer 63 CHAPTER 3 RESEARCH CONTEXT tain places but also because of the high prices in most of the cases thus our intention is to support both online and offline ac cess to data The off line access to data is treated in the offline browsing of web content The typical pre fetching solutions offered by off line browser utilities cannot be cast to mobile domain without taking into account the severe memory limitations of
106. ction to be done gave the same precision 5 3 5 Association Rules In section 4 7 we discussed the importance of prioritizing the LO that are selected for hoarding A possible technique to be used is the utilization of automatically discovered associations as rules for increasing or decreasing the priority for the LO of the Candi date set For example the following rules Table 9 are discov ered over all users requests on one of the examined texts Table 9 Association Rules all users Text 4 Conf Supp Supp Supp Antecedent a Consequent c a c aUo 1 100 it n ambiente 1 lemma gt it n camicia 1 derivati 3 6 3 it v mollare 1 lemma oath ge 2 100 tvstiraro lemmas it n gancio 1 derivati 3 10 3 3 100 itv rendere 1 lemma gt _ it n gancio 1 derivati 3 10 3 4 100 it v stirare 1 lemma gt it n gancio 1 derivati 4 10 4 As one can see from the table above even in a quite small tracking data set quite strong rules can be found In the example given here we required the confidence to be 100 i e should be true for every antecedent and support gt 18 i e three or four out of 16 users requested the words a and in all cases a request also of c was made Table 10 Association Rules cluster 2 users Text 4 rear Antecedent a Consequent c Rea RE n i a 1 100 it n ambiente 1 lemma gt it n camicia 1 derivati 3 4 3 2
107. d input capabilities of the mobile devices make it difficult to use rich e g complex and multimedial documents using a PDA like interface it is therefore not useful or practical to transpose a power point presentation on a PDA Context dependent information the dependency can be rela tive to location context i e the system knows the location where the learner resides and adjusts to it temporal context i e the system is aware of time dependent data behavioral context i e the system monitors the activities performed by the learner and responds to them adjusting its behaviour interest specific context i e the system modifies its behav iour according to the user s preferences Examples of context dependent systems although not related to m learning are Tourist information systems like GUIDE 17 and CYBER GUIDE 2 these systems offer information to tourists tak ing into account their current location Context aware messaging systems that trigger actions accord ing to a specific context like the ComMotion system 70 which links personal information to locations and generates events e g sound or message boxes when a user moves to a certain location Other such systems are CyberMinder 24 and lcron 37 they allow the user to define more complex conditions like time and location dependent conditions General utilities like Friend finder GeoNotes BusLo cator 75 178
108. de of the infrastructure providing the service a possibly prohibitive task In such cases one can fall back to a less convenient two step process through the notion of a stub one would then use a local instance of the needed service i e one might have an actor on the server that asks for a local service on behalf of a remote mobile user 185 Appendix C List of publications A major part of the work for this thesis has been published in peer reviewed journals conferences and workshops in the area of computer supported learning and e learning Here the list of the papers is given ordered by publication date Trifonova A Ronchetti M 2006 Hoarding Content for Mobile Learning International Journal of Mobile Communications IJMC Vol 4 No 4 pp 459 476 e Trifonova A Georgieva E Ronchetti M 2006 Has the Time for Univer sity s Mobile Learning Come Determining Students Readiness submitted in Journal of Educational Technology amp Society endorsed by IEEE Technical Committee on Learning Technology e Kennedy I G Fallahkhair S Fraser R Ismail A Rossano V amp Trifono va A 2005 A Simple Web based Adaptive Educational System SWAES to appear in Special issue on Modeling and Simulation of International Journal of Technology Instruction Cognition and Learning TICL e Trifonova A Knapp J Ronchetti M 2005 E learning versus M learning Experiences a Prototype and First Experimental
109. ded pruned As a next step for improving the hoarding algo rithm we decided to concentrate on the aim to minimize the miss_rate As mentioned before every miss in the hoard might be critical for the users understanding of the studied material and thus the low miss _rate is probably the most important factor for the hoarding process The addition of time measurements shown on Figure 30 in the hoarding algorithm as a guarantee for real review of the material was a simple step that helped excluding some of the misses the surrounded ones on Figure 29 However still lots of misses were appearing It was also mentioned that in ELDIT words are always linked to their infinitive form thus that even different forms of the same word are requested by the same URL For example all derivations of the word sentimento like sentimentale or sen sazionale are requested and thus saved in the log files by the link it n sentimento 1 derivati Nevertheless in certain texts a word and its unusual form or conjugation might be very important for the understanding of the text This means that certain words might be critical for these texts for users understanding or for answering the comprehension questions Such critical words might be included into a Critical Set which we will use for im proving the hoarding performance Our supposition is that such words will appear very often in the requests i e a large number of t
110. ding set or in prioritiz ing the LO On the other hand the user knowledge profile should consist of everything that the system knows about what the user already knows Example is the system awareness of the user s compe tence in a certain subject i e beginner intermediate advanced or a list of all the topics already covered by the user previously Users can be also grouped based on their knowledge but in con trast to the user behaviour the profile of the user knowledge will be mainly used for pruning the entries from the hoarding set i e for excluding objects in order to decrease the size of the hoard We can distinguish static data about the user and dynami cally changing data The static data include for example the user age gender mother tongue and etc On the other hand the dy namic data is our current knowledge about the changeable over time user parameters and should be reviewed in certain periods of time For example the user browsing pattern might change drasti cally few days before an exam date thus the hoarding system should be able to quickly recognize such changes and react ac cordingly 95 Chapter 5 5 Contextualization of the Solution and Experimental Outcomes The previous section gave the outline of the strategy for solving the hoarding problem in a general level outlining some of the techniques that are possible to be used in concrete implementa tions What we describe till now gives more abstract vie
111. e time of ELDIT Our goal was later to compare the usage of the online desktop and the mobile offline systems having in mind that the mobile ELDIT complements the main system Figure 10 How often do you Figure 11 For how long do you use ELDIT use ELDIT 68 Our preliminary conclusions were that language learning is a good choice as a field of the use of mobile devices The ELDIT system and especially its text corpus is especially suitable for experiments on our hoarding problem As people have different learning styles our expectation was that some should use the sys tem to study in small gaps of waiting time while others will pre fer using the PDA just as an electronic dictionary available any time How these habits and stiles should influence the hoarding for a mobile learning system will be discussed further the dictionary 42 both Figure 12 Do you use the texts or the dictionary of ELDIT On the other hand on Figure 12 one can see that ELDIT texts are used by more than 50 of the system users which gave us the certitude that though the mobile ELDIT will contain only part of the original system it will be useful for the users 3 3 3 How Details on Mobile ELDIT development In the developing the mobile version of ELDIT we followed the guidelines we extracted from the literature see Section 2 3 for small and simple modules and also applied the principles de scri
112. e and topical Different learning approaches involving mobile devices should be considered and observed to find which ones are the most effective given the conditions in which m learning hap pens e The lack of convenient input tools pushes the research in exploration of new forms of user interfaces for example sound or mobile scanning tools as input output e The small screens of the existing mobile devices give many research opportunities Digital materials used in e learning should be at least partially reused but a specific adaptation is required for them to serve m learning needs 24 The best way to do the adaptation would be of course automatic customization conversion In a general sense re search in the area of device independent presentation of data serves also other domains but an investigation is needed to find out which are the special requirements of m learning e Related to e learning are the services that students and teachers need and that are typically provided by Learning Management Systems Providing such services via mobile devices is an applied research direction e E learning always depends on the connectivity of the end user With the mobile devices there are periods of poor connectivity or no connectivity at all M learning could therefore be delivered in three different ways pure con nection pure mobility and mixture of the previous two Pure connection is when the mobile device is
113. e can easily construct the candidate set for every next step level of hoarding Later this candidate set will be pruned its size can be decreased by dropping some of the objects that are not likely to be requested Table 1 Links between LO LO LO LO ees LO LO x 1 0 1 LO 0 x 1 1 LO 1 1 x 0 oor x LO 1 0 1 x 88 The generation of the candidate set should start from the starting point predicted for the next offline session It should generate a candidate set of the LOs connected to this point and afterwards should be followed by pruning of those candidates When the pruning of this depth level 1 candidate set is finished a candidate set should be generated for every LO that is still in the set thus going one level deeper Again pruning should be done on the newly generated candidate set and the cycling procedure should stop when the estimated user s browsing depth is reached 4 6 Pruning Pruning is the step when the hoarding system decides if a LO is probable to be seen by the user or not and in this latter case ex cludes this material from the hoard This should be considered the most important together with prioritizing step and at the same time the most fragile one in the hoarding process An alternative to the pruning might be a prediction of the exact path that the user will be following but in a real system unless a very strict follow ing of the learnin
114. e category we might include accessing services that can be seen as dynamically generated content The functions that are offered by such devices are therefore not different from what can be done with other de vices in the same way as mobile telephony is not intrinsically dif ferent from residential telephony but the change of boundary conditions induces a new use of the media Also the different in terface that such instruments have small screen small or no key board has an impact on what is reasonable useful and even pleasant to do on such devices For instance reading a digitalized book on a Palm is today barely acceptable and reading it on a cell phone is probably unacceptable for most people Even browsing the Internet is an experience not comparable with doing it on a PC So while some research concentrates on how to best per form the same action in a changed environment some other fo cuses on what actions are best suited to new conditions On this 12 last aspect the ability to contextualize i e to take into account where the user is in space and time and what the user is doing in order to propose the best suited activity is a big challenge that goes under many names the most popular of which are ambient intelligence and ubiquitous pervasive or sentient computing Research on pedagogical use of the new media is a wide open field On the more technological side infrastructural re search on mobile computing is of g
115. e for bigger packages the reply time increases We should mention that it depends strongly on the percentage of misses because of the particu larity of a slow reply when a miss is encountered For the fig ure we used 20 of miss rate which would probably be too high percentage in a real hoarding system and was slightly above the average in the experiments described previously Replay delay 45 3 5 25 Delay Time Seconds 15 AmS 0 5 4 6 8 10 12 14 16 Zip Size MB e Nonisses o All 23 misses t Onlylmisses m No Zip Figure 43 Response Time depending on the package size Though this is not really a problem but rather a fact that should be taken into consideration we would like to share another observation which we found important Generally us ers expectations for the PDA system functionalities and speed of work are much higher than the one available nowa days Except from the people that had used PDA devices be fore all our users were complaining that they expect the de vice to behave more similar to the desktop PC Our experiments were done with PDA devices that were es pecially bought for this purpose as none of the users that were eager to participate owned one However we discovered 129 CHAPTER 5 EXPERIMENTAL OUTCOMES that generally with small exceptions they were not willingly exploring any add
116. e g user behaviour user knowledge and user preferences Nevertheless extraction of knowledge about user preferences is out of the scope of the current work The pre processing of the raw data on one hand is the proc ess where the log files and all other available data should be parsed and integrated into a database or other suitable format to perform knowledge extraction algorithms Generally the pre processing is one of the most resource consuming processes but in the context of analysing user behaviour in a mobile learning system this part might be on the server and will most likely be performed during the offline periods of the user This means that in our context this is not a critical point e g even quite slow speed of the pre processing and extraction of the data will not lead to user s impression of a slow system As we previously discussed the optimal way for developing a mobile learning system is to make it sit on top of an e learning solution in such a way that to take advantage of what is provided there already In section 3 3 3 we showed that our Mobile ELDIT uses the same XML data with the online ELDIT substituting on 99 CHAPTER 5 CONTEXTUALIZATION AND OUTCOMES the server side the Content Redesign Engine with a specific one for the mobile devices this was one of the general principles de scribed in Section 3 2 Further one of the very important func tionalities for the hoarding process of the m LMS is the po
117. e learning context Discussed approaches though assume always that the hoarding engine will know in advance which are the devices that will form the ad hoc network and use this information to fill in the caches On the other hand another possible approach would be just to use these networks to try to satisfy the hoarding misses 7 2 Mobile ELDIT Improvements Include simple dictionary As described the learning content of ELDIT is very elaborate and continuously growing For every word in ELDIT are provided explanations and translation a num ber of examples idiomatic expressions derivations etc grouped by concrete meaning Yet in Mobile ELDIT only a portion is loaded for offline usage due to memory limitations The main goal of this thesis was to research on the possible techniques to solve the hoarding problem i e to satisfy all user requests How ever in certain cases the user might like to access an arbitrary word not following the links that we have provided from the texts or other words The student might want to check the spelling of a synonym or to see the meaning of a word which he she hears while on the street In the current version of Mobile ELDIT such an arbitrary search is not provided as a functionality However a possible and welcome solution will be the addition of a simpler dictionary the entries of which can be shown to the user also in 141 CHAPTER 7 CONCLUSIONS AND FUTURE WORK the cases of hoarding miss Such
118. eden Divitini M amp Haugalokken O K amp Norevik P 2002 Improving communication through mobile technologies Which possibilities In ternational Workshop on Wireless and Mobile Technologies in Educa tion 2002 V xj Sweden pp 86 90 Feng Ming Whei amp Su Cheong Mac 2004 The Application of Infor mation and Communication Technologies on Digital Learning to Reduce the Digital Divide Journal of Cyber Culture and Information Society 7 pp 87 110 available online on 04 November 2005 at http www nhu edu tw society jccic 07 fu 7 05 pdf Figg C amp Burston J 2002 PDA Strategies for Preservice Teacher Technology Training 14th World Conference on Educational Multime dia Hypermedia and Telecommunications 2002 Denver CO USA Flynn M amp Pendlebury D amp Jones C amp Eldridge M amp Lamming M 2000 The Satchel System Architecture Mobile Access to Documents and Services ACM Journal of Mobile Networks and Applications 5 4 pp 243 258 Gamper J Knapp J 2003 A Data Model and its Implementation for a Web Based Language Learning System Proc Twelfth International World Wide Web Conference WWW2003 Budapest Hungary May 20 24 2003 Garner I amp Francis J amp Wales K 2002 An Evaluation of the Im plementation of a Short Messaging System SMS to Support Under graduate Students European Workshop on Mobile and Contextual Learning
119. eeseeeteeneees 120 k Nearest Neighbours Prediction value of k 00 122 k Nearest Neighbours Prediction correctness 0 122 Association Rules all users Text 4 ccceseeseeseeeeeetees 123 Association Rules cluster 2 users Text 4 cesses 123 ix Chapter 1 1 Introduction E learning is growing very fast and many Universities and com panies are already supporting in some way an e learning solution There is now little doubt that the World Wide Web is a very suc cessful educational medium On the other hand wireless and mo bile technologies have been developing very fast over the last few years New devices and technological solutions appear on the market with great speed and the research and development com munities are trying to find the best possible ways to use them Small relatively inexpensive devices like PDA Personal Digital Assistant smart phones and even the common nowadays cell phones with already reasonably powerful characteristics enable computational and data access while on the move As a conse quence mobile applications are appearing in different fields like commerce 4 healthcare 63 64 tourism 32 etc The rush in the field of wireless and mobile technologies creates opportunity for new field of research also in the learning domain Though the use of mobile devices for educational pur poses was explored for the first time quite long time ago in early 1970s in the Dynaboo
120. em ory making the rest work much slower and even making pro grams crash As our experiment was connected with offline delivery of ma terial we introduced a client side proxy that should simulate Internet access even in offline periods The problem that ap peared was that Internet Explorer the browser available on windows based mobile devices does not send requests to the local proxy if it does not find an Internet connection by itself This made it necessary to use another web browser Unfortu nately all other browsers at that moment were commercial products Very recently a free Mozilla browser for Pocket PC has been developed However at present its early version still does not fit to our requirements Problems were found also considering the presentation in the browser we have used Special German and Italian letters are 127 CHAPTER 5 CONTEXTUALIZATION AND OUTCOMES not always presented correctly thus the browser should be chosen carefully in advance v The first version of our prototype works with a large number of small files several thousands We have observed that the file transfer from the desktop PC to the mobile device is a very slow operation when lots of files are being copied e g about 5 10 times slower for transferring small files compar ing with transferring one big file This means that packaging is strongly needed The process of deleting big quantity of small files is also very slow v The use of packagi
121. er 2004 Publication of IEEE Computer So ciety Technical Committee on Learning Technology TCLT pp 78 83 Trifonova A Ronchetti M 2003 Where is Mobile Learning Going Proc of the World Conference on E learning in Corporate Government Health care amp Higher Education E Learn 03 Phoenix AZ USA Nov 7 11 2003 pp 1794 1801 Colazzo L Molinari A Ronchetti M Trifonova A 2003 Towards a Multi Vendor Mobile Learning Management System Proc of the World Con ference on E learning in Corporate Government Healthcare amp Higher Edu cation E learn 03 Phoenix AZ USA Nov 7 11 2003 pp 2097 2100 Alfio Andronico Antonella Carbonaro Luigi Colazzo Andrea Molinari Marco Ronchetti and Anna Trifonova 2003 Designing Models and Services for Learning Management Systems in Mobile Settings Proc of Mobile and Ubiquitous Information Access Mobile HCI 2003 International Workshop Edited by Fabio Crestani Mark Dunlop Stefano Mizzaro Udine Italy Sep tember 8 2003 Springer LNCS vol 2954 2004 pp 90 106 Trifonova A Ronchetti M 2003 Context Dependent Services in an M Learning Environment the Printing Case Proc of IADIS International Con ference e Society 2003 Edited by Palma Dos Reis A Isaias P Lisboa Por tugal 3 6 June 2003 pp 635 638 188
122. erent printer a black and white printer if the user is reading a text only document or a colour printer if the user is looking at pictures Preferences might involve opting by default for cheaper services at expenses of print quality or vice versa Also the knowledge about who the user is is important there might be different restrictions on printers usage for teachers and for students or limit over the number of printed pages The second step is to choose the resource printer that best suits user s needs taking in account the context information Therefore the location info and characteristics of all printers must be known to the party that takes the decision As mentioned be fore the choice could be made locally or on the server If we consider a local case then all the information needed for the decision making should be stored on the device In a lim ited mobility the data might not be massive and might not endan ger the availability of the device s memory One might also imag ine that when moving to a different environment e g to another building the mobile device could discard all the info regarding the previous environment and download the info relative to the present surroundings In this way though the system might omit some real time info like the printers queues at the current mo ment thus producing poor solution it is probably better to walk a few more steps to an empty printer that quickly reaching a b
123. ether with a study of similar surveys done by other bodies The goal was to discover important parameters that might influence the success of a provisioned mobile learning system These parameters should be closely observed and cautiously considered in design phase Our findings show that students attitude is strongly related to the ways and frequencies of usage of e learning Students that are comfortable and use willingly e learning are much more posi tive to m learning A very important factor for a successful m learning solution will be the choice of devices to be used Generally students opin ion is that except cell phones small mobile devices and their ex ploitation are very expensive Most of them will not be eager to buy device unless they see very strong positive ways of its usage Mobile phones that have much increased recently resources seem to be the best choice in University environment though the wide variety of models might be a problem Nevertheless the price of the service is always considered by the students The general attitude to technology is also a strong factor Males tend to be more interested to experimenting and trying new things while females often prefer traditional approaches How ever once accustomed to a certain media type girls tend to use it more often This should lead to the expectation that in a newly developed m learning system the initial users will be more male We could not find any difference
124. ex pense of the answers giving as reason the price and the limitations of the devices which are respectively 35 8 2 40 5 12 45 5 16 When asked about the expected relation between m learning and the quality of university education most negative as might be expected are the students that do not use e learning more than 60 negative From the people that use only Univer sity s platform 56 are negative and from those that use more than one about 50 None of those who ever used m learning consider it to be without a future It can be noted the trend of 5 difference in the number of people that have tried an m learning solution from the people that do not use e learning the percentage is close to zero from people that use only their own university s platform it is about 5 and between those who use more platform it becomes about 10 The difference in the attitude to try m learning is even bigger about 10 Positive are about 50 of the first group slightly more than 60 of the second and almost 70 in the third group from the Italian students Respectively the numbers for Bulgarian students are about 70 80 and 90 In brief the students attitude to mobile learning seems to be closely related with the students ways of usage of e learning the more they use e learning more positive they are to the next e learning step In our opinion the roots of this are deeper in the students feelings tow
125. f course a hit rate lower than 100 would be acceptable as long as the miss rate remains at 0 it would only imply a sub 81 CHAPTER 4 SOLUTION OUTLINE optimal usage of the available resources i e a waste in memory without affecting the perceived system performance Set of LO Set of LO selected by the used by the student in one session LE Figure 20 The expected picture Though the ideal picture Figure 19 above is to select all and only those items that will be used by the user it is obvious that in a real system such an ideal situation is almost impossible to reach Most probably we will have some desirably big overlapping be tween the cached by the hoarding algorithm LO and those LO really requested by the learner see Figure 20 As mentioned before the hoarding sub system should be able to analyse how successful was the previous hoarding and improve its further predictions For this we should be able to check which parameters or combinations of parameters of the user model and or domain knowledge have bigger impact on the goodness of the algorithm By analysing the goodness of the prediction of the hoarding algorithm we can try to tune its work For example if a user indi cates a LO miss as fatal the algorithm should check why this LO was not cached e g if this entry was pruned or was given a small priority and later the rules for pruning and or prioritizing should be reconsidered accordingly This is act
126. f knowledge about the learner important for the hoard ing 4 1 Measuring the quality An important point is to measure the quality of the hoarding and to try to improve it continuously An often used metric in the evaluation of caching proxies is the hit ratio Hit ratio is calcu lated by dividing the number of hits i e found LOs by the total number of uploaded predictions cache size It is a good measure for hoarding systems though a better measure is the miss ratio a percentage of accesses for which the cache is ineffective Kuen ning and Popek 57 defined a miss cost as a main difference in the evaluation of a caching and a hoarding system In cach ing pre fetching systems the misses in the prediction reflect as a time penalty as the missing content should be retrieved from the web This differs from the mobile case where with unavailable Internet connection a miss in the hoard might be fatal In order to quantify this measure it is possible to demand a user rating on every miss using some different impact values In some cases of the learning scenario this technique has little sense because it might be doubtful if we can trust the user s opinion about his her 80 own knowledge and expertise and most probably every requested learning material is in fact important for the study process In 57 is also defined time to first miss measure a simple count between the start of the disconnected operation and the first hoard miss
127. ffline period Steps 1 and 3 show the browser request for a page that is captured by the proxy and after finding what was requested in its own cache the proxy sends back the response On the other hand step 2 shows what happens when the requested page is not in the cache the proxy sends a request to the Mobile ELDIT server 2b which on its side gets the raw data 2c from ELDIT the XML files redesigns them and sends the response back to the proxy 2a Step 7 shows that the proxy might decide to contact the server and to update the content of its cache during the online period even when there is no request from the browser At the same time the tracking data might be also send to the m LMS During all offline periods 76 steps 4 6 even if no cache entry is found the proxy responds to the requests with a meaningful message In our real experimentations with Mobile ELDIT the de vices were always offline as the PDA devices we utilized did not have wireless connection However the implementation of the system supports the step 2b as shown on the figure above How ever steps 2c and 7b were never required as we were keeping a copy of the raw XML files of ELDIT locally on the m LMS This was done for facilitating the experimentations but the scenario described is easily realizable The Mobile ELDIT system described here was used by a dozen of self motivated users part of which were preparing for the exam of bilingualism and part tha
128. g materials conform able to the resources of mobile devices regularly to send infor mation via SMS MMS about news or changes to ensure abilities for the students do download and read off line files on mobile de vices to present briefly and clear the information on subjects to ensure more close and fast connection between students and teachers to ensure active connections with other e learning sys tems Finally more than 66 of the students see the future of m learning as a support system for the traditional forms of instruc tion The percentage of the students which consider m learning can work as an autonomous system is almost equal to the percent age of those that think m learning has no perspective a bit less then 10 About 15 do not have opinion about its future 41 CHAPTER 3 RESEARCH CONTEXT 3 1 5 Deductions Differences according to e learning usage We have studied if the ways students use e learning affects their answers about the m learning and we noticed that it does On the first place there is a difference in the attitude to hav ing a mobile device other than the cell phone A noticeable trend in the answers of Italian students is that the percentage of A PDA device is not useful for me answers is related to the e learning usage 50 for those who do not use e learning 40 to those who use only the University s platform and about 30 for those using multiple e learning systems This happens at the
129. g sequence is required by the educator this would be almost impossible Student s Knowledge Knowledge Base Figure 24 User knowledge as a subset of the knowledge base Pruning should be done of LOs that are not interesting for the user and of those that the user already knows have mastered One possible schema is to determine the user knowledge by assessing him her at the beginning of the learning with the system For the 89 CHAPTER 4 SOLUTION OUTLINE well defined questionnaire the system might determine with a that provided by the system knowledge base see Figure 24 By a good accuracy the user knowledge set purpose of the algorithm the user knowledge is always a subset of If the system does not provide any initial assessment then the user knowledge set is empty at the beginning Nevertheless the goal of any educational tool is to increase the students knowledge over the provided knowledge base so in general the set representing the user knowledge should be dynamic continu ously growing If some particular exception is not determined important to correctly determine what subset of all knowledge base of the student At this point is already clear that it is very base is the student s knowledge then the system should prune the LOs that are in the knowledge In our Mobile ELDIT system we decided not to test the user knowledge at the beginning but rather to try to automatically We did pru
130. ge Additionally there could be another group of characteristics that we call user pref erences which is not substantial for the hoarding at this stage thus we do not discuss it here Depending on the mobile learning system it is possible that not all the parameters can be discovered or they might be discovered through different techniques The data about the user might be obtained by any combination of questionnaires tests and quizzes or automatically by tracking the user and analysing the log files The process for retrieving auto matically the information about the user should consist of few steps like preparation of the data for analysing and application of different knowledge extraction algorithms During the first one the log files are pre processed and integrated into a database and afterwards in the second step interesting relations and deduc tions are found The user behaviour can be described in terms of browsing styles e g consecutive random interest driven etc preferred type of educational media e g prefers video to combination of text and pictures speed of read study fast medium slow etc Based on 94 the user behaviour we can group the learners and do mining based on the similarities and differences between the groups and be tween the members of the same group shown in previous sec tions This should help us mainly to predict what will be needed i e this data will be used to fill in the hoar
131. gement Sys tem The idea is to insert a software layer between the service requestor and the service provider As we discussed such soft ware layer should in general refer to an external server for at least two reasons the mobile component cannot be aware of all possi ble settings that are available in different places and the optimal choice might depend by factors that could be dynamic and there fore unknown by the mobile component The external server must obtain context data from the requestor At this point two choices 184 are possible either the server fully provides the customized ser vice or it provides a meta service i e it only identifies the best option and then passes this information back to the requestor The requestor then performs the actual customized service In some cases like in the printing problem this last solution might be highly unpractical in other cases however it might be a viable so lution and might even be preferable since it diminishes the work load on the server As we have seen implementing this middle ware in a seamless way can require digging into technical details of the infrastructure e g of the OS In the particular case of the printing it required writing or at least modifying a device driver that is not a trivial task In other cases like for instance in the case of a service provided by a Learning Management Sys tem it might mean entering in the possibly proprietary co
132. h tree values scale of importance Later we saw that this is quite useless as after the first access every word that is requested is important for the study process 110 The students that were preparing for the exam were grading the miss almost always with the highest grade The ones that were not aiming at the exam were also giving a high value and only in some rare cases were giving lower grades Different importance of different types of words Again in the context of hoarding we examined the usage of different types of words nouns verbs etc We found out that about 50 of the requests are for nouns followed by verbs 30 see Figure 33 However we discovered that from the point of view of the hoard ing verbs are a critical point especially for the pruning exclud ing phase In our initial experiments we considered a word to be known to the user if he she had the possibility to look at its entry from a text but has decided not to revise it Sometimes we con sidered a verb to be known to the learner from even several con secutive texts but later on the same verb could be requested again This happens because the verbs are linked to the entry based on infinitive form so the user might be unfamiliar with a particular conjugation which is more rare or difficult Words usage according to the word type 60 50 40 m Both Lang 30 B Italian 20 o German 10 J 0 7 7 7 n v a S
133. hat were never requested will decrease on the account of those requested rarely by few users In other words when trying to apply the described in the previous section strategy for pruning we will have every time less items to prune A useful thing to be done is again to group users by similarity and do the same statis tic only based on very similar users We have tried to do automatic grouping of our learners based on different criteria and combinations of parameters Our objective was to find if there are meaningful ways to do such grouping with the parameters extracted from the tracking data The goal of this experiment was not only to see if the algorithm can automatically split the users but also to see of there is some persistency in the clustering based on different chunks of data the separate but consecutive texts as they were presented to the users What was important to know is if the grouping done based on part of the information the part extracted from the first itera tion s of the user with the system preserves in time and could be used for our needs Automatic grouping of the users was performed based on the words usage for 16 users over six texts The data that was fed in the clustering algorithm was in table form For every text a ta ble was created where rows represented the users and columns represented the words linked inside the text The cell value was 118 set to if the word has been requested by the user
134. hat wireless m learning compared to nowadays e learning will provide much more multimedia oriented materials More Voice Graphics and Animation based instructions richer collaboration and instant communication The paper shows how the web services model maps to m learning Authors emphasize on the necessity that the architecture is an open standards based model Apart some dif ferences in the concrete technological solution also here the sys tem differs from our proposed model mainly due to the absence of the offline support module and the supposition of always avail able wireless connection though they mention as future work the possibility to download entire course content on demand In 22 an architecture for ubiquitous learning u learning is discussed which should incorporate e learning and m learning Basically the proposed architecture does not differ from our ap proach The author cites our proposal with the only negative comment that no concrete implementation is provided while their proposal starts from an established e learning environment which they extend to include m support However we do not agree with the author s opinion that offline periods should not be supported He assumes that in reasonably short period of time students will be able to afford GPRS or other connection whenever needed which will satisfy all requirements of the u system Within MobiLearn project partially described in section 2 2 the OMAF framew
135. haviour during the usage of a desktop and a mobile version of the ELDIT learning platform In this direction we have started a collaboration with Isti tuto Svizzero di Pedagogia per la Formazione Professionale ISPFP and the system will be used in real classroom environ ment with pupils The collected tracking data will be further used for analysing the users for similarity and differences in the study styles and habits Particularly interesting will be to see if there is an automatic way to distinguish the self motivated learners that we had until now from the teacher guided pupils and how we can use grouping based on this to further improve hoarding Hoarding with Different Learning Materials Multimedia During this thesis the experimentations were made with the pre viously described Mobile ELDIT system The learning material of ELDIT is a low granulated split into small chunks and so we 138 were experimenting with big number of items in the hoarding set The learning material was text based so the LOs we were working with were of a small size On the other hand we see clearly that often the materials held in e learning systems are much bigger than the chunks we experimented with Though we did out min ing always in an abstract way which should permit the deductions to be applied to data with different properties it is possible to dis cover divergence It is well known that the use of multimedia is one of the main pedagogical adv
136. he Start Manu and choose Today Pressing the OK button will actually switch off the proxy HOW TO INSTALL AND SETUP THE NETFRONT BROWSER 166 After downloading the NetFront installation appropriate for your device you should run the exe file and follow the instructions on the screen To run the browser press the Start Menu NE on the device and choose Programs Click on the NetFront icon shown up to start the browser NetFront3 When you first open the newly installed NetFront browser you should do the following 1 From the Tools menu choose Browser Settings 2 Inthe General tab a C a Uncheck Use cache Cache L Use cache Delete All Size 512 keytes Press Home button and in the newly opened window write as HOME page http www mobileELDIT com TextsList Enter the address of the HOME page http www mobileEldit com TextsList a E Note If more users will use the same device at the same time or consecutively please take a look at the instructions How to set up Usernames on page 165 Press OK button In the Network tab Check the Use proxy option 167 APPENDIX A b Write localhost without the quotes in the field on the left c Write 3128 in the field on the right e g the port Proxy C suto retrieval Use proxy Proxy host port localhost 3128 No Proxy Host Setting 4 Press the OK button in the upper right corner 5
137. he hoarding might lead to termination of the study process or even worse to misunderstanding of the material We have to as sure that the algorithm works more precisely One of the main reasons for errors of the hoarding algo rithm is the simplicity of the pruning rule that we have used in the current experiment In the cases that we have obtained 100 miss rate cases are shown with a red line surrounding them the reason was that the user had requested a text and without reading it pressed the back button and continued with other material This misled the algorithm to conclude that the user knows all the words that were provided in the text This in its turn led to ex cluding those words from further including in the hoarding set Later on the user requested the same text again this time really 107 CHAPTER 5 CONTEXTUALIZATION AND OUTCOMES reading it As the algorithm already had decided that all words are known to the user and excluded them every word requested was missing This problem can be solved by monitoring the time spent on a page so as to be able to infer whether the page was actually read Later on in the section 5 3 2 User behaviour observations one can see that in Mobile ELDIT a common time needed for reading a text is more then 3 minutes so if the time spent on a text is less then 180 seconds the user most probably did not really read it 5 3 2 User behaviour observations The main objective of the thesis i
138. he potential increases too At the University of Helsinki the LIVE Learning In Virtual Environment experiments made on SMS system and with WAP phones were very positive 89 The project went on by introduc ing digital imaging and sharing photos between the participants teachers The conclusions were that it is very possible that the introduction of MMS and the other 3G services in the large scene will lead to more and more possibilities for m learning Another project 31 on evaluation of a Short Messaging System SMS to support undergraduate students was done at Sheffield Hallam University The experiment was with 67 under graduate psychology students The implemented system was for managing learning activities to guide prompt and support the 20 students in their learning rather than for learning The findings were overwhelmingly positive with students perceiving the sys tem to be immediate convenient and personal Positive results were underlined in the outcomes from a sur vey in Norway almost 100 of the students in that University have cell phones and an SMS system would be widely accepted 26 Once again an SMS system was considered to be used to spread information about lectures and classes corrections in the schedule and etc In certain cases students find it more convenient than e mail or WWW as the information reaches them in real time Let us now consider cases of more structured interaction One of the e
139. he users will review them For example one can see on the Figure 36 below that the words sen timento and scaricare are requested by three out of four users 113 CHAPTER 5 CONTEXTUALIZATION AND OUTCOMES note that these are the infinitives while in the text a verb s con jugation and a specifically difficult form might be given as in this case assentivano is a form of one of the fourteen derivations of sentimento text it ab general 027 85 words pari B circondare reditto passare finanziario criticare pistola affare calmare scaricare passegiata racogliere D Figure 36 Overlapping in users requests costante giovane We have also noticed that there are words throughout all the tracking data not only for a particular text that appear very of ten Examples are the verb essere and the noun nome that are requested an order of magnitude more than other words Though their so frequent appearance in the logs probably has a very good pedagogical and linguistic explanation we are actually interested in the fact that the automatic discovery of such words is very simple and in the same time adding them to our so called criti cal set might improve drastically the hoarding process We have performed an analysis on the frequency of use of the words in every text and created an ordered list of them accord ing to the number
140. hesis chapters that describe these consecutive steps their different aspects the prob lems that we run into and their solutions as they came and fol lowed in time For this reason also conclusions and related work are given immediately after each separate subtopic The thesis manuscript is organized as follows Chapter 2 the following section presents a review of the re search in the Mobile Learning domain as it was done in the be ginning of the thesis in the end of 2002 It contains examples and descriptions of m learning projects and systems classified by their research aims In this manuscript this review of the literature has not been much updated on purpose for showing the situation of the m learning domain at the time the deductions and the guidelines were made However its classification structure and 8 deductions are valid also for the activities and projects of the forthcoming years of research in the field up to now In this sec tion we give the definition we adopted and used throughout the thesis discuss the accumulated from previous work conclusions and guidelines which were considered during the rest of the work Chapter 3 presents and discusses the research context The first part section 3 1 is dedicated to the results of the survey performed in order to determine the readiness of University stu dents for accepting mobile learning in their everyday life and cur riculum Though such surveys have been perfo
141. ide to visit Further it should also contain the objects that are linked to those objects that the user will access and so on The construction of the candidate set should be up to the depth level that is generally reached by the user For the first access this value can be taken as the average or the maximum depth of all previ ous first sessions The links between the pages give us the structure of the web site a learning material in particular thus we can extract the links between the LO by parsing the pages and keeping this data 87 CHAPTER 4 SOLUTION OUTLINE in a more useful format for computations These links might be either bi directional or not We can build a table that represents these links in the way shown on Listing 1 Listing 1 Creating the LO links table for every LO create a row for i1 1 number_of_LO i if current_LO contains link to LO set cell 1 else set cell 0 An example table that can be a result from this algorithm is shown on Table 1 On the first row one can see that LO contains link to LO and to LO but not to LO3 There is a bi directional link between LO and LO see row 2 col 3 and row 3 col 2 In this way we can easily observe the set of objects that the user will be possibly requesting if he she decides to browse deeper in the site i e to go one level of depth further Those would be the ob jects directly linked to a particular object From this table w
142. ight not be so clearly separated Wap browser Application Mobile Device Web browser Mobile Content Management Packaging and Synchronization and Presentation Adaptation Pana Context Discovery eLMS Web Services Interface Presentation Lhyer Business Logi ay er User Profiles fae ee Storage oe OKI Layer Figure 5 General M Learning Architecture 56 We identify three main modules in an m learning system which stands between the mobile device and the e learning system They are the following e Discovery of context e Content management and specific adaptations e Support for disconnected operation For helping the support of the services during offline periods and also gathering parts of the context information in some cases a small module will be necessary also on the mobile device Let s see the interaction between different modules by giv ing a simple example We can imagine a scenario in which a user requests an interaction with the mobile learning system from her PDA The system shows to the user a list of services which it can provide and the user selects to request more data about a seminar The system provides to the user the information about the subject speaker and location of the seminar and if the user is interested also creates a reminder which is triggered by the system depend ing on what time the user needs t
143. ile domain and whose solution should be included in every system with a large quantity of data In order to attack the main problem the full context around hoarding had to be constructed and is described throughout the thesis In this sense the thesis appears to be multidisciplinary as it treats also important questions about the construction and evalua tion of an m learning application We have started with the choice of a concrete area for experimenting in mobile learning and hoarding The chosen field was language learning and a prototype of a mobile language learning system was built We discuss the general and concrete approaches to develop and build it Motiva tions for our choices are given at every step We describe in detail the hoarding problem and the strategy to solve it with the goal to provide an efficient hoarding solution Experimental results are presented together with the practical experiences gathered from interactions with the users Here the main points of contribution are listed It should be mentioned that a great part of this thesis has appeared in different articles published in international journals and conferences The full list up to date is given as an appendix at the end of the manu script e The first contribution of the thesis is the drawing of the at tention of the researchers and developers in the mobile learning domain to the importance of the hoarding problem As we discuss above in our opinion this is
144. ined include results from the hoarding in vari ous stages and explanations about the differences of the steps taken Interesting observations over the users together with their experiences impressions and opinion were collected via inter views and questionnaires Though they are not directly connected to the hoarding problem they are reported at the end of the sec tion Chapter 6 is devoted to review of the work closely related to the hoarding problem The support for disconnected operations is ne glected by e learning society but in the mobile computing in gen eral it is an important question Different approaches are pre sented which are developed for specific cases of disconnected mobile systems and are compared to the hoarding in the learning scenario we pose Chapter 7 gives a short overview of the thesis work and is dedi cated to the conclusions and deductions made throughout Chapter 8 gives some ideas for the future work based on the cur rently proposed hoarding approach and improvements on the de veloped Mobile ELDIT system Finally a complete bibliography is followed by some appendixes that comprise other interesting work conducted within the thesis At the end a list of journal and conferences articles published dur ing the thesis is given Chapter 2 2 Mobile Learning State of the Art Computing technology has been applied to learning for decades but it has really flourished with the advent of the Web In recent
145. ing will increase the quality of the University education very often give as motivation one of the following expectations the availability of real time informa tion availability and accessibility to more information increased freedom in sense of location independence higher integration of the study process into everyday life time saving more interesting form of the study process thus higher motivation to do it Some students see the problem that m learning will be used by a few people and thus even if the presumed mobile learning 37 CHAPTER 3 RESEARCH CONTEXT system is very nice and useful the University education in general will not become better for the major part of the students Some of the answers put into mind the doubt that there is no full understanding and sometimes there is even misunderstand ing of m learning and its potential For example in one student s comment it became clear that he excludes the possibility that an m learning platform is web based or another student that wrongly believes it is impossible to visualize lecturers slides on mobile devices probably considering only cell phones etc In the questionnaire we have only a 2 line loose definition of what mobile learning is It should be also mentioned that in some cases the students wrongly connected the idea of mobile learning with the one of video registered lectures accessed via mobile devices which was by oversight influenced by a previously given q
146. ith amazing speed both from hardware and software point of view however their screens will remain comparatively small Often also the navigation is difficult Equipped with a small phone style key board or a touch screen for the PDAs users might lose more time in searching where on the page is the information they need than in reading it We can think about alternative ways of naviga tion for example voice commands The memory available on a mobile device is also relatively small It is possible to use exten sion packs on some devices like PDAs which reduces some of the restrictions but also due to their additional cost we can not presume their availability Location is a new thing to be considered Although up to now we have been talking only about limitations confronting m learning and e learning there are also advantages The small size of the de vice and the wireless connections make them available anytime and anywhere The mobility opens variety of new scenarios Ser vices involving location discovery are for example a stu dent teacher receiving directions on how to get to a certain room or alerts for seminars lectures that can be triggered while taking into consideration the current place and the time to get to the needed classroom location aware printing of the learning content etc 3 2 3 The Architecture We presente the functionalities offered generally by Learning In formation Systems LIS The services approach expo
147. itional functionalities that the device and its pre installed software were offering Unless we have specifi cally suggested them to use certain functionality for example taking notes on the device they were using only the Mobile ELDIT system The probable explanation of this fact and a similar example was found in 101 the user motivation to waste time into exploration of the new device decreases if the user is not owner of the device 130 Chapter 6 6 Related Work As can be seen in the state of the art section mobile learning in corporates a wide variety of applications that use different techno logical approaches What is missing in most of the proposed ar chitectures and systems is that they consider either only online access to the content and services or they are designed especially for small content data that fits all into the device memory exam ple the http www hotlavasoftware com mobile learning courses This is valid not only for mobile learning but for mobile domain in general see for example 16 The point to consider is that in some scenarios like the learning one the content that is to be delivered can be quite large Only some transcoding proxies take care also for caching web pages for offline usage e g AvantGo We think that delivering content for offline usage is an important issue as still mobile devices are often disconnected be cause of the lack of network access in certain places or because of
148. k project 82 the term mobile learning or in short m learning can be more and more often found in the literature in the recent years While 10 years ago only occasional papers could be found 11 in the last few years conferences and workshops are being organized on this topic The domain of mobile learning can include a wide variety of applications and new teaching and learning techniques dis cussed in Section 2 M Learning review of the literature The common criterion for entering in the mobile learning domain is to use a mobile computational device in some teaching and or study ing activities or education supporting services As the m learning domain is explored only in the recent years many new research 1 CHAPTER 1 INTRODUCTION topics are emerging in various areas including technological is sues pedagogical and methodological ones problems related to content and user interface adaptation and etc In their tries of finding the best way to apply mobile devices in education people are experimenting with different fields Courses modules were created throughout different projects for people with numeracy and literacy problems for kids university students or working adults for teachers for studying computer science subjects psy chology or language learning Mobile learning has been often considered as the next step in distance learning and as an integral part of any form of educational process of the future
149. ld be useful and a pleasure to work with The newly developed Mobile ELDIT is a version of ELDIT 30 an innovative system for online learning of the German and Italian languages and it allows the users to ac cess a subset of the ELDIT learning materials from mobile de vices namely PDAs Later in this document we will show our results at different stages of our research on hoarding using the Mobile ELDIT system and experiments done with it These re sults have the goal not only to confirm that the proposed general approach to hoarding works in practice but also to show how the different techniques and parameters influence its work Also us ers experiences and feedback gave us important indications for the successful m learning future e Our originally developed as a proof of concept system called Mobile ELDIT was successfully used initially by a dozen self guided learners It appeared to be a viable and complete real world mobile learning system based on an innovative language learning system Some of our first users were using m ELDIT as an additional tool in their preparation for the exam of bilingualism They successfully passed the exam and report that studying with the PDA really helped them in it Mobile ELDIT was successfully introduced in a Miltimedia Language Centre in Merano Italy for few months in 2005 Additionally since the be ginning of 2006 Mobile ELDIT has been used in a school envi ronment for teacher guided
150. ld offer all possible features including rich multimedia and video Only in some rare cases the students presume simple software with 40 clear functionalities though often assumption is the simple and intuitive interface fast and comfortable system Other interesting views It should reproduce the university environment It should provide all services provided in the Secretariat like certificate re quests and etc It should facilitate the communication between students and teachers and push the students to interact between themselves via forums instant messaging e mail etc Different lectures should be available on the net in video slides and text or at least providing full references First of all it should be fast and constant the service should be always available I think that the technology already exists PDAs cell phones notebooks They only have to be made useful and be used The big problem is often the cost It will be very nice to be able to integrate the lecture notes with the video registered lecture To sum up the main services which m learning must provide to support rich and actual educational information it must be an in formation system to support traditional learning by providing the following a timetable abstracts of lecture themes test and exam results messages to carry out tests and questionnaires etc to ensure fast and convenient access to learnin
151. le ELDIT transactions ceeeceeeceteeeeeteceeeeeeeeenees 76 The ideal hoarding Set ccceccecsceeseesseeseeeseeeseeeeenteesseenees 81 The expected picture cecccceesseesceeseeseeeeeeeseesseseeceeesseenaes 82 The hoarding starting Step ccecceeseeseeesceseeeseeseeeteeeseenees 84 Web based material Structure ceeseeeescneeeeeeeeeeeeeeeees 86 Browsing path cceccesesseescesceceeeseeeeeeeeeeereeeeneeneenseenaes 87 User knowledge as a subset of the knowledge base 89 Overlapping of LO accessed from different locations 90 The process of extracting knowledge about the student 98 Log file collected from the device side proxy 0 101 Database containing the content data description and user tracking IMformatiOn c cee ceseceeceseeeteceeeeeeeeeeeseeeeeeneees 102 Example data showing the decreasing of the hoarding set sy idateed ditties agen alte ation maaan ensvaa utente hee 106 User session length in Mobile ELDIT cesses 109 Ne of texts read in one m ELDIT access n 109 Vil Figure 32 Figure 33 Figure 34 Figure 35 Figure 36 Figure 37 Figure 38 Figure 39 Figure 40 Figure 41 Figure 42 Figure 43 Figure 44 Example of consecutive browsing behaviour of a user 110 Words usage according to the word type ceseeseeeteees 111 Typical pattern for a user preparing for the bilingualism NADI a ded psc tett ee odseetads atu
152. le computing E learning is grow ing at a very fast rate nowadays most universities have at least some degree of support for e learning companies are investing in the field and the need for continuous education pushes for e learning solutions On the other hand it is likely that mobile tele communication will continue to grow and to add new services Competing and complementary wireless technologies like wire less LAN Bluetooth GPRS and UMTS will multiply potential handheld applications IDC forecasts that 63 millions handhelds will be sold by 2004 and that approximately 38 of them will be smart phones integrating PDA functionality with features for communication Most mobile clients will support Java J2ME making it easier and less costly to develop portable applications Given such scenario forecasting the success of m learning seems to be an easy bet It is more difficult to understand in detail how m learning will help reaching the goals of a better learning and how it will be different from the rest of e learning According to 177 APPENDIX B the literature see the conclusions of the State of the art section 2 3 but also 117 and 95 successful m learning will be charac terized by the following properties 5 minute value the ability to use small fragments of time e g waiting time for learning e g doing quizzes using a discus sion forum communicating reading material Simplicity the limited display an
153. mation for analysing the user behaviour for the hoarding purposes The server has the important functionality of adapting the content to the PDA by rendering it into proper format for the 70 device screen and displaying limitations content redesign for analysing the collected information about the user user model ing and for predicting the learners future needs in order to pre pare the material that will be used during offline periods hoard ing and pre fetching Also on connection the cache is filled in with the predicted by the hoarding subsystem set of learning ob jects Note that throughout the manuscript we use the term learn ing object LO for referring learning units and more concrete separate HTML pages in the mobile ELDIT system Nevertheless it might be any digital chunk of learning content that is in some way connected to the other chunks For the needs regarding content adaptation of the Mobile ELDIT system the only context information that has to be discov ered is the device hardware and software limitations Knowing the screen size the browser type and the device s browser support for scripts and frames allows the Content Redesign Engine mo dule to create the proper look for the Mobile ELDIT pages As a first step we chose the easiest way to discover the context through the device browser s HTTP request that is captured on the server site GET http science unitn it mEldit text 056 HTTP 1
154. n There are two possible ways to ask the server to print a document one is to pass to the server the current version of the document and the information about the application that is using the document and the other is to pass to the server a printable de vice independent version of the document such as a postscript file The first solution requires the server to carry all possible programs and to recognize all possible file formats which makes this option inconvenient To achieve the second is much easier one needs a postscript printer driver on the mobile device side which produces a Postscript file and then sends this file to the server On the server side the file is printed on the chosen printer It is possible to print Postscript files also on non Postscript print ers e g using on the server the Ghostscript program that is available for different operating systems 181 APPENDIX B At this point what we call printing on the mobile de vice actually means 1 print the document to a postscript file 2 pass to the server the context information and the generated file 3 have the server choose the printer send the postscript file to it and pass back the info about the chosen printer One last problem remains open all this should happen when the user chooses the print menu item This means that one should write a pseudo printer driver that when invoked performs all these actions This
155. n adaptation can include adaptation of the struc ture adaptation of the media format quality or even type etc This module should be also used to adapt the presentation for auxiliary services not only presentation of content Packaging and Synchronization For allowing offline usage we need a mechanism for selecting what is needed by the user and also for taking care of content s coherence and synchronization with the system During the off line usage it is better to continue the tracking of the user activities and feedback the statistics to the LMS This module should be able to predict which learning path the user is most likely to fol low and assign weights to the learning objects depending on how important they are for the next user session The objects with higher weights should be uploaded to the device first afterwards the materials with smaller weights should be uploaded until the device s available cache is filled The module should be able to analyse how successfully the previous uploads were done and im prove further prediction 3 2 4 Related Work A work closely related to ours on defining architecture for mobile learning system is on defining the requirements for a mobile e learning platform presented in 59 The authors discuss the pos sible m learning scenarios in respect of e learning platforms and the functionalities an m learning platform is best suitable for Also the characteristics of the mobile device
156. nce unitn it foxy MobileELDIT Form ph The sample package contains three texts of three thematic groups in both difficulty levels AB more difficult C sim pler in both Italian and German languages 2 A special virtual machine for your PDA called ewe http www ewesoft com You can see list of all supported devices and download the latest version from the EweSoft web site I m providing only 162 the ewe virtual machine installation for PocketPC with which the system was tested at the following address http www science unitn it foxy mELDIT Ewe143 CAB PocketPC zip You also need to use a different browser NetFronts3 A trial version for PocketPC is available on the developer s web site http nfppc access co jp english agree html Here you can download the installation for PocketPC2002 with which the system was tested http www science unitn it foxy mELDIT NF31PPC2AREN R10D ZIP HOW TO INSTALL THE EWE VIRTUAL MACHINE To install the EWE virtual machine on your device you should you have downloaded you should find the ewe installation i e a cab file which is appropriate for your device processor type You should copy this file and run it connect the device to your PC and use ActiveSync In the zip file a on the device using File Explorer Follow C J the instructions that appear on the screen EWE virtual machine will be installed automatically and the cab file itself will
157. nchronization with 46 PC e mail GPRS and etc are rarely utilized It is noticeable that also here the authors suppose that often the students do not use the full potential of their devices even on cell phones because of the high costs of the services One should pay attention that the cost factor did not change over a few years Another very similar study was done at University of Oulu Finland with the aim to explore university students conceptions of their needs for mobile tools and what kind of features they would appreciate 66 Their survey also reveals that 100 of the participants posses a mobile phone but use it mainly for calls and SMS 83 5 have never used a handheld device and quite a lot do not have laptops 41 The already existing wireless network is also very rarely used only by 17 4 of the participants The re sults of the study states that 86 of the subjects want to read their emails via mobile device 57 want to use library services via mobile devices and 30 5 of the subjects are interested in using the learning environment in a mobile device Students attitude is that mobile learning environment could be used mainly for in formation delivery and discussion Some also mention that in their opinion mobile devices are not adequate for visualization of content In 42 a survey wais done on the students attitude to use W CDMA phones in classrooms Their conclusions are that stu dents are hap
158. nd most probable path he she will follow during the next learning session During the work described here we observed that users generally followed very consecutive path i e when given a list of texts the users 137 CHAPTER 7 CONCLUSIONS AND FUTURE WORK were reading them one after the other as they appear in the list rarely skipping some of the texts and even more rarely going back to one that was skipped However we didn t do more formal and deep automatic analysis on usage and access patterns because of the small quantity of tracking data we managed to collect In our experiments we used the assumption that the student will have consecutive access pattern and we were concentrating on the other hoarding steps In our opinion this is a weak point in our experimentation phase as it is probable that in other systems that contain different type of learning material or data with different structure the research on access patterns will be crucial and should be thoroughly considered Larger Scale Experimentations Results shown in the current manuscript are based on small scale experiments It is clear that hoarding requires analyzing of user behaviour and acquired knowledge Certainty for the correctness of the deductions depends strongly on the quantity of data that was analyzed A logical step would be to check the correctness of the deductions in a larger scale In this context especially interest ing will be also to confront users be
159. ne can see this also in the students opinion about prices for any kind of device 10 more Bulgari ans consider its cost higher compared to Italians However about 10 more Italian students do not use any e learning solution and the general attitude of Bulgarian students both to e learning and m learning is definitively more positive About 10 more Bulgarians think e learning enhances the quality of their education and more than 20 more Bulgarians are eager to use m learning The difference rises to 30 when the question is if mobile learning will increase the quality of instruction Our suggestion about this difference is that in the last year or so at the University of Rousse different surveys and questionnaires are given to students to find out the ways to improve the quality of instruction and often students see the changes based on their sug gestions In those surveys mobile learning is often mentioned and students are probably more informed and more optimistic about its success Differences according to the studied subjects We did not find any specific differences depending on the studied subjects The small exception is the fact that the percentage of the students of University of Trento from non technical specialties that use any e learning platform is bigger compared to engineering specialties unexpectedly more than 80 vs about 63 On the other hand non technical specialties generally use only the Univer sity s
160. ne periods is called hoarding Related to hoarding terms are caching and pre fetching Though in similar contexts to ours sometimes they are used as synonyms of hoarding they are more often used when consider ing online conditions and Web performance Caching is a tech nique for keeping content that has been requested by one user available on the nearest server for a certain amount of time so other requestors can access it faster Pre fetching on the other hand is a technique which tries to guess what will be needed to the client in the near future cache it and in this way improve the clients experience In the context of mobile learning we prefer us ing the term hoarding which in some sense combines caching and pre fetching Different schemes of caching and pre fetching are proposed and the goals are the reducing of network traffic mini mizing the access latency bottlenecks servers workload etc in the WWW world Although the goal of hoarding content for off line usage is shifted a little from the one of Web caching some of the techniques can be reused However while in the online case one can balance between the accuracy of the cached set and the added traffic in the situation we consider much higher accuracy is required and the added limitation is the memory availability The learning scenario has characteristics that expose some addi tional information to be considered and thereby possibility to pro vide an efficient soluti
161. nected PDAs and their achievements were shown on a whiteboard The results from the pedagogical point of view were again very positive as the students were very shy they preferred to keep quiet and the teacher could not find out the real level of their knowledge Compared with traditional classrooms virtual environment and technology motivate more participation and col laborative dynamics between instructor and learners 2 3 Guidelines for m learning applications We have been trying to catalogue research on m learning in three main areas infrastructure content and communication collabo ration We shall here conclude first by providing some guidelines for m learning applications and then by summarizing the direc tion in which we believe valuable contributions are expected The nature of mobile devices with their small screens and poor input capabilities leads to the assumption that they can not replace standard desktop computers or laptops But the same properties can make them efficient in learning domain We report here some guidelines that can be found in 95 28 Modules should be short and last no more that 5 10 minutes Users should be able to use their small fragments of waiting or idle time for learning by reading small pieces of data do ing quizzes or using forums or chat Simple funny and added value functionality The computa tional power and other properties of mobile devices make it 23 CHAPTER 2 STATE
162. ned and the hoarding set was smaller On Figure 29 we show how the hoarding set is getting smaller for one of the users of the system We have chosen a par ticipant with very common behaviour to demonstrate the general ideas and further we discuss some particularities we noticed in other users behaviour On the figure below the x coordinate is the step for the calculations which we chose to be one text the dots in the lower part show the real user requests the squares rep resent the size of the hoard and the line shows the trend of how the hoard decrease with every next text that the user was reading the triangles show the miss rate calculated as the percentage of accesses for which the cache was ineffective if it is not zero First Hoarding Experimentations 100 4 90 80 70 60 50 Hoard Size percentage form the full material set 40 30 20 10 Miad 2 ce 3 ee s s oe 3 e S H 0 a NE n e RESTS CO I OL Lavery fe aes 0 5 10 15 20 25 30 35 40 45 50 Hoarding Step Used Items m Uploaded Items A Error Rate Figure 29 Example data showing the decreasing of the hoarding set First of all the graphic on Figure 29 shows that hoarding process works even in this rough first iteration and the very simple prun ing rule we have used here One can see that within 50 steps the hoard size decreased to about 30 which makes the hit_r
163. need Also deductions of the student s knowledge should not be done based on this first access as they might be misleading Consecutive browsing behaviour A very interesting observation is that most of the users show a strict consecutive browsing be 109 CHAPTER 5 CONTEXTUALIZATION AND OUTCOMES haviour Figure 32 We expected that the users will read groups of texts in different order depending on their mood or specific in terests However generally they were reading the text in order of appearance in the list we provided Lingus del testo Tabane 2 Log File Observations gt Spanche des Texter Deutech AB 12 06 2004 Sat Famiglia bambini Famigtia bambini adacasone r edercarjome AB c Familie Kinder Jugend Famille Kinder Jugend j Erdchuny Demanen Vitex mablen o Ich lebe allen iorenti itda e bimp o Cosin m imida europea Pee otk o Maledetto cocer amp o La frets dela mamma aiik marma o Limmwersth dei bambis se weng o Kinder und Kamere We Deutschlands Kinder Caco gren Mba den ae pe Kinder an de Macit o Ge takan e la fima vecon o Paese che vai o Nuove rechae o Concrete per natura o ABa scusla dela pace Arse musica lesieratura Arse musica lestoratmna cinema teatro cinema teatro o Larmunes t munca Drage 4 Oz o A Wentees imendo 4 o Dracula e Frarkensten wl cutale o Holywood nel coset o Telia d Antone 4 de Alen Divan J
164. ness of the hoarding and discussed their pros and cons This theoretical approach gives the basis of later providing a customized solution for a concrete system leaving possibilities for future comparisons with other results Talking about practical results we have acquired and pre sented a number of positive results over Mobile ELDIT that show the correctness of the theoretical deductions It should be mentioned that though lots of work was done considering hoarding in mobile learning context we are still far away for having an optimal general solution One of the biggest problems in the experimentation phase appeared to be the need of big quantity of tracking data which to be fed in the automatic knowledge extraction algorithms As the m learning field is so new there is no such data available in advance and gathering it requires time Lots of interesting issues that could not be researched be cause of time and other constraints appeared and would be nice to be explored in the future Here I would like to list some ideas for 136 improvement of the currently proposed approaches and tech niques both for hoarding and for the prototype Mobile ELDIT system Some general ideas for interesting research directions re lated to supporting offline access to learning materials and to mo bile learning in general to which we come across during the thesis are also included 7 1 Hoarding Improvements Optimization of the hoarding process probabl
165. ng helps solving the above mentioned problem see Figure 41 but also brings additional load to the system First of all the system needs more time to load the cached content if it is zipped see our measured times on Figure 42 base on an Acer n 10 device Transfer Times of the Hoard Loading delay 1400 120 1200 100 g F 1000 a0 E g 800 8 3 v 60 gt 600 E E 40 E E 400 a a 200 20 o4 0 0 3 0 1 8 2 27 8 7 2 61 17 4 6 8 10 12 14 16 o Not zipped Hoard Size in MB Zip Size MB a Zipped Not Zipped Zipped Time to Load ZIP Figure 41 Transfer between desk Figure 42 Load Times for top PC and PDA device Zipped packages We should point out that on a system that does not use pack aging the load time is practically zero while one can see on the figure above that the load time for even quite small pack age size are quite high Nevertheless the interviews with our users showed that the load time is not as disturbing as the re ply delays of the proxy In other words the users would not mind to wait for about a minute or two while the system 128 loads as this happens once in a while as long as the re quested content is not delayed too much On Figure 43 we show that the introduction of zipping for our system has a positive effect on the proxy response time for smaller package sizes up to 8MB whil
166. ng is by de fault arranged on courses lectures classes etc A logical se quence is the development and experimentation on transforming traditional courses in a form appropriate for mobile devices The M Learning project is one of the projects that have a special section dedicated on creation of a WAP portal for educa tional purposes More concretely this is the part developed at Ul tralab i e http www ultralab ac uk projects m learning The technical aspects in the creation of a WAP portal for educational purposes do not differ from a common WAP portal As the target users for this project are young people age 16 24 with literacy problems the group studies the problem of keeping the interest of the young adults to the useful learning materials by exposing also modish and exciting subjects A special attention is paid on the pedagogical aspect of education The M Learning project team is also producing offline m learning materials for people with literacy and numeracy prob lems 20 102 A great potential is encountered from the cogni tive and pedagogical point of view Learning modules are created by using standard tools like Macromedia Flash in its version for mobile devices The preliminary conclusions are that new tech nologies have great impact on students interest in the subject studied In this case this was one of the main wanted repercus sions The positive results of many more systems developed to combine WAP
167. ning of the LOs based on our supposition that the user gather this information by analysing the user browsing behaviour knows a certain LO Overlapping of LO accessed from different locations Figure 25 up to a word entry which are the LO in this case thus we have some overlap in the data that can be accessed from different loca tions e g the same words will be presented in more than one It was previously mentioned that our data is very low granulated 90 text On Figure 25 we schematically show the LO sets of three texts The X symbols show the words that the user requested to see when reading the text in which the X belongs At the first user access we did not have any knowledge about his her language skills thus no pruning was done If the user was reading on the first session Text_1 and it was predicted that for the second off line session Text_2 should be prepared then we can prune the LO that the user had the possibility to access the last time but de cided not to do it Thus we subtract from the whole set of LO for Text_2 the words that we consider the user knows It should be pointed out that it is possible that the user only opens a page with a text but doesn t really read it In such a case the LO that were contained in the set will be wrongly considered as known by the user Thus this elementary rule might be too simple and lead to big hoarding miss rate If used with a combi nation of other rules the accuracy
168. ns of their needs for mobile tools for scaffolding learning activities Proc of Ed Media 04 June 27 July 2 2005 Montreal Canada Lonsdale P Baber C Sharples M Arvanitis T N 2003 A context awareness architecture for facilitating mobile learning MLEARN 2003 London UK May 19 20 2003 MacKenzie S amp Kober H amp Smith D amp Jones T amp Skepner E 2001 LetterWise Prefix based Disambiguation for Mobile Text In put 14th annual ACM symposium on User interface software and tech nology 2001 Orlando Florida pp 111 120 MacKenzie S 2002 KSPC Keystrokes per Character as a Charac teristic of Text Entry Techniques 4th International Symposium on Mo bile Human Computer Interaction 2002 Pisa Italy pp 354 358 Marmasse N Schmandt C 2000 Location aware Information Deliv ery with ComMotion Second International Symposion on Handheld 150 71 72 73 74 75 76 77 78 79 80 81 82 and Ubiquitous Computing 2000 HUC2K Bristol UK Sept 25 27 2000 LNCS 1927 Springer 157 171 Milrad M amp Perez J amp Hoppe U 2002 C Notes Designing a Mo bile and Wireless Application to Support Collaborative Knowledge Building International Workshop on Wireless and Mobile Technologies in Education 2002 V xj Sweden pp 117 120 MobilEdia news from August 31 2005 available on September 14 200
169. nt formats by diverse applications and services Such option would be possible to implement inside our more abstract frame work by including a way to translate the information properly Still the work presented in this paper proves the viability also of our ideas The authors also see the web services as most appropri ate way for integrating their context aware sub system with a mobile learning system In 90 architecture for m learning based on web services is discussed The analyses show that this technology is appropriate for supporting mobile equipped users in learning scenarios The authors find one of the biggest challenges in the ability of such system to convert in satisfactory time the data LO from one format into another They find the solution in preliminary before request creation of different versions A major problem that we 61 CHAPTER 3 RESEARCH CONTEXT find in this work is that the only way the system would support the offline usage of material is by manual users request of pre prepared modules students could easily access and download the entire course content anytime anywhere on their mobile de vice The authors also suppose that in all cases the entire course will fit into device memory which is in contrast with our assump tions More recently another web service based architecture was discussed in 98 A very positive attitude to m learning can be felt from the paper with authors believing t
170. o believe that most probably the reason is in the number and variety of learning materials published on every 32 university platform for every single course or specialty Generally when the university platform offers good set of materials students prefer to use this unique source otherwise they look for other e learning platforms or even just sources on the Internet Nevertheless there are gender differences that appear in the same manner in both groups Bulgarian and Italian students The percentage of the girls who use only their own university s e learning platform is about 10 higher than the boys The percent age of girls who use different platforms is almost two times lower than the boys The percentage of girls who do not use any plat form is higher compared to the percentage of boys More generally the most often given reasons for using e learning are convenience increased availability of the learning materials higher speed of access to the materials the possibility to search through the lectures and other digital sources access to large volume of useful learning information and etc Some people mention that the quality of the lecture is increased by the exis tence of e learning resources connected to the studied subject and that this gives ability for a personal improvement and qualifica tion Other often given reasons in favour of the more modern ap proaches and extensions of the traditional university education are the
171. o get to the seminar room Later the system gives to the user directions on how to get to the semi nar room and during the seminar lets the user watch the slideshow of the presentation also on the PDA display The user might take notes and attach them to the presentation slides When the presen tation finishes the user might go home and print his notes from his home PC On the next day at the Faculty a friend asks about the seminar and the students decides to share with the other one his notes by printing them on the nearest printer The system gives to the user directions on where the nearest printer is located Let us now connect the scenario described above with the functionalities as they are executed by the different modules of the architecture presented on Figure 5 First the user request is captured and in order to proceed the system need to know who the user is and what is the device used This is done automatically by the Context Discovery module which based on the first re quest or additional interaction already holds the information about the user id and the capabilities and limitations of the device both software and hardware Based on this data the system can 57 CHAPTER 3 RESEARCH CONTEXT check the user role student teacher guest etc and access rights in the eLMS and decide what services can be offered in this mo ment and propose the proper list to the user After the next inter action with the user the m lea
172. o use a browser as an interface to the learning ma terial Most of the browsers on the mobile devices nowadays still do not support frames and have only limited support for script languages This leads to the need of specific adaptation of the content The adaptation is also needed because commonly web pages are designed for screen size at least 800x600 hence they are hard to be read and or navigated from devices with smaller screens ELDIT does not make an exception Different adaptation techniques can be used to tackle this problem 12 The adaptation can be done at server side it can be done in a proxy between the server and the client or it can be done on the client side Every one of these solutions has its pros and cons As mentioned earlier the data of the ELDIT system consists of XML files example shown on Figure 15 both for the texts and for the word entries For displaying the data on a desktop PC or a laptop dynamic HTML pages are produced on the server site This is done on the fly on every user request in order to facilitate adaptation to the user Another reason for generating the pages on the fly is that the data is often updated and new data is added con tinuously by the linguists These pages contain frames and JavaScripts for easy navigation and the word entries are highly interlinked 72 iojx File Edit View Help k2xml version 1 0 encodi lt DOCTYPE word SYSTEM 1S0 8859 1 2 gt Program Files
173. ods on every single text 30 40min as they were picking up more word entries since they were trying to answer the comprehension questions This means that on average one text was read in one session Others that want to improve their language skills might read certain texts in 3 5 minutes thus reading more texts in a ses sion Sometimes rather rarely a user was reading a large number of texts see Figure 31 for a longer period gt 1h Session Length Number of Texts per Access User Sessions Texts count Figure 30 User session length Figure 31 Ne of texts read in in Mobile ELDIT one m ELDIT access Random behaviour on first access The final goal of our experi mentation is to support the automatic selection of learning mate rial for offline periods For this we need to be able to predict what material the user will need in his her next learning session Our observations show that during the first use of the system the learner is exploring what is possible to be done in this unknown environment so his her actions are quite unpredictable For ex ample it is very likely that the user will click on a word that he she is familiar with just to see what information is available This leads to the impossibility to exclude even easy basic word entries e g the word essere to be Later on the users start really studying and do not click on words that they do not
174. of pruning should be noticeably higher For example one can look at the time needed for reading certain page if the time was below a given threshold the material can be considered as not read 4 7 Prioritizing Setting the priorities to the LO that are still in the hoarding set af ter the pruning process is also a very important step This is be cause even when pruned the set might be still bigger than the available mobile device memory and only part of it will fit in The priorities in the hoarding context should mean how important the object is for the next user session and should be higher if we sup pose that there is a higher probability that an object will be used sooner In this sense the predicted starting point of the user s next offline session should be always assigned a maximum prior ity For prioritizing the LO we can analyse the accesses done previously by all the users and extract interesting and important knowledge Aggregated data like the correlation between the ob jects based on their contemporary usage in other users sessions is one thing that can be easily discovered and is very helpful For 91 CHAPTER 4 SOLUTION OUTLINE example a well known association rules see 36 discovery can be applied to determine from all previous learning sessions the relations between LOs that are very strong i e associations dis covered with confidence near to 1 and big enough support value Note that when se
175. om here later was easier to extract portions of the data which were fed into the knowledge extraction algorithms The da tabase shown on Figure 28 is composed of two logical parts which are actually interconnected one is the data about content the upper part on the figure and the second is the pre processed tracking information the lower part on the figure fles File_ID EE 3 groups D mante Y File_ID BIGINT 20 FK Group_ID BIGINT 20 FileName VARCHAR 255 Text Title VARCHAR SS Subject VARCHAR 255 FileSize BIGINT 20 al aT el ext_ From_File gt i gt links X texts_groups Accessed Text ID FromFile BIGINT 20 FK Group_ID BIGINT 20 FK D Link VARCHAR 255 Text_ID BIGINT 20 FK Expl VARCHAR 255 Q baseFileID BIGINT 20 bf ind Q baseFileID session_history Session_ID BIGINT 20 FK Q Position BIGINT 20 accessed_texts hd Item VARCHAR 255 Text_ID BIGINT 20 FK Q Date DATE Session_ID BIGINT 20 FK Time TIME Total_Time_Seconds BIGINT 20 Q Time_Spent_Seconds BIGINT 20 Q Links_Count BIGINT 20 Display_Delay INTEGER 11 Session_ID ID BIGINT 20 __ _ daily _session be sessions User_ID BIGINT 20 FK Session_ID BIGINT 20 Start_Date DATE User_ID BIGINT 20 FK Q Weekday VARCHAR 255 Start_Date DATE Tot
176. ommunications Vol 3 No 3 pp 303 323 Attewell J 2004 Mobile technologies and learning A technology up date and m learning project summary available on September 14 2005 online at http www sda org uk files pdf 041923RS pdf Battiti R Brunato M Villani A 2002 Statistical Learning Theory for Location Fingerprinting in Wireless LANs private communication Battiti R Brunato M Villani A Statistical Learning Theory for Loca tion Fingerprinting in Wireless LANs Technical report DIT 02 0086 University of Trento October 2002 Bickmore T W amp Girgensohn A amp Sullivan J W 1999 Web Page Filtering and Re Authoring for Mobile Users The Computer Journal 42 6 pp 534 546 Bickmore T W amp Schilit B N 1997 Digestor Device Independent Access To The World Wide Web Sixth International World Wide Web Conference 1997 Santa Clara CA USA Bohnenberger T amp Jameson A amp Kr ger A amp Butz A 2002 Lo cation Aware Shopping Assistance Evaluation of a Decision Theoretic Approach International Symposium on Mobile Human Computer Inter action 2002 Pisa Italy pp 155 159 Broadbent B 1997 Designing Training for Mobile Computing Pre sented at the Annual International Conference of the American Society of Training and Development May 1997 145 12 13 14 15 16 17 18 19 20 21 22 23
177. on Despite of its importance this issue has not been addressed seriously up to now Moreover people have avoided facing this problem for years saying that mobile devices characteristics are continuously growing and soon a fast Internet connection will al ways be available In the fall of 2000 Clark Quinn 80 wrote The vision of mobile computing is that of portable even wearable computation rich interactivity total connectivity and powerful processing a small device that is always networked allowing easy input through pens and or speech or even a keyboard when necessary though it may be something completely different like a CHAPTER 1 INTRODUCTION chord keyboard and the ability to see high resolution images and hear quality sound 39 66 We point your attention to the words portable total connec tivity and always networked About five years later in 2005 such situation is still not reached which for mobile users means that learning content is not accessible during periods of discon nection We can still see the neglecting of the need to support off line delivery of learning material to users equipped with mobile devices An example can be seen in 22 where talking about ubiquitous learning environment and discussing the u learning ar chitecture no offline access is considered to be supported The ar gumentation from Des Casey is that it is reasonable to assume that GPRS and similar
178. on the devices we were us ing and we were experiencing much bigger delays comparing 173 APPENDIX A with NetFront browser However it seams that in a short time Minimo will be the browser of our choice Why the proxy is needed Mobile ELDIT works offline utilizing a caching proxy called FoxyProxy thus does not require Internet connection In fact the proxy catches the requests of the browser and extracts the needed data from the cache In case that the requested entry is missing it informs the user The proxy also collects log files of the requested entries with associated request times This is needed for the research on the hoarding problem for details see http www science unitn it foxy papers html Can I browse the Internet through NetFront browser or it is only for utilizing Mobile ELDIT NetFront browser is not especially for utilizing Mobile ELDIT Though because our system is still under development in or der to use NetFront for browsing the Internet ie to access pages that are not part of Mobile ELDIT you should set up NetFront NOT to use the proxy FoxyProxy I don t have my own device Can I use Mobile ELDIT any way Can I borrow a device How Where Mobile ELDIT can be found also at the Mediateca Multilin gue at Bolzano and Merano Few PDA devices are available for use and for loan to the adult members of the Mediateca http www provincia bz it cultura bilinguismo multilingue m ediateca_ebook i htm
179. onnected on purpose but the user wants to work and when the connection fails during his work online Depending on the application and the data needed the requirements of the first situation can be met by using AvantGo or other client side caching mechanisms see e g 114 Although thick clients can be used to maintain the syn chronization and the caching there is still the problem of the small amount of memory available on the mobile devices The data should be carefully separated and only the necessary pieces should be uploaded Other approaches provide special services for mobile de vices The delivery approach can be different such as the Satchel architecture 61 29 which provides a distance access through a special browser to documents and other resources needed during work Mobile web services could be also used as Microsoft and 14 IBM released versions of Web Services Toolkits for mobile de vices This enables access to Web services on enterprise servers but although there is a big potential in Mobile Web Services there are some disadvantages and problems that should be overcome One of them is loss of network connectivity the service is not available if there is no connection and the question of how a sys tem should recover from a failed web service stays open It is also not clear how the services are discovered in peer to peer networks and how to manage the resources of the devices These issues are the objective of rese
180. ood starting point for hoarding predictions whenever an instructivism approach is applicable It should be also mentioned that in our scenario the material the texts were just listed and not specifi cally ordered which allows to the user the freedom to navigate as preferred Still the users show this consecutive browsing behav iour In contrast with our system often for reading certain material there is a pre condition given by the educator that the user should be already familiar with the previous topics Depending on the mobile learning system additional information about those 86 pre conditions can be known to the system and considered in the predictions Evaluation of the user competence on a subject will surely contribute 4 5 Generate candidate set As mentioned earlier one of the steps of the hoarding algorithm is to construct the candidate set of learning objects to be hoarded When using web based material the user clicks on the links of one page to go to another one and can either continue to browse further or can go back to a previously viewed page Figure 23 Starting age O lt zK X Sees level 1 Ss N L p x FAAR 1 i f NO de eae yel 2 Figure 23 Browsing path This means that the candidate set should contain the objects linked to the starting point i e the objects that the user might de c
181. ork was developed 21 This work was published about the same time as ours The framework which is also abstract provides a different view to a module base m 62 learning without specifying functionalities to be provided or the ways to do it The main idea is to define a layered model and specifications for interfaces between the layers so that to provide interoperability for already available and new services Another mobile learning prototype based on detailed mo bile learning architecture is described in 34 Authors aim at the support of adaptation for mobile users As the adaptation dimen sions are content user model device connectivity and coordina tion it turns to be quite general In fact from the point of view of functionalities it allows is very similar to our proposal The main difference is the absence of support for offline work The shortage is known to the authors and is mentioned as crucial future work The work was also published nearly at the same time as our work A lot of work has been done in the area of content adapta tion for mobile devices and of device independent representation of web content Different approaches are proposed for describing device capabilities different architectural approaches are devel oped for using this information for adapting the content accord ingly A comprehensive review of the current device independence technologies and activities could be found in 108 a and 12 Trans
182. ould also 116 decide to add a threshold bigger than 0 for pruning for example 5 or 10 i e to prune rarely used items This would make the hoard decrease even more but would possibly increase the er ror_rate Thus it is important to mention that this threshold for pruning is one of the parameters that should be carefully chosen in a real world system The second effect namely the continuing existence of hoarding misses happens because the user actually requested an item that was never before requested by other users and thus was pruned In our opinion this effect is mainly because of the small number of users that we had for the experiments In a situation when tracking data from much more users will be available we would expect this to happen rarely In the cases when the device s available memory is still smaller than the predicted hoard after the pruning step we could use the usage percentage as a criteria for ordering the items for hoarding step 4 of the general algorithm described in Section 4 The prioritizing should be done after pruning the items which be long to the user knowledge set Nevertheless in some cases when the percentage of usage for a certain word LO in the general case is very high the item might be considered to be included into the hoard even if it is in he user knowledge set These would be the cases when a particularly difficult word form e g verb conjugation is used The thresholds of the pe
183. p 62 67 ClickZ Wireless Trends amp Statistics from August 29 2005 available at http www clickz com stats sectors wireless article php 3530886 on September 14 2005 Collett M amp Stead G 2002 Meeting the Challenge Producing M Learning Materials for Young Adults with Numeracy and Literacy Needs Proc of the European Workshop on Mobile and Contextual Learning 2002 Birmingham UK pp 61 62 Da Bormida G Bo G Lefrere P Taylor J 2003 An Open Abstract Frameworkd for Modeling Interoperability of Mobile Learning Ser vices Proceedings of 5th International Conference On Enterprise In formation Systems 23 26 April 2003 Angers France Des Casey 2005 u Learning e Learning m Learning Proceed ings of E Learn 2005 Vancouver Canada Oct 24 28 2005 Dey A K 2001 Understanding and Using Context Journal of Personal and Ubiquitous Computing 5 1 pp 4 7 146 24 25 26 27 28 29 30 31 32 33 34 Dey A et al 2000 CyberMinder A Context aware System for Sup porting Reminders Proc Second International Symposion on Handheld and Ubiquitous Computing 2000 HUC2K Bristol UK Sept 25 27 2000 LNCS 1927 Springer 187 199 Divitini M Haugalokken O K Noverik P A 2002 Improving Com munication through Mobile Technologies Which possibilities Proc IEEE Workshop WMTE 02 29 30 Aug 02 V xj Sw
184. paring for the exam concentrated on one difficulty level namely the level for which they wanted to conduct the exam see Figure 34 German Speaking German Speaking Preparing for the Exam NOT Preparing for the Exam 60 70 50 60 50 40 30 20 10 10 0 0 de ab de c it ab it c de ab de c it ab ite 40 30 20 Figure 34 Typical pattern for a Figure 35 Typical pattern for a user preparing for the bilin user not preparing for the bi gualism exam lingualism exam Typically they were reading texts in both languages concentrat ing slightly more on the texts in their native language since in this case it is harder to compose correct answers remember that students have to answer questions in the other language Alterna tively users that were not preparing for the exam were concentrat ing on texts in the target language e g Italian for German speak 112 ing user and this without considering the difficulty see Figure 35 Only driven by curiosity they also browse texts in their na tive language 5 3 3 Hoarding with Critical Set In our first hoarding experiments we showed that the simple rule for hoarding which we used worked in the sense of decreasing the size of the hoard and thus increasing the hit_rate but there are sometimes a large number of entries that are wrongly ex clu
185. pated in a few days experi ment Some of those in the first group used the system for few weeks period just before their exam of bilingualism Others util ized it for much longer period almost a year thus that we can have more data and try to analyse also evolving in their behav iour All this data has been put into the database as shown above which later gives a possibility to easily extract portions and con verting the data into a format appropriate for every specific knowledge extraction algorithm 104 5 3 Hoarding results In our hoarding experiments with Mobile ELDIT system the us age data was collected in the following manner as a first step the users were given a set of ELDIT texts and were asked to study them at their convenience Whenever the users felt that they had finished with the current portion of texts they were given an other set As mentioned before part of the users were preparing for the bilingualism exam others were just studying the language without aiming at passing the exam Only in certain cases the us ers were given the option to choose the texts they would like to read Nevertheless whenever the full data set was not fitting de vice memory the sets of words were chosen randomly Our initial hoarding experiments had the aim to explore the basic hoarding system that uses very simple rules for the pruning and has shown that the hoarding will really work Step by step we tried to use more complica
186. pected duration of time in which the system will be used offline topics preferred by the user etc Different other options could be fore seen for instance the proxy might be aware of the cost of the connection and behave in different ways according to that i e synchronizing the cache when the cheap connection is available Internet through LAN or cradle and using only the cached con tent whenever possible on expensive connections Another functionality of the proxy is the tracking of the u ser s activities When connection is available or the device is be ing synchronized the log files should be uploaded to the mobile Learning Management System m LMS There the mobile ver sion of the user models should be updated and the Packaging module will be aware of the user s needs and adapt accordingly The m LMS should be responsible for calculating and updating 75 CHAPTER 3 RESEARCH CONTEXT the user models which will differ from the user models in a stan dard LMS o E o 2 re a Lay Online Offline period F Ly Intsinet Connection i Figure 18 Mobile ELDIT transactions The Figure 18 above shows an exemple transactional sequence of requests responses between the mobile device where the web browser and the proxy sit and the two web servers ELDIT eLMS and m ELDIT mLMS The figure shows two online pe riods in grey and one o
187. py to do it but different models of the devices give different possibilities functionalities for different students which the lack of homogeneousness might be a problem The use of new media increases pupils interest to certain activities still the phones are mainly used for communication and collaboration via voice calls and text picture messages A study conducted at University of Dublin Ireland 74 led to a strong expectation of the authors that the future of learn ing is bound to mobile and wireless They testify fast growth in computational devices and Internet use for educational purposes in their institution At the same time they note that funding for providing enough devices to students might be a problem In their 47 CHAPTER 3 RESEARCH CONTEXT opinion students will be reluctant to pay for such devices for themselves When talking about general statistics and predictions about the usage of mobile devices lots of studies have been done for dif ferent countries USA Japan Republic of Korea Morocco Norway and etc 43 All of them show that cell phones are spreading very fast and are owned and used by nearly 100 of the young people They are used at various locations and SMS is quite popular On the other hand PDAs and smart phones are con sidered business oriented devices and are rarely owned by stu dents see 5 thus are probably not the best choice in the context we consider here Other sources 72 and 19
188. r and his her learning history and the setting where the learning process takes place As mentioned earlier an important factor in mobile envi ronment is to embrace also the context The MobiLearn project www mobilearn org is one of the biggest and most important European led research project and is an example where also the context information is taken into consideration in the architecture phase Participants in the project discuss the importance of loca tion dependent learning like presenting learning content on the spot i e information given to the students while visiting museum As MobiLearn is a large International project it has broad band of goals like creating a general framework for m learning creation of pedagogical paradigms exploration of adaptation for mobile devices and realization of a new business model together with prototypes implementation understanding in depth the process of learning in different contexts etc The target users are workers and citizens in their everyday learning activities e g citizens vis iting cultural city and its museums or family members using sim ple medical information on the spot 2 2 3 Communicating and Interacting with People Interaction can have little structure messaging or be highly structured for reaching a goal as in collaborative and problem based learning In both cases new technology has much to offer In the case of highly structured interaction pedagogical models 19
189. r the users perception of benefit which such a sys tem brings into their learning process Our interviews have also shown that missing portions of the learning material are perceived very negatively In this context the existence of an efficient hoarding subsystem plays an indispensable role for the overall measurement for positive effectiveness of the sys tem 5 4 2 Problems Found v As it was mentioned before wireless and mobile devices market is very dynamic Fast changes happen also in the software field for mobile devices few operating systems are available that are incompatible between each other but are of ten incompatible also within versions This fact triggers re search in device independence technologies but often parts of the developed system should be written specifically for a given platform A possible solution that we tried to explore is the usage of java technology for overcoming the problem of developing the same thing twice Our experience shows that in this early stage there is no even Java Virtual Machine JVM with equal behaviour for the different platforms Even worse using the same JVM our system was not equally sta ble across different OS versions Our expectation though is that this problem inequality of JVMs and other standardiza tion issues will soon diminish and hopefully disappear Until then careful planning should be done on what hardware plat form to use and what should be the software
190. ractically useless as the user will rarely review over and over the same study material According to our knowledge on the scene of learning and related technologies hoarding has been hardly explored until this current work In the rare cases when offline is considered in conjunction with e learning like Backpack Mobilizer for Black board details available at http www syberworks com or mobile learning see few alternative approaches discussed at http learning ericsson net mlearning2 project_one presentation_i paq html the content is supposed to be manually selected for downloading or is fully downloaded by the system without any consideration of the available space The only profound and more formal study on the possible ways to treat the problem of discon nection in a mobile learning scenario was found in 44 and 27 They propose two different architectural models but their main concern is to track the user activities and synchronize automati cally the learner s learning progress records Both architectures do not consider the problem of what material to be pre fetched Furthermore we are not aware of continuation and practical de velopment of a real system based on these architectures 134 Chapter 7 7 Conclusions and Future Work This thesis work though aiming from the very beginning at solv ing a very concrete problem namely the hoarding of content for mobile learning in its depth appears to trigger a wide v
191. raphical characteristics as real world is polluted Kids have to take vir tual probes from the water and or air in the polluted area or sur roundings analyse the results interview people and read informa tion about similar situations and finally find out how to sublimate the environment During the game they collaborate by doing dif ferent probes and analysis and giving the results reports to their classmates or leaving them probes and reports in a certain place in the area map where other kids can find and use them The kids have to collaborate because they are forced by time limita tions Advanced wireless technologies IEEE 802 11 Bluetooth and GPRS are used in a project for development of ad hoc class room and eSchoolbag system at the Aletheia University in Tai 22 wan 15 The so called Paperless education is being observed together with the acceptance from the students the term paper less education and research on the topic is made also in http www paperlessclassroom org The traditional classroom was replaced by the new developed electronic tools electronic blackboard rubber colour chalk and so on Pupils were strongly encouraged to communicate and to learn together in groups Applications for recording the data and taking notes have been developed for Palms and the pedagogical effect of them has been analysed 3 93 Again in Taiwan 65 students were equipped with net work con
192. rcentages should be experimentally set 5 3 4 Combined Hoarding As discussed previously for speeding up the process of decreasing the hoarding size a possible strategy is to try to predict what learning objects study material is known to the user instead of as we did in our first experiment to wait until the concrete por tion of the material is shown to the learner In other words after the first interaction of a learner with the system we should try to discover what the user knows e g what objects should be put into his her knowledge base set We might do this by finding similar in their knowledge users and if a certain user has shown that 117 CHAPTER 5 CONTEXTUALIZATION AND OUTCOMES knows certain word we can suppose that another user with similar knowledge will also know it The two steps that have to be ap plied are a to use some algorithms to discover automatically similar ity between users b to predict user behaviour based on known study patterns It is also important to mention that with higher number of partici pants we would expect great diversity in users behaviours and in the sets of words they know The large quantity of tracking data will lead to lower the performance of the hoarding algorithm and more specifically of the pruning as done until now and described in the previous section What is referred here is the fact that with increasing number of users we would expect that the number of words t
193. rding is the process of automatically selecting learning content which has to be prepared and pre fetched on the mobile device s local memory for the following offline session Hoarding is highly needed in the m learning context for two main reasons On the first place is the demand to support what is called any time any place educa tion This means that on the mobile device e g a PDA the kind of device often used in m learning which might be often discon nected from the Internet the needed learning content should be available locally for allowing access during the offline periods On the second place comes the desire to hide from the student the technologies that lie behind this ubiquitous learning We would like to free the user from tedious operations of manual prepara tion and planning his her next study session Moreover often we cannot even count on the student s own judgment for his her knowledge and future needs In order to attack the main problem the full context around the hoarding had to be constructed and is described throughout the thesis In this sense the thesis appears to be multidisciplinary as it i treats also important questions about the construction and evaluation of an m learning application We have started with the choice of a concrete area for experimenting in mobile learning and hoarding The chosen field was language learning and a pro totype of a mobile language learning system was built We di
194. reat help to m learning since many problems in m learning are in common with m anything We shall try to review these different aspects The rest of this section will therefore follow the structure we outlined here We will begin by covering infrastructural research since that is a common denominator We then will examine the problem of ac cessing content from the learning perspective and we will move to facilitating interaction with other people 2 2 M Learning Research 2 2 1 Infrastructural Research Access to the web through personal electronic devices with their small screen size has been an interesting problem for lots of re searchers Unfortunately today most web pages are designed to be displayed on desktop computers with colour monitors having at least 800x600 resolution This leads to at least 4 to 1 often greater ratio of designed vs available screen area making direct presentation of most pages on the small devices aesthetically un pleasant un navigable and in the worst case completely illegi ble Work is being done in the area of device independent access to web content In this context different approaches are proposed for describing device capabilities HTTP Request Header CC PP UAPROF etc Also different architectural approaches are de veloped for using the information of devices capabilities and adapting the content accordingly The adaptation could be server based XML XSLT Cocoon Axkit proxy based
195. ricket system 79 is based on a combination of ultrasonics and radio SpotON 40 uses signal strength of radio signals In some recent research 15 CHAPTER 2 STATE OF THE ART the discovery of the position is based on the wireless signal of Wi Fi networks For mobile learning the infrastructural research described so far is just the technical base for reaching certain practical goals for developing concrete systems and to explore the possi bilities offered by technology for the learners and educators In other words they are to be used further to allow either access to content and services or communication and collaboration be tween participants During the initial phase of the doctorate some work was done on utilizing location information for providing context dependent service to mobile learners This work is pre sented as Appendix on page 177 2 2 2 Accessing Content Accessing content is one of the most important functionalities in e learning and it takes a big part of the research efforts in m learning too Generally based on the infrastructural research men tioned above for transforming data into format suitable to mobile devices some research specializes in adapting courses for mobile devices and in building learning WAP portals The most obvious use of mobile devices for educational purposes is in fact a direct application of the e learning techniques on smaller devices in stead on a desktop PC For grown up people studyi
196. rmed in other countries by different researchers a review of similar studies showed that the results might widely vary We were seeking to reveal the situation in our environment so an online survey was performed with Italian and Bulgarian university students The second part of section 3 3 2 is dedicated to the general mo bile learning architecture that we proposed and according to which we developed further our Mobile ELDIT system that con tains the hoarding subsystem In 3 3 is presented the real world system that was developed dur ing the thesis as a proof of concept of the theoretical deductions that we have made and for experimenting with different hoarding strategies Here we explain in some details the particularities of the ELDIT system part of which became our Mobile EDLIT We explain how and why we chose ELDIT describe the design prin ciples we followed for the development and the practical solu tions chosen Chapter 4 discusses the hoarding problem and the general ap proach to solve it This section starts with the formalized algo rithm that has to be followed for successful hoarding It is fol lowed by detailed step by step discussion of different sub processes of the hoarding Chapter 5 is dedicated to the contextualization of the solution i e to the concrete implementation of the general approach de scribed above into our Mobile ELDIT system Experimental out 9 CHAPTER 1 INTRODUCTION comes that we obta
197. rning system requests information about the seminar from the eLMS and triggers the Mobile Con tent Management and Presentation Adaptation module Knowing the capabilities of the device from the Context Discovery module the data is redesigned and returned to the user After wards the user requests the reminder to be set up for her The sys tem needs additional context information namely the user loca tion in order to calculate the needed time to get to the seminar room Once again the Context Discovery module is triggered to track the user current position which is changing constantly as the user moves and is checked regularly Meanwhile as the system knows that the network is not accessible in the seminar room it triggers the Packaging and Synchronization module The eLMS might contain a large amount of materials concerning the seminar the presentation itself including explanations from the lecturer related links additional papers and examples etc As the system already knows the limitations of the device the Packaging mod ule selects with certain confidence what part will be more useful and important during the seminar for example only the presenta tion In order to fit the device memory the system also asks the Presentation Adaptation module to resize the images used Be fore the presentation the chosen part of the material is seamlessly uploaded to the user s PDA and is accessi
198. rt of ELDIT to develop for mobile de vices in the sense of their usefulness for the users and interest for mobile learning experimentations About 90 persons have com pleted it up to October 2004 Some of the outcomes were valuable for the development of the mobile version and will therefore be listed here In Figure 8 one can see that users that had used ELDIT dur ing their preparation for the bilingualism exams that are more than 50 of all users find the system very or quite useful 55 50 y 9 0 i Bn ee eae eee DB Figure 8 Do you find ELDIT useful for preparing the bilingual exam a Level AB and b Level C 1 Very useful 2 quite useful 3 not very useful 4 useless 67 CHAPTER 3 RESEARCH CONTEXT This positive attitude made us believe that a mobile version of this system would be used and will be useful for the users prepar ing for the exam of bilingualism Figure 9 shows that the main consideration in understanding an unknown word falls almost equally on definitions examples and translations In the mobile ELDIT they are presented on the first screen for word entry see on page 74 Figure 17c The rest of the information is presented to the user only if specifically requested structure schema in definition 9 80 examples 27 80 Figure 9 What is considered mainly for understanding words meaning Figure 10 and Figure 11 show the declared by the users usag
199. s are discussed to 60 gether with predictions of their impact on foreseeable learning scenarios What differs drastically in this work from our point of view is that the mobile platform functionalities are a direct map ping of the functionalities of an e learning platform and only those that are impossible to deliver are excluded In our opinion is important to foresee the support also of new services that are proper only in the mobile case like location dependent services In 67 context awareness architecture for mobile learning is presented Similar to our Context Discovery module their Context Engine is responsible for gathering the context data A very good description of context is given in a hierarchical struc ture with the notion of context states and sub states dynamics and historic dependencies of processes The main difference from our Context Discovery is that authors suppose that all the context information is collected on the mobile device including data ob tained from sensors Though we basically agree that very often the mobile device is active participant in the process of context discovery in our vision some context data can be extracted di rectly from the infrastructure i e location and will not always require adding extra load on the device see Appendix B Also in our opinion to have easily extensible system we should sup port the presumption that the context data might be needed in dif fere
200. s from technical specialties were more then 60 Also here the main age group was under 25 78 3 of the partici pants and the boys were more than the girls 61 vs 39 Dif ferent years of study were almost equally covered 3 1 2 Availability of devices their usage and attitude to prices Devices One of the main concerns when trying to introduce a new service or technology in our case m learning is who will be able to use it This is partially dependent on what devices will be used how many users posses those types of devices but also if 28 the users are prone to spend money for acquiring a device if a new service appears that needs one Our study shows that more than 95 of Italian students and almost 90 of Bulgarian ones posses a cell phone see Figure 1 Personal computers are also often owned by participants of this study 81 2 of Italian and 75 3 of Bulgarian students In ad dition Italian students often have also laptops 55 only some have PDAs 7 videophones 7 and Smart phones 3 7 These percentages are quite small for Bulgaria less than 2 Availability of devices 100 7 90 4 80 70 60 4 50 4 40 4 30 4 20 4 10 4 top top tohet 96 1 81 2 55 0 o L L ae a Ee g O a Va amp lt O c RO Q es lt L oe x s o R oO we ae B Italian participants m Bulgarian participants Fig
201. s in the students attitude to m learning connected to studied subjects and specialties nor to nationality Nevertheless all students expect a strong support of wide variety of services well developed and often updated m learning platforms with strong integration to e learning solutions In our opinion the results presented here show clearly that the future of mobile learning is bright though lots of effort should be done to satisfy students high expectations to ensure a high rate of its utilization 49 3 2 General mobile learning architecture As mentioned before e learning is growing very fast and many Universities and companies are already supporting in some way an e learning solution Online courses web based education computer supported training and even virtual university are al ready wide used terms The rapid development of wireless infra structure and the advent of mobile devices in people s everyday life push the research in uniting those two domains which results in the emerging of mobile learning Considering the functional ities of e learning system we analyse the possibilities to extend it to provide services for mobile devices This includes distributing didactic material user identification and authorization gathering of data relative to the user system interaction provisioning of mobile services etc We find suitable an architecture that provides interoperability between the e Learning Management System eL
202. s to support automatic analyses on user behaviour and using the extracted knowledge in the hoarding process Nevertheless as there was no previous research in the particular context described above on which we could base our experiments we were forced to combine the automatic knowl edge extraction with semi manual and manual analysis combined also with some questionnaires and interviews with the users of our system Based on them we extracted some important knowl edge and characteristics that were further used for improving the automatic hoarding or for confirming the correctness of the knowledge automatically extracted In order to be clear further we list them here and give details right afterwards Measurements of the overall usage times and number of texts read in a single session Noticed random behaviour on first access Noticed consecutive browsing behaviour Reported changes in the behaviour with time Reported importance of missing words Noticed different importance of different types of words Reported usage of additional material and notes taking Noticed differences based on the target language Overall times and number of texts read in a single session Gen erally people were using the system between 10 and 40 minutes On Figure 30 we show the session length distribution that was ex tracted automatically from the system Important to know is that 108 the users preparing for the exam report spending longer peri
203. scuss the general and concrete approaches to develop and build it Mo tivations for our choices are given on every step We describe in details the hoarding problem and the strategy to solve it with the goal to provide an efficient hoarding solution Experimental re sults are presented together with the practical experiences gath ered from the interactions with the users Finally suggestions for improvements and further research issues are given Keywords Mobile Learning Hoarding Offline Access to Learning Content Disconnected Operations Caching Pre fetching General Ap proach Concrete Techniques ii Acknowledgments First of all I would like to thank my family my parents and my brother for always supporting me morally even from the long dis tance that is separating us now There were times in which with out the conversations and discussions with them I wouldn t be able to go ahead and reach this moment They helped me pass the hardest minutes and hours I had during these four years abroad four years of new experiences new cultures and unexpected situations I only wish I could share with them more of the nice and pleasant periods that I have had here I would like to thank a lot Judith Knapp who not only provided me a Strong basis for my thesis but also shared with me her valu able experience in non technical details of the research life She gave me priceless advices on the ways research goes on how to communic
204. se the learners and the instructors can be physically separated they may never or rarely meet for face to face lectures discussions etc and thus the whole learning proc ess is technology mediated In the second scenario the traditional learning approaches can be supported with complementary ser vices like online delivery of the learning materials support for collaborative work virtual communities etc In many cases both aspects are simultaneously present The goals of e learning sys tems and the functionalities they offer can differ the needs and goals of know how transfer in an industrial company are quite different from the educational needs of a university The func tionalities can be broadly grouped in four categories access to re sources data specific e learning services common services and presentation We intend to first list the main services and then discuss how these services must be modified with the introduction of small ubiquitous devices e Resources Support of learning objects LO any digital material link to other resources active element like simulations etc Break ing the educational content into small pieces allows modular ity and reusability of the content These chunks of digital re sources can be rearranged in modules like lectures and courses To facilitate this process they are usually described by additional metadata as prescribed by the IEEE Learning Ob ject Metadata LOM standard for
205. sing web service interface to access these functionalities allows flexibility interoperability and possibility for extension In this section we present an architecture that will provide access to learning materi als and other services to users equipped with mobile devices Our goal is to have an architecture which is 55 CHAPTER 3 RESEARCH CONTEXT a General to be able to provide all possible services of fered to the e learning users from the corresponding eLMS but also to support services that are new in the mobile context b Generic to support different mobile devices digital pones smart phones PDAs tablet PCs and etc with dif ferent characteristics and be easily extensible for the new generation devices To achieve this goal we believe that the mobile system should sit on top of the traditional e learning system and to provision adap tation of the existing e services like user identification authori zation distribution of didactic material gathering of data related to user system interaction and etc In addition it should take care of mobile specific services On one side we have the mobile device which will request access to the mobile system from a web browser WAP browser or a specific application On the other hand we have the eLMS which exposes an in terface to the services it provides We note that only some of the possible services are shown on Figure 5 In the business logic layer these services m
206. sion This pushed us to the second possible approach namely analysing and extracting knowledge from the log files gathered only on the mobile system Nevertheless it is important to have the awareness that the user behaviour on the online desktop system might differ some times even drastically from the one that the user will have on the mobile system In fact in the interviews we had with the users of the Mobile ELDIT that were familiar and were using also the 100 desktop ELDIT people share those differences and we report them further on in Section 5 4 5 2 2 Approach 2 using the mobile system One of the advantages that mobile learning gives us compared to e learning is the possibility to easily distinguish one user from another In e learning environments the problem of having multi ple users using the same computer or the fact that often the users are behind a proxy server is generally solved by asking the user for username and password on every session In a mobile learning system one can have the advantage that the mobile devices cell phones and PDAs are very personal devices generally used only by one person thus the problem of identifying the user which of ten appears in web based systems is much looser here The log files might be collected on the server side mLMS with bigger certainty for correct identification of users than in the general e learning case though the problem of the device being behind a proxy still exi
207. ssibil ity to analyse the user behaviour based on the system usage In this sense one should take also here the advantage of the possibil ity to analyse and extract important knowledge about the user by using the information from the tracking data that is possibly gath ered in the e LMS This is especially important when the same user might be using the online and the mobile version of the sys tem in the same period of time but it might be also useful to get familiar with the general behaviour of the users in the system it self 5 2 1 Approach 1 using the online desktop system In our case this approach was inapplicable as we met the problem of identifying the users of the online system In ELDIT there was no specifically developed tracking subsystem on the server and a number of different servers were responsible for different parts of the platform like authentication content generation collection of statistical data and etc Another particularity was that for access ing the online ELDIT the user had to register and log in but in practice the system was developed in such a way that the users were not obstructed to register with different username on every usage As tracking the users was not one of the objectives of ELDIT it seemed not to be a problem for the online system On the other hand this led to the impossibility uselessness of analys ing user behaviour based on the online system which to be used as a basis for the mobile ver
208. ssons learned Computer Vol 36 Issue 9 Sept 2003 p 30 37 101 Team Results of the Fall 2004 CALS Mobile Learning Survey 2004 http www cals ncsu edu cfprod apps calswebsite filelibrary PDAreport pdf last checked on September 14 2005 102 Traxler J 2002 Evaluating m learning European Workshop on Mo bile and Contextual Learning 2002 Birmingham UK 103 Tretiakov A amp Kinshuk 2004 A Unified Approach to Mobile Adap tation of Educational Content Proceedings of the 4m IEEE International Conference on Advanced learning Technologies 2004 August 30 Sept 1 2004 Joensuu Finland IEEE Computer Society pp 101 105 104 Trifonova A Georgieva E 2005 Determining the Readiness for Mo bile Learning UNITN Technical Report available online at http eprints biblio unitn it 105 Vainio T amp Kotala O amp Rakkolainen I amp Kupila H 2002 To wards Scalable User Interfaces in 3D City Information Systems 4th In ternational Symposium on Mobile Human Computer Interaction 2002 Pisa Italy pp 354 358 153 CHAPTER 7 CONCLUSIONS AND FUTURE WORK 106 Vavoula G Lefrere P O Malley C Sharples M Taylor J 2004 Producing guidelines for learning teaching and tutoring in a mobile en vironment In International Workshop on Wireless and Mobile Tech nologies in Education 2003 March 23 25 2004 Jhongli Taiwan 107 Virtanen V amp John
209. stically increased access from the university increase of about 40 is due to the increased number of computer halls for free access for enrolled students This is on behalf of usage from Internet cafes Another assisting factor is the augmented speed of Internet con nection in University s classrooms With the advances and with much more publicity of mobile technologies it seems natural that also the interest for trying their application in education will in crease Very similar to our study is one on the actual usage of mo bile technology by students done at the Norwegian University of Science and Technology Trondheim 25 The study shows that mobile phones are widely used in Scandinavian countries while PDA devices are still limited We should mention that in contrast of our survey the one presented here is a small scale only 25 participants who participate in the same course The study was oriented to improving the communication and collaboration be tween students and teachers through mobile technologies which is specifically needed for the concrete course In fact students ranking of the services they expect to be provided by a provi sioned mobile learning system does show that they need more collaboration Nevertheless their study shows very similar result to ours about the mobile phones usage i e almost everybody has a mobile phone and the main functionalities used by the students are phone calls and SMS WAP calendars sy
210. sts However in this case the offline periods will not be covered So we found a much better solution in conveying the tracking task onto the mobile device the tracking data is stored locally and when connection is available is transferred on the server In Mobile ELDIT the user was asked to give some initial information about himself only on the first interaction and later is freed from any direct interaction with the system In this way the identification of the user is done once when the system is set up and is included into the log files name TextsList 19 11 04 19 10 02 a it c general 032 22 11 04 10 06 25 1 it n accoglienza 1 derivl pbsO 22 11 04 10 07 25 1 it n accoglienza 1 full 22 11 04 10 07 47 6 it v compiere 1 lemma 22 11 04 10 10 10 1 it v essere 1 lemma 22 11 04 10 10 56 o URL Date Time Delay Figure 27 Log file collected from the device side proxy The architecture that was developed for the mobile ELDIT and described at Section 3 3 3 comprises a client side proxy see Figure 13 on page 56 which collects the tracking data The user 101 CHAPTER 5 CONTEXTUALIZATION AND OUTCOMES interactions with the system are written into log files and contain the visited URLs on the mobile device together with the date and time when the links were clicked as shown on the Figure 27 above For our system a database was created where the needed data to be hold including the tracking data aggregated values and etc Fr
211. study time in the sense of students being much more unproductive and thus is out of con sideration Other repeatedly given argument against m learning is that PDAs and cell phones are not able to give more to a learning sys tem than what already exists in e learning e g Internet is good enough Better e learning I believe it will not be that handy as e learning and I don t find it useful to do with the cell phone or a PDA in much less comfortable manner what I can do with a laptop Further students mention as obstacles the high cost of the devices and the connection the small devices are not enough technologically advanced to be useful for education mentioned the small screens small space etc not comfortable for long pe riods of usage might lead to health problems e g May cause eye troubles Other interesting answers I hate the computer I would use it only if it does not substitute the professor I will feel myself followed From Bulgarian side there is more potential interest more than 80 of the participants would like to try m learning The most frequently mentioned reasons are that there are no limits in terms of time and place it will give more easy and convenient ac cess to learning materials this is a modern educational technol 36 ogy it will be attractive and useful Students often mentioned more than one reason to use mobile learning On the other hand the
212. system work in most of the cases worse then a LRU least recently used algorithm enhanced with some heuristics Another system for experimenting with the hoarding prob lem is the WebScrooge Hoarding Agent 14 which deals with the Web browsing Their strategy includes user defined priorities re cency of document use and predicting of access patterns in order to provide reasonable Web availability during periods of low or intermittent connectivity The user requests are captured by a lo cal proxy which first searches in the local cache for the requested page and if the entry is not found it is retrieved from the server and meanwhile stores it locally In their system the hoarding agent is called profiler a module that is responsible for peri odically calculating the priorities of the cache entries and keeping the cache in a state of equilibrium The tests were done with hoard set size of about 25 of the full URLs set Though on the first sight this system seems very similar to our work it has major 133 CHAPTER 6 RELATED WORK differences First of all the hoard updates of this system were in terms of seconds 10 to 240 This means that disconnection pe riod they suppose if exist are very short Their experiments show also better results with frequent hoard updates about 30 seconds They get about 31 better performance in the sense of hit rate comparing to LRU algorithm As we discussed earlier in our case the LRU is p
213. t if the user will be viewing objects LO and LO it is most probable that the object LO will not be viewed This in some cases we can lead to in creasing the set priority of certain material while in other cases we can set it to much lower level which will sometimes lead also to exclusion from the hoarding set Note that for the example above we considered only asso ciations with confidence 1 and any support greater than 0 In real 93 CHAPTER 4 SOLUTION OUTLINE situations the best values for these parameters should be experi mentally discovered Generally the confidence value of the dis covered associations can help also in placing the items of the candidate set in an ordered list Also other data mining and or machine learning algorithms should be considered and tested to see their appropriateness for the hoarding process and how they can be combined best When no other rules can be applied the possibility to fit the predicted set into the limited device memory should be checked If there are still too many LO with the same priority that are pre dicted to be uploaded the choice should be done randomly 4 8 User modeling There are different ways to model user behaviour depending on the application and its needs In the context of hoarding we rec ognize two groups of characteristics that will be used differently in the hoarding process We schematically call the first user be haviour and the second user knowled
214. t the learner accessed In some cases might be possi ble to consider with a certain confidence that a portion of the material which the student reviewed is mastered or we can do some mining based on how long the user needed to review this particular portion In other cases it might be very important to look also at what the learner could access but decided not to view It is important to consider that on the first interaction the user is commonly unfamiliar with what can be done what interac tions are allowed what will be received on different actions etc This means that user actions might be based on his her curiosity rather than driven by his her knowledge or by the content This leads to the assumption that the mining on the data gathered by the system on the first user knowledge should be more attentive and extracted rules might be unreliable 4 4 Predict the starting point As mentioned in the previous section the web based learning ma terial provided by an educator will generally be structured in some manner and will have a certain starting point or index page shown on Figure 22 gt Starting Index J Page 85 CHAPTER 4 SOLUTION OUTLINE Figure 22 Web based material structure This is the starting point of the learner for his her first learning session with the system It can be also often a starting point of every following session especially if
215. t was studying and practic ing their language knowledge without the goal to pass the exam The tracking of their requests was gathered and analysed the out comes of which are presented further in section 5 Contextualization of the Solution and Experimental Outcomes 77 Chapter 4 4 Hoarding Outline of the Solution We define hoarding in the learning context as the process for automatically choosing what part of the overall learning content should be prepared and set available for the next offline period of a learner equipped with a mobile device We can split the hoard ing process into a few steps listed here that we will discuss further in more detail l 2 Predict the entry point of the current user for his her next off line learning session We call it the starting point Create a candidate for caching set This set should contain related documents objects that the user might access from the started point we have selected Prune the set the objects that will probably not be needed by the user should be excluded from the candidate set thus mak ing it smaller This should be done based on user behaviour observations and domain knowledge Find the priority of all objects still in the hoarding set after pruning Using all the knowledge available about the user and the current learning domain every object left in the hoarding set should be assigned a priority value The priority should mean how impor
216. talian languages All texts are divided into two difficulty levels and are split into thematic groups The users might browse through a number of texts and connected words that are previously packaged and saved in the cache of the PDA In contrast to ELDIT it does not contain all ar bitrary searched by the user word but only the ones pre fetched in the device local memory As the system is under development the number of misses the items that the user wants to see but are not available might appear Mobile ELDIT works offline utiliz ing a caching proxy called FoxyProxy thus does not require Internet connection The proxy provides access to texts and words entries leaving to the user impression of working online It also collects in few log files information about the browsing paths of the users and times spent on particular pages Regular backups of these log files are done on the external memory if such exists The system should be utilized through a special browser called NetFront3 which should be installed on the device see Instruc tions section further on 158 PRE REQUIREMENTS There are practically no pre requirements for using Mobile ELDIT Basic skills on web browsing and PDA usage are enough What mobile devices could be used to utilize M ELDIT The system is tested ONLY on few PocketPC devices Acer n10 iPAQ 3800 and iPAQ 1940 Theoretically the system us age is limited to devices for which the ewe virtual
217. tant the object is for the next user session and should be higher if we suppose that there is a higher probability that an object will be used sooner Sort the objects based on their priority and produce an or dered list of objects Cache starting from the beginning of the list thus putting in the device cache those objects with higher priority and con tinue with the ones with smaller weights until available mem ory is filled in 79 CHAPTER 4 SOLUTION OUTLINE An effective hoarding system will highly depend on the system s knowledge about the specific user for which materials are to be prepared Thus the hoarding process should be split into two parts 1 the first interaction with the system when no knowledge is available about the concrete user and 2 every next after the first access when the system has some knowledge about the user and continuously gathers more on each iteration This system s knowledge includes user preferences learning style personal learning abilities the level of expertise in the studied field and topic It can be acquired in different ways by direct assessment of the user by questionnaires and quizzes but also by observing and analysing the user behaviour during his her usage of the sys tem thus automatically discovering user s learning style prefer ences acquired knowledge etc We should point out that our cur rent work is mainly focused on this last mode automatic gathering o
218. tc Some sources show see section 6 7 of 73 that about 95 of all the lexis in a text should be known to the learner in order to have good comprehension Facts like this one can help hoarding but we were not able to reach this point in the current work Another example is to add knowledge about the learning styles strategies and sequences in order to help the hoarding pre diction For example as subject of our prototype was language learning one can study different theories of language learning and find out pedagogical rules to be used in hoarding In 56 for in stance is stated that there is a natural order in the acquisition of grammatical structures regardless of the first language of a speaker 140 Cooperative Hoarding within Ad Hock Networks Ad hoc networks have been frequently discussed recently as a re search issue The idea to connect devices between themselves and using each other s services or resources is not new For different learning scenarios we can easily imagine students with the same study classes being also physically close to each other Generally students that follow the same subjects will need similar often overlapping learning materials Base on this assumption the physical closeness of students that might need similar materials it would be interesting to experiment with collaborative hoarding in ad hock networks The ideas for two possible approaches are sketched in 60 though not considering th
219. ted rules and to add intelligence to our system thus to improve the hit_rate and to decrease miss_rate see Section 4 1 Later steps were based on in depth user behaviour observa tions Some interesting and important outcomes from those ex periments which are not directly related with the hoarding proc ess can be also found at the end of the section 5 3 1 One User Hoarding For obtaining the first hoarding results we observed only one user at a time We gave to the user a short list of available texts and we considered this to be always the user starting point as described in the general algorithm discusses in Section 4 see page 79 for reference Then for creating the candidate set we selected all the words that were accessible from the chosen text and then we did pruning based on what we thought the user already knew One can see that in this experiment Step 1 was solved in the sim plest way limiting the user choice to a small number of texts which were all pre fetched on the device and thus eliminating the need to predict the starting point Our goal was mainly to test the automatic pruning which meant essentially to discover automati cally the user knowledge set We used the following assumption 105 CHAPTER 5 CONTEXTUALIZATION AND OUTCOMES the user knows all the words that were presented to him her in a previous text and whose links were not followed In this way on every next iteration more words were pru
220. tem 101 5 3 HOARDING RESULTS c cccesesssceesssececsescececseeeeceesaeeecseaaeeeeseeeeees 105 3 3 1 Ones Ser HOGTAING no ienie ei 105 5 3 2 User behaviour observations oecon 108 5 3 3 Hoarding with Critical Set osses 113 5 3 4 Combined Hoarding eeeseseeeeeeeeeeeeeereereeresreesereees 117 3 3 ASSOCIAHON Rules aare de a eara ia e Rowdies 123 5 4 OTHER OUTCOMES FROM THE MOBILE LEARNING SYSTEM 124 5 4 Positive OULCOMES oir eenen 125 D 4 2 PrODLEMS Found vice cs Suvaiincscceacesieagereseesessastanud a ia 126 6 RELATED WORK iecsssscscisssessccccdsvassssvssiocsescsssoesecossicucosesbessesssdecess 131 7 CONCLUSIONS AND FUTURE WORK j 00 csscccessssccessseeees 135 7 1 HOARDING IMPROVEMENTS cssccccsesseeecseececeeseeecsesaeeeesneeeenees 137 7 2 MOBILE ELDIT IMPROVEMENTS ccecsesssceeeceeeesssaeeeeeeeeeeses 141 7 3 OTHER RESEARCH ISSUES ccsssccccessseceessececeessececseeeeeenseeeeees 142 BIBLIOGRAPHY sci isescsscosivsssessiacsessossecesscesuevensestassscaseavevspiessessossensees 145 APPENDIX A Mobile ELDIT User Manal scsscccsssscesscesseeessce 157 APPENDIX B Context dependent services in an m learning environment the printing CASC cccccccercsesescscsrssssccscecseessesscessesssees 177 APPENDIX C LIST OF PUBLICATIONS ccssscssscesssscssccesseecnsce 187 vi List of Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8
221. that you keep the memory inside the slot while using Mobile ELDIT so the system will do regular backups of the col lected tracking data For details on what is collected see the FAQ Why the proxy is needed You should enter the type of the memory you use in the follow ing file On the PDA Program Files ewe mELDIT Storage_Name txt You can open it on the mobile device by simply clicking on it 164 NOTE On different devices memory types and OS versions the name of the external memory will vary Some possible names are CF Card SD Card Storage Card Scheda SD CF etc You can see the correct name of your memory in the File Explorer on the mo bile device You can recognize it by the icon 1 If different users will use the system with the same device please read the instructions on how to set up usernames HOW TO SET UP USERNAMES There are two modes to set up user names 1 probably better choice Set the user name in advance and do not allow the user to change it 2 Let the user choose each time his her username For 1 before starting the proxy you should edit the file User_Name which you will find in On the PDA Program Files ewe mEldit It is a simple text file and can be edited directly on the device The name might be actually a nickname but in any case it should be a string with no spaces Example AlbertoCattaneo or Albertol It s up to you to decide the username generation procedure
222. the decision about what devices to use in the project An analogy and differentiation is made be tween e learning d learning distance learning and m learning and in this context they try to foresee the future of m learning and the methods and technologies that should be used for successful m learning In the attempt to find the best way to apply mobile devices in education people are experimenting with different fields one of them is language learning At Stanford Learning Lab 92 an exploration of m learning has been done by developing proto types that integrate practising new words taking a quiz accessing word and phrase translations working with a live coach and sav ing vocabulary to a notebook They envisioned that a good ap proach would be to fill the gaps of time by short from 30 seconds to 10 minutes learning modules in order to use the highly frag mented attention of the user while on the move The research in dicates some very useful directions like the length of the learning materials the personalization of interaction and the frustration of the user and low perception of the learning materials because of the poor technological implementation i e poor navigation through the materials poor cellular connections etc One thing often discussed in e learning field is adaptation of the learning content both in the ways it is presented and its structure to the specific learner s needs Logically the research on adapt
223. the meanings are provided one after the other starting with the most common one and giving the above mentioned information for every meaning It is possible that the word link provided from a text leads directly to the con crete sense which is used in the context of the text and is not the first one in the list If the word used in the text is a derivation form a certain words it is possible that the link leads directly to a screen like the one shown on Figure 6 In the word entry page lots of words are also clickable and connected to the basic word entry to the chosen word i NetFront v3 1 a d 1918 amp http www mobileeldit com it v y amp COMPOSTI Senso 1 Musik 3 die musicassetta Musikkassette die Pana a a var u dZ Ag NetFront v3 1 4 4619 21 amp obileeldit comyit n musica 1 full Qy amp amare la musica Example Mio marito ama molto je musica o3 anny colleziona atschi d quatsiasi tho e genere musicale studiare musica Example 2op averci tifettuto 2 lungo Mawo ha deciso of studiare musica e o imparare a suonare uno strumento File View Tok 22 6 Z 3g aj File View Tools 2 Ss zg E Fig 5 Sample word entry Fig 6 Sample word entry Full Collocations Derivations As mentioned at the beginning of the word entry page links might be provided to additional information On the Figure 5 one can see part of the full entry of the word musica
224. ther answers include expectations of much increased accessibil ity of learning materials real time information better time sched uling time saving allowing more freedom and flexibility Some students describe their view of m learning as a way to substitute traditional learning e g I will not travel 30 km to participate to the lectures or I will be able to watch the lectures while lying on the grass near the lake or travelling in the train 35 CHAPTER 3 RESEARCH CONTEXT In some cases students do not think they belong to the po tential users of m learning but still have a positive attitude for example It might be useful for those who can not participate in lectures nor use e learning others say they have not the possi bility to use m learning as they don t have the needed devices On the other hand quite often the explanation of a negative attitude is I m not interested or I don t see it useful There are students who prefer more traditional approaches books paper notes taken by hand and etc These are generally the same people that do not use e learning for the same reason Also here as often mentioned for e learning some students feel that there might be the lack of interaction between teacher and students more than 25 of the negative answers There is also the opinion that this new technological approach will bring more distraction than con centration generally will increase the
225. this index page contains an ordered reference of other materials like lectures sequences ex ercises etc A possible approach for predicting the starting point of user sessions is to keep statistics on what is the starting point of a session considering what the end point of the previous session was Our initial experiments on mobile ELDIT see Section 5 3 2 show that after the first learning session which we con sider almost unpredictable and rules that could be extracted by analysing it are unreliable the users generally show a very co herent behaviour if a list of materials is presented to a user he she almost always starts from the first item of the list then goes to the second to the third and so on It is also valid for the sessions the user continues from the point where he she finished last time This rule is rarely changed and if it happens it is based on some specific interest of the user For example we were giving a list of texts that were thematically grouped and the users were generally browsing starting from the beginning of the list In in frequent cases when a certain topic was especially interesting to the user he she was skipping the previous subjects and reading directly what was of interest and later returning to what was skipped We can not be sure that in every kind of learning mate rial the users will show the same sequential behaviour We how ever believe that supposing a continuous user browsing is a g
226. ting point of the students first session This might be an index page or a list of all lectures of the course Based on the observa tions of all previous users the system can be aware of often used sequences of material used on first request and can also estimate the average or maximum depth ing their first session Still it might be that users have very differ ent behaviour In the context of pre fetching the content on the first user access the system should hoard as much as possible data trying to satisfy all user as shown on the figure below s requests 5 amount of basic data texts and much auxiliary material diction In a system like m ELDIT this means to deliver only a limited ary entries Set of LO used by the student Set of LO selected by the hoarding algorith The hoarding starting step Figure 21 The system can try to detect the user expertise level on the study topic by a questionnaire for example and to narrow the hoarding set using some domain knowledge e g if certain material should be proposed to beginner users and if the current user is advanced the material should be excluded from the hoarding set An initial evaluation of the user knowledge could be provided by the educa 84 tor assessment though it is out of the scope of this work As we are mainly interested in extracting automatically important knowledge about the user we would like to look at the tracking data of wha
227. tion 3 2 2 M Learning According to the definition for m learning we adopted and de scribed in section 2 1 mobile learning can be viewed as any form of teaching or studying that happens when the user is interacting through a mobile device It might include various scenarios and here we try to transfer the services provided by an e learning plat form enumerated previously into the mobile context We can easily see that there are services that need to be adapted to fulfil the limitations of certain devices there are other services that are infeasible to transfer but also new services appear provoked by the mobility The connectivity is one of the main differences if we compare a mobile device with the PC the usual medium for delivering e learning Nowadays mobile devices might be connected to The Net via lots of technologies WAP GPRS UMTS Bluetooth WiFi etc Although it is predictable that in the future always on will be wide spread currently it is not the case Mobile devices often have periods of disconnection either intentionally when the connection is too expensive or not when no infrastructure is provided 54 Devices hardware and software characteristics have a big im pact on what content is possible and meaningful to be delivered Usually the web content is designed for desktop PCs thus un pleasant and even rarely useful from a small screened device Nowadays mobile phones became more powerful w
228. to the general architecture mentioned that demonstrates in practice the research issues triggered through the thesis Explana tion of why and how it was developed is given together with some technical details about its modules 3 1 Survey on the readiness for mobile learning To start planning the development of a mobile learning system we analysed the available literature to benefit from other people s experience Looking at the wide variety of mobile learning pro jects their aims diversity of devices used and even the target au dience we were wondering if the previously gathered knowledge applies in our scenario In this context we tried to investigate the factors that might help in predicting the success of an m learning application in a more concrete situation the University envi ronment We performed a study on the readiness of the University students for using mobile technologies in their study process The survey was done in May June 2005 in parallel at the University of Trento Italy and the University of Ruse Bulgaria An online questionnaire was developed and the students of different facul ties were asked to fill it in For the majority of the questions par ticipants were asked to choose from predefined single or multi choice answers However for some questions they had to write an 27 CHAPTER 3 RESEARCH CONTEXT explanation of their opinion argumentation for certain answers and also to describe their idea view or
229. try c right bottom m ELDIT basic word entry screen 74 We have converted the screen on the left of Figure 17a into a series of interconnected screens on the mobile device When a user wants to see a word entry first the main screen is displayed Figure 17c Afterwards the user might select to view more de tailed information Figure 17b by clicking the links that were added during the XLST transformations on the server The searching possibility is excluded from the mobile system because as mentioned before it would allow the user to request arbitrary words not only the one connected to previously viewed content This would force the inclusion of the entire dictionary as the user actions will be unpredictable which is not desired at this phase For supporting offline use of the learning material and for collecting tracking data a client side proxy is developed As men tioned before the proxy is responsible for receiving the browser requests and retrieving the content from the server or from the lo cal store cached pages in Figure 13 when there is no connec tion available at the moment The client side proxy could also seamlessly upload the content that will be used in the future based on the prediction done in the Hoarding Engine Generally uploading might be done on a special user re quest where the user might also be given an option of setting dif ferent parameters e g provisioned disconnection time ex
230. ually one of the particularities that mobile learning offers Often mobile devices are definitely personal devices used only by their owners or as in the Mobile ELDIT application we developed are borrowed for certain period of time but used only by one person during that pe riod it is possible to easily and securely identify the user 82 4 2 Definition of session in the mobile learning context In the Internet world a session is defined as a continuous period of time during which a user s browser is viewing Web pages or a Web application within the same server or domain source MSDN Library It is a series of transactions or clicks on the web pages links made by a single user There are different criteria to decide if a session is over or not The most commonly used one is the inactivity period of the user resumption of the activity by the same user after a timeout has occurred is considered as the start of a new session On the other hand for hoarding in a mobile system the im portance falls on the time between two possibilities of the user to synchronize with the main server In this sense we find more use ful to define a session as the time between two synchronizations of the mobile device with the main online system The default session length might be one day as commonly synchronization is generally done once per day but during the system usage other session lengths might be observed and explicitly set for every user
231. ucation 14th World Conference on Educational Multimedia Hypermedia and Telecommunications 2002 Denver CO USA Judith Knapp 2004 A New Approach to CALL Content Authoring PhD Thesis Universit t Hannover Institut fiir Informationssysteme 148 47 48 49 50 51 52 53 54 55 56 57 58 Jung Li 2004 Context Aware Support for Computer Supported Ubiq uitous Learning In proc of the 2nd IEEE International Workshop on Wireless and Mobile Technologies in Education WMTE 04 Ketamo H 2002 mLearning for kindergarten s mathematics teach ing IEEE International Workshop on Wireless and Mobile Technolo gies in Education 2002 V xj Sweden pp 167 170 Ketamo H 2002 xTask adaptable working environment JEEE In ternational Workshop on Wireless and Mobile Technologies in Educa tion 2002 V xj Sweden pp 55 62 Kinshuk amp Lin T 2004 Application of learning styles adaptivity in mobile learning environments Third Pan Commonwealth Forum on Open Learning 4 8 July 2004 Dunedin New Zealand Kinshuk 2003 Adaptive Mobile Learning Technologies Global Edu cator available at http www globaled com articles Kinshuk2003 pdf last accessed on Jan 16 2006 Kistler J Satyanarayanan M 1992 Disconnected Operation in the Coda File System ACM Transactions on Computer Systems Vol 10 No 1 February 1992 pp 3 25
232. ues tions about video lectures in an e learning system accessed via Internet or on CD Probably this led also to lots of the negative reactions to m learning as people see the devices not strong and comfortable enough for looking video on them In our opinion a different probably more positive outcome might be expected if the students are given more concrete scenarios and situations em bracing mobile devices in different learning processes It was noticed that quite a big part of those who answered that they would not like to use m learning still think that its exis tence will increase the quality of instruction about 10 Most of them are students that think that attendance at the lecture and im mediate contact with the colleagues and professors is very impor tant Nevertheless they consider m learning might be very useful if for some reason other students can not be present at the lec tures About 25 of those who declare they possess a PDA type device say they would not like to use any m learning system Reasons include lack of humanity and personality too much dis traction while learning demands good concentration A few peo ple declare that they would tolerate m learning only in the cases that it is additional support or in cases they are hindered from par ticipating in the face to face lectures 38 We must also mention some original students answers The quality of education doesn t increase in dependence of the place
233. ugeedickal m Passa o Ersen af Radern Kumai Musik Lism Kunst Musi Lisans Film Theater Film Theaser Se De Heuzang abgrechabet o Kammchen alt Angsthase o Huje anf dem Betochern o Buste B der aus Beton 3 o Hamd ale Finstar o Rokippden a Figure 32 Example of consecutive browsing behaviour of a user Behaviour goal changes with time Some users reported changes in their own behaviour a short time before the exam date For ex ample a user at the beginning was using the system mainly during the week ends and more often in the mornings also this can be seen from the log files He was reading texts in the language that is more difficult for him Italian as the goal was to learn new words As the exam date was drawing nearer the last 2 weeks be fore the exam the user was using the system much more often almost every day after work workdays evenings and was read ing mainly texts in the mother tongue German trying to answer the questions in the target language Italian The users that were not aiming at the exam do not report changes in their behaviour Importance of missing words In our first prototype version we asked the user to grade the importance of every missing entry Our initial idea was to try to distinguish a group of words that are critical for the understanding of the text and others that were not that important On every miss users were given a form for grading the importance of the miss wit
234. ulum Two School District Partner ships Computer Supported Collaborative Learning Conference 2002 Boulder CO USA 152 94 Sm rdal O amp Gregory J amp Langseth K J 2002 PDAs in Medical Education and Practice IEEE International Workshop on Wireless and Mobile Technologies in Education 2002 V xj Sweden pp 140 146 95 Steinberger C 2002 Wireless meets Wireline e Learning 14th World Conference on Educational Multimedia Hypermedia and Tele communications 2002 Denver CO USA 96 Stone A amp Briggs J amp Smith C 2002 SMS and Interactivity Some Results from the Field and its Implications on Effective Uses of Mobile Technologies in Education IEEE International Workshop on Wireless and Mobile Technologies in Education 2002 V xj Sweden pp 147 151 97 Stone A amp Briggs J 2002 ITZ GD 2 TXT How to Use SMS Ef fectively in M Learning European Workshop on Mobile and Contextual Learning 2002 Birmingham UK pp 11 14 98 Sushil K Sharma and Fred L Kitchens 2004 Web Services Architec ture for M Learning Electronic Journal on e Learning Vol 2 1 Feb 2004 pp 203 216 99 Symbian Limited 2005 Analyst Statistics Smartphone market over view Fast FAQs on http www symbian com about fastfaqs html last accessed on 12 12 2005 100 Tatar D Roschelle J Vahey P Penuel W R 2003 Handhelds go to school le
235. ure 1 Availability of devices The devices that are mainly involved in the research in m learning are mobile phones and PDAs So we asked the students about the reason for them not to have one The answers differed in Italy and in Bulgaria Considering PDAs for most of the Italians the main reason is that such device is not useful for them 59 4 fol lowed by the high price of the devices 41 6 expensive wire less services 13 9 devices limited resources 11 1 etc For Bulgarian students the major concern is the high price both for the devices and for the wireless services more than 75 of all answers and only after this comes the answer that the device is not useful for the respective student 12 2 and the limited re sources of the device 5 29 CHAPTER 3 RESEARCH CONTEXT Prices Considering participants attitude to prices of the devices the general opinion is that while personal computers and cell phones have acceptable costs for PCs more than 65 of all par ticipants and for cell phones more than 54 of IT and 60 of BG students the prices of laptops and PDAs are considered high by more than 52 Moreover lots of the students have no interest and thus no opinion about the prices of Smart phones more than 60 of Italians and about 50 of Bulgarians and PDAs 35 38 while the no opinion percentage is rather small for other devices as shown on Figure 2 below 100 90 47
236. users are numer ous and have different interests thus the prediction accuracy is quite low comparing to what is needed in our scenario but could be compensated by the fact that the Internet connection is perma nent The idea of hoarding for disconnected devices in distributed file systems has been first described in 52 though in contrast with us they do not consider mobile devices in the sense of PDAs but rather they consider laptop computers They propose the Coda File System to explore the usage of caching of data not for im proving performance but for increasing the availability They propose architecture for hoarding and for keeping the coherence of the utilized files The initial system was based on client server architecture which tracks the local file modifications and saves a Client Modification Log The project has lately evolved into UbiData project 38 and the direction taken is in double middleware architecture for ubiquitous data file access They introduce incremental hoarding where the idea is to use a version control system to maintain object differences and also study the automatic data selection problem A metadata server is included to store the users mobile profile which keeps a list of user files that are considered interesting They define a hybrid priority metric for choosing the hoarding set of files The hybrid prior ity is calculated by taking into account the recency of use the
237. ustering algorithm automatically discovered three clusters Still in most of our further experiments we forced two clusters to be produced As discussed earlier our current goal is to predict the user behaviour and needs Based on the clustering shown in Table 4 we experimented on using prediction algorithms The data was split into training 60 and testing set 30 We used the data collected for texts 1 to 5 and the clustering as shown in the Table 4 for predicting the grouping for the last text Table 7 k Nearest Neighbours Prediction value of k Valueofk RMS Eror RMS Error 1 0 273861279 0 433012702 2 0 273861279 0 433012702 3 0 273861279 0 348792827 lt Best k 4 0 273861279 0 387887144 5 0 273861279 0 405825766 6 0 273861279 0 371410282 The algorithm details about the algorithm can be found in 36 analyses what is the best value of k i e the number of neighbours to be compared during the prediction see Table 7 Table 8 k Nearest Neighbours Prediction correctness Predicted Actual Value Residual Value correct wrong correct correct correct correct The prediction results are shown on Table 8 Experiments with different random separation of the users into training and testing set gave error of 17 33 i e 1 or 2 out of 6 wrong predic tions The prediction based on 3 clusters and also manually 122 choosing the text over which the predi
238. usy one Another issue has to be considered To perform printing 180 from a device to a certain printer the device needs a driver for that printer A desktop computer can have installed on it drivers for all supported by the system network printers which for a mobile de vice is not suitable solution One could think of downloading on demand the needed drivers but sometimes installing a driver re quires rebooting the machine so also this solution is not sensible A second possibility is that info about the printers is kept on some server the mobile client could contact the server passing its own context info and getting back the indication of the chosen printer After all this is what is typically done in a multi user OS where printers are never directly accessible by the users to stay away from nasty concurrency problems and has the advantage of enabling accounting and permission checking The main draw back of having a centralized server is the scalability of the solu tion in terms of performance the central server becomes a bot tleneck reliability the server becomes a critical single point of failure and geographic scale it makes no sense of thinking of a central server that knows about all printers in town One can overcome these weaknesses in a standard way i e by having a federation of servers each being responsible for a sub region and being able to forward requests to other servers with some degree of replicatio
239. ved from the server A significant problem that should be mentioned though de scribed in the Pocket PCs user manual is that the battery of Windows based devices discharges quite fast When a device is frequently used it discharges in 1 2 days but the main problem comes from the fact that even when not used the battery discharges in about a week time The discharged de vice forgets the software installed by the user and all user s data This leads to the necessity to do backups of all impor tant data on an external memory quite often It is also very inconvenient as all the programs that were installed should be re installed FAQ Why should I use NetFront instead of IE browser The reason to use other browser not IE is that IE does not send requests to the local proxy when no Internet connection exists This does not allow the local proxy to provide offline access to the learning material Can I use other browser instead of NetFront3 Yes Other browsers could be used but the system is not tested with other There is a requirement that the chosen browser allows the usage of local proxy and the browser should be set up properly i e to send the requests to the lo calhost proxy on port 3128 A possibility could be to use Minimo Mozilla for Pocket PC which could be downloaded at http www mozilla org projects minimo However the last version we have checked v 0 0009 still had some problems with user interface
240. ver a certain printer or pages limitations a decision should be taken This could be done locally or on an ex ternal resource a server The system should find which is the most convenient for this user and accessible in that moment printer inform the user about the choice possibly giving him the possibility to modify this choice Finally the document should be printed optionally allowing the user to monitor the status of the printer queue Printing on the nearest suitable printer contains ele ments that can be found in many other mobility related problems What makes the printing problem so prototypical as we shall dis 179 APPENDIX B cuss in a later section is the fact that printing is a service and can be used from any application and it uses services those provided by the operating system There are various architectural choices that can solve the printing problem We shall briefly discuss them outlining advantages and disadvantages of each of them The first step is collecting context information In order to find the closest printer we need to know where the user device is i e their physical position But then occurs the problem that the nearby printer could not be always reachable some room might be locked at night or during week ends The behavioural infor mation is also important Knowing what the user is currently do ing and what applications is using at that time could provoke the usage of diff
241. vider Specified Web Clipping Approach for Mobile Content Adapta tion 4th International Symposium on Mobile Human Computer Interac tion 2002 Pisa Italy pp 324 328 114 Zenith 2002 Enabling The Mobile Work Force With Zenith s Mobile Application Extension MAX System Zenith Infotech White Paper available online Last accessed September 1 03 http www zenithinfotech com max whitepaper max_whitepaper pdf 115 Zhang J Helal A and Hammer J 2003 UbiData Ubiquitous Mo bile File Service Proc ACM Symposium on Applied Computing SAC Melbourne Florida March 2003 154 116 Zhao Gang Yang Zongkai 2005 Learning Resource Adaptation and Delivery Framework for Mobile Learning in 35th ASEE IEEE Fron tiers in Education Conference October 19 22 2005 Westin Indianapo lis Indiana 117 Zobel J 2001 Mobile Business and M Commerce Hanser Fachbuch ISBN 3446216189 155 Appendix A During the thesis manuscript we gave a detailed explanation of different aspects of Mobile ELDIT system but we concentrated on the technical perspective Here we would present the system from the point of view of the user A demo version of Mobile ELDIT was published online for free download in the beginning of 2005 and here is how the user gets introduction and instruc tions about usage and possible problems Mobile ELDIT User Manual Index GENERAL INFORMATION ABOUT MOBILE ELDIT 158 PR
242. w to the solution and is general so that it would be valid not only for our system as the concrete implementation might differ from case to case and in our view will often depend on the concrete system s specifics In this section the way we mapped the algorithm dis cussed before to the actual implementation in Mobile ELDIT will be described Both measurements and observations that we ob tained from experiments on the system are provided 5 1 Methodology for looking at the outcomes We have clearly shown by now that this thesis has the main goal of attacking the hoarding problem In the previous section 4 1 we described how we plan to measure the goodness and success of our hoarding strategies in terms of hoarding size hit rate and miss rate and we discussed pros and cons of other possibilities However the supplementary goal we have in the current work is to analyze the successfulness of the newly developed mo bile learning system Mobile ELDIT see 3 3 from a less tech nical point of view and to gather experience for further improve ments For this we have performed questionnaires and interviews with the users of our system at different stages Initial surveys have been performed for determining the students previous ex periences with similar systems with computational and mobile devices in general and expertise in the targeted languages Addi tionally a questionnaire was filled in by the users in a later stage 97 CHAP
243. ween users A possible ap proach is to calculate the User_similarity as a count of occur rences in the same group throughout different texts With the example data given above forced two clusters we receive the following similarity table Table 5 The most similar ones are 119 CHAPTER 5 CONTEXTUALIZATION AND OUTCOMES marked with shade that have been preserving their maximum values after the fifth and for sixth texts Table 5 Users Similarity U U2 Us Us Us Us Uz Us Us Uio Urs Ure Uas Ura Uis Uie U 50 50 67 33 67 33 67 33 83 67 50 33 33 33 33 U2 50 67 50 83 50 50 50 83 67 50 33 50 50 83 50 Us 50 67 83 50 83 17 83 83 67 83 67 17 17 50 17 Us 67 50 83 33 100 33 100 67 83 100 83 0 0 33 0 Us 33 83 50 33 33 67 33 67 33 33 50 67 67 67 67 Us 67 50 83 100 33 33 100 67 83 100 83 0 0 33 0 Uz 33 50 17 33 67 33 33 33 50 33 50 67 67 67 67 Us 67 50 83 100 33 100 33 67 83 100 83 0 0 33 0 Usp 33 83 83 67 67 67 33 67 50 67 50 33 33 67 33 U10 83 67 67 83 33 83
244. working in this new field and it is increasingly difficult to have an overview of what is going on since most papers are dispersed in many conferences and some reports are only available as grey literature Here an overview of what is going on is presented By 11 CHAPTER 2 STATE OF THE ART no means were we able to discover all the interesting papers that have been published in the field but from the inevitably partial view we tried to let emerge the trends that characterize the field 2 1 Defining M Learning There are number of different definitions of mobile learning Of ten m learning is described as e learning through mobile compu tational devices Alternative definitions emphasise on the mobil ity of the learner rather than the device Here we focus more on the first definition and while in general by mobile device we mean PDAs and digital cell phone more generally we might think of any device that is small autonomous and unobtrusive enough to accompany us in every moment in our every day life and that can be used for some form of learning We shall begin by enu merating the different ways such a device can help us In first place they can allow us to interact with people via voice and through the exchange of written messages still and moving im ages A second possibility is to consider them as tool for access ing content which can be stored locally on the device or can be reached through interconnection Under the sam
245. x y z and in another classification as a more semantically meaningful symbolic expression like floor and room number or Professor X s office Depends on the method chosen the accuracy varies The research shows 7 that the average error percentage in this second method is lower thus it is more reliable The position determining system that we use comes in two variants the first returns raw data physical coordinates the second returns a semantic description of the location The second option we find more suitable for our goal because we can explic itly take account of the local topology meaning walls aisles etc In a system where more than one positioning systems will be used one can think of introducing a semantic server which translates data from the format used by the device GPS WLAN Bluetooth into format proper for the server that offers the printing service We experimented with a system 6 based on the strength of the IEEE811b signal coming from different antennas that uses available hardware and infrastructure and therefore requires only adding a software layer In a conceivably wireless networked city such method would work indoors and outdoors 3 Generalization We have seen that the printing problem can be mapped on a more general one where the focus is on providing some context dependent service while using basic services provided by the in frastructure e g by the OS or by a Learning Mana
246. y can be done in dif ferent directions During the current thesis we did not get to the point of measuring the optimality of work of the algorithm rather the goal was to prove the correctness of the ideas and strategies for creating the candidate set and for pruning Throughout we tried to keep the steps as separate as possible in order to have a clear idea of the processes and the successfulness or the faults on every step On the other hand the concrete algorithm should be optimized and one idea for this is to try to optimize it by combin ing the steps of generating the candidate set and pruning it into a single step i e instead of first adding all connected to level n items of level n to the candidate set and later pruning what is not needed it is possible when an item is selected for including to apply at the same time the pruning rules ad decide on the fly However our work was mainly on testing possible techniques and we were targeting the improvements in terms of hoarding accu racy rather than optimal implementation Once the needed by the system accuracy is fulfilled many other further optimization of algorithm speed could be taken into consideration It should be mentioned however that most time consuming processes for the hoarding system might be done during users offline periods Usage Patterns Observations The hoarding process includes various steps one of which is the prediction of the user s starting point a

Download Pdf Manuals

Related Search

Related Contents

manuale rt  Model 805 STARTREK TELEPHONE  APP48-AP V 2.0 APPsystem Manual de Instruções  Bedienungsanleitung  Sistema de tuberías de extremo liso para tuberías de    Sweex Wireless LAN USB 2.0 Adapter 54 Mbps  Rollei Movieline SD-15  fiche-enseignant-Choco Leo  

Copyright © All rights reserved. Failed to retrieve file