Home

circumstantial knowledge management - e

1. Network SABATINI_2is loaded Figure E 3 Screenshot for the network management screen The components of this screen are detailed here APPENDIX E USER MANUAL 133 10 11 12 13 14 The network selection panel allows to choose an existing network When a network is selected in the box its description is shown The button select must be pressed in order to load the network The network removal button deletes the network selected in the box 1 from the system The network creation panel allows to create a new network with the specified name and description The network is created when the create button is pressed The map area shows the network plane and the network items nodes and links The picked elements are highlighted The load image button shows a dialog to choose a background image for the network plane The zooming slider allows to zoom in and zoom out the map area 4 The number under the slider is the zoom percentage over the image real size The node creation button will create a node in the place where the user clicks over the map area 4 The node removal button allows the user to select a node in the map area 4 to be deleted Moreover its input and output links will be removed as well The node motion button allows the user to move a node from one position of the plane to another The user must drag and drop the node in the map a
2. 4 10 Mockup for the networks management screen ooo o a 4 11 Mockup for the simulator screen o e 5 1 Classes for the ModelKnowledge subsystem o 5 2 Classes for the EditionKnowledge subsystem 9 5 3 5 4 5 9 5 6 5 7 5 8 5 9 6 1 6 2 6 3 6 4 6 5 6 6 A l C 1 C2 C 3 C 4 C 5 C 6 E 1 E 2 E 3 E 4 E 5 Circumstantial Knowledge Management for Human Like Interaction Classes for the DataGateway subsystem o 82 Classes for the Edition subsystem 2 nop a Seog Pa Bede de Sew 84 Classes for the Simulation subsystem 0 00004 85 Geometric state of the plane ir yl ee Ee fy a e 88 Classes for the presentation tier 2 1 a a 90 Relational graph for the model knowledge 92 Relational graph for the edition knowledge 93 Instructions for the experiment 4 9 0962 eg be ee basa he ee 99 Questionnaire for the experiment o 100 Results for the subjective evaluation comfortable 102 Results for the subjective evaluation intuitive 103 Results for the subjective evaluation reliable 103 Results for the subjective evaluation agile 103 Map of the campus illustrating the first scenario 116 Survey completed by subject 1 0 0 0 0000 eee eee 123 Survey completed by su
3. 53 Validation test 05 Plane selection o oo 53 Validation test 06 Plane visualization o o 54 Validation test 07 Network items management 54 Validation test 08 Agent simulation 55 Validation test 09 Simulation Services o o 55 Validation test 10 Routing and navigation services 56 Validation test 11 Big networks soii iso a 56 Validation test 12 Interoperability o o 57 Traceability matrix Sa dd A AA gale ae N 58 Costs projection for physical resources 2 023 04 eee 06 ees eee 62 Costs projection for human resources 1 a 62 Total costs projection Yt A ef 62 Routing instructions oia e RE RO L p E 89 Results for validation test 01 Database connection 96 Results for validation test 02 Situations management 96 Results for validation test 03 Features management 96 Results for validation test 04 Network selection 96 Results for validation test 05 Plane selection 96 Results for validation test 06 Plane visualization 97 Results for validation test 07 Network items management 97 Results for validation test 08 Agent simulation 97 Results for validation test 09 Simulation services 97 Results for validation test 10 Routing and navigation services
4. W 2006 Dialogue Systems Go Multimodal The SmartKom Experience In SmartKom Foundations of Multimodal Dialogue Systems chapter 1 pages 3 27 Springer Want et al 1992 Want R Hopper A Falcao V and Gibbons J 1992 The Active Badge Location System ACM Transactions on Information Systems 10 91 102 Weiser 1991 Weiser M 1991 The Computer for the 21st Century Scientific American 265 94 104 Weiser and Brown 1996 Weiser M and Brown J S 1996 The Coming Age of Calm Technology This page has been intentionally left blank
5. a e a oy eee eee Communicative Communicative Acts Input Acts Output Speech Display Recognition Manager Modality Processing Dialogue Management Problem Solving Manager Route Scheduler Simulator Temporal Finder KB Figure 2 1 TRIPS architecture Domain Planner Specialized Reasoners group communicates the previous two by managing the conversation and coordinating the reasoners TRIPS project solving manager is a component within the dialogue management group which deals with a database which stores the problem solving context and with specialized reasoners and whose purpose is to keep a representation of the task to be performed coordinate the generation of a feasible and coherent plan and maintain the state of the solution The other key component in TRIPS dialogue management is the conversational agent This component is in charge of dealing with the problem solver manager as it interacts with the user For this interaction to be performed the conversational agent considers the input by means of communicative acts which are interpreted along with the discourse context to discover what the user means In coordination with the problem solver manager it generates an output which can be expressed again as communicative acts 30 Circumstantial Knowledge Management for Human Like Interaction A new core architecture for TRIPS was proposed in 2001 James Allen and Stent 2001 in order to
6. e Direct edition of the model requires some technical skills or programming abilities and thus the edition tool is more convenient for unexperienced users However some experienced users may prefer the direct edition as it allows to perform some scripting tasks e Although both systems require some training using the edition tool is more intuitive as it is built in a common graphical style e The edition tool is more reliable as it shows the network graphically whereas in the case of direct edition input errors are more difficult to detect e Modifications over the network are easier by using the edition tool e The edition tool does not enable the creation of a node directly by using its coordinates it has to be created over the network using the mouse This functionality has been proposed by expert users as a future line so they can keep their current edition mechanisms and take advantage of the new functionalities and facilities Summarizing the edition tool is considered more convenient and fits its purpose as it enables fast knowledge edition and provides higher quality results with respect to the manual edition Chapter 7 Conclusions and Future Work Finally this last chapter presents the conclusions of the project Actually this project still has a long way to be improved For this reason the state of the project is revisited and future work is proposed in this chapter as well The purpose of this future work is to ens
7. setOf TGeometry 1 SitPoint setOf TPoint 1 0 1 1 Path Person z A 4 Represents IDPath CommonType S hy IDPerson CommonType _ lt Formed of TypePath CommonType Rare Situation setOf TPoint _ StartNodelD CommonType EndNodelD CommonType Cost CommonType PathGeom setOf TGeometry To be in gt Figure 3 1 Architecture for the previous situation model database For the purpose of this project several modifications will be performed over this design First while the cost for each node or each link is currently stored within the item table for this project costs will be stores in an external relation in order to increase the scalability with regard to the number of different situations having an associated cost The creation of views may be required to take advantage of the database spatial capabilities Additionally a characterization may be added to the network items nodes and links so that in the future some rule firing mechanism can be implemented over this characterization 48 Circumstantial Knowledge Management for Human Like Interaction to increase the performance and scalability of the system e g assigning cost factors to items features instead of doing so for items themselves Finally the database has to include some additional relations to support not only the model knowledge but also the edition knowledge in order to be managed by the tool 3 4 Feasib
8. such as train schedules cinema timetables or hotel bookings As the scenario information and the domain knowledge are closely related SmartKom integrates the situational and the domain knowledges together in what has been called the modeler knowledge module This module performs what is known as the decontextualization process which builds a context free statement based on a context dependent statement and on contextual infor mation This process receives dynamic spatio temporal information and a set of intention hypothesis based on the user utterances as input Then it decides what situational and domain information is relevant for resolving the contextual reference and computes scores for each different hypothesis based on its contextual coherence Finally the process returns a decontextualized intention hypothesis More information about the SmartKom project as well as some demonstration videos can be found in http www smartkom org start_en html 2 7 4 Apple Siri Siri is a personal assistant application which was introduced by Apple in the iPhone 4S smartphone in 2011 Siri implements speech recognition and the user can communicate with it in a natural way in order to ask for information to ask for an action to be performed 32 Circumstantial Knowledge Management for Human Like Interaction Pointing Sensor Gesture Interpretation Gesture Recognition Speech Recognition Speech Interp
9. 1 Template for software requirements specification 3 2 1 Functional Requirements This subsection describes the user requirements related to the system functionality PR Name heman ass model selection DE A r A E E system shall allow the user to select a situation model in order to work on it Cyce Fist Priority Esential Stability Sbe Table 3 2 Requirement FR 01 Situation model selection PR a a a s s s s s as a aaeeeo model management Description The system shall provide means to manage the situation model knowledge Cyce Firs Priority Esential Stability Sbe Table 3 3 Requirement FR 02 Situation model management FR 03 Situation model simulation Description The system shall include mechanisms to perform a simulation over the situation model Cyde Fist Priority Conditional Stability Stable Table 3 4 Requirement FR 03 Situation model simulation 37 38 Circumstantial Knowledge Management for Human Like Interaction FRO The system shall store all the model knowledge persistently Cycle Firs Priority Essential Stability Stable Table 3 5 Requirement FR 04 Persistent storage 3 2 1 1 Management This subsection describes in further detail the user requirements for the network manage ment functionality mD Fm Name Networks management Networks management DE The system shall provide means to the user to create and dele
10. 3 24 Requirement FR MA 19 Nodes properties FR MA 20 eee MA 20 AAA a a a a a e e e e eeaeee management Description The system shall provide means to the user to create edit and delete links within a previously selected network and plane Cyde Firs Priority Conditional Stability Table 3 25 Requirement FR MA 20 Links management Al 42 Circumstantial Knowledge Management for Human Like Interaction FR MAT Description The user shall be able to create links with a graphic interface by choosing the start and end nodes of the link Cyce Finst Priority Essential Stability Stable Table 3 26 Requirement FR MA 21 Links creation mpm moz Links properties Description Besides the properties common to all network items specified in requirements FR MA 08 to FR MA 13 a link may be bidirected or not Cyce Fisi Priority Conditional Stability Unstable Table 3 27 Requirement FR MA 22 Links properties PMA Edition interface zoom Description The user should be able to zoom in or zoom out the plane back ground image and the network items placed over it to obtain a higher precision in the edition process Cyce First Priority Optional Stability Stable Table 3 28 Requirement FR MA 23 Edition interface zoom RMA Situational knowledge taxonomy Description The system shall classify the situational knowledge in seven dif ferent aspects spatial temporal amb
11. 72 83 89 90 HCI Human Computer Interaction 20 25 IDE Integrated Development Environment 78 JUNG Java Universal Network Graph Framework 91 LaBDA Advanced Databases Group 16 78 106 NLP Natural Language Processing 25 26 NUI Natural User Interface 26 27 33 OCR Optical Character Recognition 26 PC Personal Computer 20 PDA Personal Digital Assistant 20 RFID Radio Frequency IDentification 44 55 76 113 139 140 Circumstantial Knowledge Management for Human Like Interaction UML Unified Modeling Language 24 VPN Virtual Private Network 127 WYSIWYG What You See Is What You Get 78 Bibliography Allen et al 1994 Allen J F Schubert L K Ferguson G Heeman P hee Hwang C Kato T Light M Martin N G Miller B W Poesio M and Traum D R 1994 The TRAINS Project A Case Study in Building a Conversational Planning Agent Journal of Experimental and Theoretical Al 7 7 48 Apple com 2012 Apple com 2012 Using Siri Do I have to say things a certain way to get Siri to respond http www apple com iphone features siri faq html Accessed March 6th 2012 Baldauf et al 2007 Baldauf M Dustdar S and Rosenberg F 2007 A Survey on Context Aware Systems International Journal of Ad Hoc and Ubiquitous Computing pages 263 277 Bernsen 2000 Bernsen N O 2000 What is Natural Interactivity In Proceedings of the 2nd International Conference on Language Resources and Ev
12. 97 Results for validation test 11 Big networks 98 10 6 12 6 13 6 14 6 15 7 1 T2 7 3 7 4 7 5 7 6 het 7 8 7 9 7 10 7 11 7 12 7 13 7 14 7 15 7 16 T 17 7 18 7 19 7 20 7 21 7 22 C 1 Circumstantial Knowledge Management for Human Like Interaction Results for validation test 12 Interoperability 98 Times for modeling tasks prota aa abe a 101 Absolute error for modeling tasks eo e 694 GR RSG S 8 102 Relative error over 1 meter for modeling tasks o oaoa aaa 102 Future work FW 01 Costs factors assignMent o a 107 Future work FW 02 Planes support ooo a o 108 Future work FW 03 Multiple selection for network items 108 Future work FW 04 XML schema e 108 Future work FW 05 Node creation by coordinates 109 Future work FW 06 Support for dynamic agents 109 Future work FW 07 Undo and redo users vale we De eo 109 Future work FW 08 Native support for the temporal aspect 110 Future work FW 09 Native support for all the context aspects 110 Future work FW 10 Efficient algorithms for shortest path 110 Future work P Well Trace tool 24a ao RAG eee 111 Future work FW 12 Inlaying in Cognos toolkit 111 Future work FW 13 Integration in Cognos Knowledge base 111 Future work FW 14 Integration in Cognos Ontology 111 Future work
13. AT te E lo E e ee E te oe Dis Loading the Application vs as EE A BR a E User Manual EL WES creen tra bs ees da Se hens Se Shen das Be A Be ds ds ai E 2 Situations Management A AI A a aes E3 Network Management oc bere a betes bee bes de E 4 Simulation Glossary Acronyms Bibliography 105 105 107 115 115 116 118 119 121 127 127 127 128 129 129 130 132 135 137 139 141 This page has been intentionally left blank List of Figures 1 1 Cognitive architecture for a natural interaction system Xir TRAPS architectire ste dl he el ee he ee He Ge Ge DE 2 2 SmartKom architecture 3 1 Architecture for the previous situation model database 3 2 Spiral lifecycle development model 3 3 Temporal planning for the development Gantt chart 4 1 Physical client server architecture 0 4 2 3 tier architecture design 434 484 4 ar A hae a a 4 3 Subsystems for the data tier 2 51 24 Laa Bae 2 ja ee 24 ORG 4 4 Subsystems tor the logic tier ig ac 22 Bae ge baelg a ek Le eos 4 5 Subsystems for the presentation tier 2 0 0004 ee eee 4 6 Entity Relationship ER model for the model knowledge 4 7 ER model for the edition knowledge 2 4 8 Mockup tor The application start sereen sra a Be Aa Re ia 4 9 Mockup for the situations and features management screen
14. The tab selector allows to move between the different screens of the application The tabs will remain disabled unless a connection to a database is established 2 The database connection panel contains all the required fields to establish a con nection to an Oracle database If the input in any of the fields is wrong its background color will change to light red to indicate this fact 3 The connect button establishes the connection to the database 4 The disconnect button closes the connection to the database that was already es tablished 5 The status bar shows the status of the connection or an error message which will be painted in red if any occurs 129 130 Circumstantial Knowledge Management for Human Like Interaction r gt A A es T Situations Management i Network Edition Simulation IP Hostname Port 1521 Instance Name Username Password _ eeeeccccce fay Connect Disconnect b Cognos S Circumstantial Knowledge Management for Human Like Interaction LaBDA group Carlos 111 University of Madrid Figure E 1 Screenshot for the main screen E 2 Situations Management This screen allows the management of situations and features Figure E 2 shows an screen shot of this interface The components are described here 1 The situational aspect selector a
15. action Not required Table 6 10 Results for validation test 10 Routing and navigation services 98 Circumstantial Knowledge Management for Human Like Interaction Validation Test Test 11 table 3 59 Y Ok Not required Table 6 11 Results for validation test 11 Big networks Validation Test Test 12 table 3 60 Interoperability Not required Table 6 12 Results for validation test 12 Interoperability In conclusion requirements FR MA 13 FR MA 14 FR MA 15 and FR MA 27 are not validated Any of these are defined as essential so their validation is not absolutely required for the system validation However in all the cases a corrective action is proposed and the validation of all the requirements is left as future lines with the maximum priority 6 2 Evaluation A good system evaluation would be able to indicate whether the system successes in its purposes and whether it provides some advantages over other systems So far no similar systems with the same purpose were found so this kind of evaluation cannot be performed However for the edition process a comparative evaluation could be achieved in order to observe whether the tool provides some advantages over direct edition of the model knowledge For this purpose an experiment has been designed in order to model a given network by two different ways The first one will edit the model directly over the model knowledge base using some auxiliary procedures
16. application It serves as an entry point for all the system functionality Additionally this subsystem also provides an interface which allows the user to authenticate and establish a connection to a database containing the knowledge model 4 2 3 2 UI_Edition Subsystem This subsystem provides the interfaces which are required for carrying out the management operations These operations can be divided into the situations and features management and the networks management Given that the network management may require a very complex interface two different screens will be used for this subsystem the first one for features and situations management and the second one for network edition 4 2 3 3 UI_Simulation Subsystem This subsystem provides an interface which serves for running a simulation over the situ ation model The simulation will consist on an agent which will move across the network while a requested service such as navigation is executed 4 3 Data Storage Two different data stores are required for this project The first store contains the knowledge base of the situation model Meanwhile the second store keeps all the data which is required to manage the model i e it contains all the entities which are required to perform an edition operation over the situation model 4 3 1 Model Knowledge Base Figure 4 6 shows an ER model which shows the conceptual design for the model knowledge store As figure 4 6 shows networks a
17. are 2 years depreciation of 50 per year and none of them are considered to have any residual value when their lifetime expires 62 Circumstantial Knowledge Management for Human Like Interaction Oy Unitary cost 1 Computer Server Sun Fire X4150 Server 2450 165 E Personal Computer Intel Quad Core 4GB RAM 650 110 Fungible goods 327 Table 3 62 Costs projection for physical resources Table 3 63 shows the detailed estimated costs for human resources based on the time planning estimated on the previous section Requirements Engines Table 3 63 Costs projection for human resources Finally table 3 64 summarizes the total costs projection for the entire project Table 3 64 Total costs projection Chapter 4 Analysis and Design The purpose of this chapter is to provide a high level abstraction of the software design given the requirements specification from chapter 3 in order to provide a solution which covers these requirements Sections 4 1 and 4 2 describe respectively the physical and the functional architectures of the system The latest one includes the system data and the system logic which is divided into subsystems and the integration of these subsystems After that section 4 3 describes the design for the databases including the different entities and the relationships among them Finally section 4 4 shows a high level design for the user interfaces descr
18. class are private and therefore they are supposed to have these accesor methods by default ModelKnowledge NetworkItem Feature id int Oo description string geometry JGeometry cost double costFactors Set lt Cost gt features Set lt Feature gt 2 isActive bool hierarchyLevel int radius double parentNode Node name string costFactors Set lt Cost gt 0 1 Situation name string aspect SituationalAspect description string situation Situation costFactor double Q name string description string items Set lt NetworkItem gt startNode Node endNode Node isBidirected bool findLinkByld links Set lt Links gt id int Link Figure 5 1 Classes for the ModelKnowledge subsystem e The SituationalAspect class has only a name which represents the situational aspect itself As the different situational aspects were defined in the requirements and were a reduced set of values we will implement this class as a Java enumerated type which will contain the different possible values for the situational aspects e The Situation class is identified by a name and a situational aspect and can also contain a description It may be noticed that we have preferred to set the situational aspect as a field of the Situation class rather than considering a set of sit
19. comfortable 1L 2X 3U 441 sU 10 211 3H 411 5X I found the functionality which I required easily the system is intuitive 1XI 21 31 41 5 11 21 311 41 5x o the task successfully and without errors the system is 10 28 30 40 50 10 20 38 40 50 ne the task did not take too much time to be completed the system is IW 20 30 40 5 1 20 30 40 5 2 What do you think are the advantages of System A I think it s more fun and also the fact of starting with a plane and created facilitates the beginning of the task 3 What do you think are the disadvantages of System A It is costly and imprecise Also it dosen t show graphically the results making it more difficult to detect an error Finally the user must have some notion of creating the code 4 What do you think are the advantages of System B E is more agile more comfortable faster and more precise It shows graphically the results at all times It is more intuitive because it is graphical 5 What do you think are the disadvantages of System B It hasn t disadvantages 6 In what situations do you think System A would be convenient Never 7 In what situations do you think System B would be convenient Always 8 Please express here any other observation The interface is very nice Figure C 1 Survey completed by subject 1 1 Evaluate the next statements in a scale from 1 to 5 being 1 strongly disagree and 5 strongly agree for each of the t
20. diverse inputs for the different modalities SmartKom was designed as a four years project It lasted from September 1999 to September 2003 and 25 7 million euros were allocated to it This funding mainly came from the German Federal Ministry of Education and Research BMBF and from other private partners Blocher 2006 As SmartKom is motivated for non desktop scenarios the situational context takes an important role of the interaction process For this reason the SmartKom system imple ments a context model to manage the circumstantial knowledge The context model is highlighted in figure 2 2 which shows the general architecture for the SmartKom system Herzog and Reithinger 2006 following the component and connector viewtype defined in Clements et al 2003 The context model for the SmartKom system is described in Porzel et al 2006 This model contemplates four different types of context the dialogical context the knowledge on what has been said the ontological or domain context the knowledge of concepts on the domain the user context the knowledge about the interlocutors and the situational context the material aspect of the context such as the time or the place One of the main ideas of SmartKom is that the kernel system can be used in different scenarios manifestations of the system such as home office mobile etc modalities such as speech gesture mimics or affectives and domains topics of conversation
21. emphasize the distinction between domain planning and discourse planning thus increasing the portability and maintainability of the system More information on TRIPS system its domains and some demonstration movies can be found in http www cs rochester edu research cisd projects trips Addition ally more information of its preceeding project TRAINS can be found in http www cs rochester edu research cisd projects trains 2 7 2 Deep Map Deep Map Malaka and Zipf 2000 is a research framework of a tourist information system whose development took place in the early 2000s The purpose of Deep Map is to build a mobile solution which serves as a trip planner and as a city navigator The system would take in consideration the user interests and needs the social and cultural background and some other contextual information such as the weather conditions or traffic While the system would provide complex functionality and combine several technologies it was thought to still be accessible for untrained users The core component for Deep Map is the geographical information system GIS which manages the location information and relates it to the user needs Additionally as the user may require some historical information about any particular location the architecture also needs some temporal capabilities in its databases For that reason Deep Map contains a 4D database which represents spatial information in three dimensions as well as tem
22. fields to create a new network A chooser element to select a plane for the selected network and a button to remove the selected plane A new plane form with all the required fields to create a new network plane A panel showing the image for the selected network plane A button to load a new image into the plane A slider to zoom over the image A set of buttons to add and remove network items as well as to select network items to be edited CHAPTER 4 ANALYSIS AND DESIGN 75 Circumstantial Knowledge Management for Human Like Interaction ala x Start Situations Management Edition Simulation m Networks m New Network Network Network M Name Name The selected network description Load Remove Descr Description Planes New Feature Plane Plane M Name l Name Load Remove Add Network Edition Operations _ Add Node Remove Node Add Link Remove Link Select renin Select inc rn Item Properties Descr Description X X coordinate Y Y coordinate Cost Basic cost Features V Feature 1 Feature 2 Load Image Zoom a Status Figure 4 10 Mockup for the networks management screen e A network item edition form with all the required fields to edit a selected network item Finally figure 4 11 shows a moc
23. for each task For this planning it has been taken into account that there is only one developer which is working in the project 20 hours per week i e 4 hours per day 5 days per week The estimated time for the whole development process is of two months and two weeks CHAPTER 3 PROJECT DESCRIPTION 61 starting on March 12th and ending in May 25th Additionally two extra months are considered for non development tasks such as research on the state of the art and require ments elicitation and specification thus considering a total project time of four months and two weeks mar 2012 apr 2012 may 2012 Duration BCEE mooo a fon Development 19 03 2012 16 04 2012 EO es Testing 17 04 2012 19 04 2012 Documentation 20 04 2012 20 04 2012 Iteration 2 23 04 2012 09 05 2012 Planning 23 04 2012 23 04 2012 EN EE HEN e ee EEC RON rea aa e se Iteration 3 10 05 2012 25 05 2012 Planning 10 05 2012 10 05 2012 Analysis 11 05 2012 11 05 2012 Figure 3 3 Temporal planning for the development Gantt chart 3 7 2 Costs Projection The costs projection for this project are detailed in this section Table 3 62 shows the detailed estimated costs for physical resources Regarding the technological resources computer servers are considered to have an average lifetime of 5 years depreciation of 20 per year while for personal computers the average lifetime
24. form description of objects and its evolution through time Additionally a tool will be developed to manage the circumstantial knowledge stored in the database This tool will provide means to edit the spatial network graph and to define objects associated to certain spatial and temporal conditions Finally another tool will be developed in order to perform a simulation over the situation model The main aim of this simulator is to show the power of the situation model itself before the integration with other modules of the human like interaction architecture takes place Some scenarios where the implemented situation model can be applied are described in appendix A 1 3 Project Context Figure 1 1 shows a model based cognitive architecture for a natural interaction system proposed by Calle 2004 In this architecture the Presentation Model and its associated interface components parse the user inputs and synthesize the outputs Different tech nologies can be implemented for this module such as speech recognition and synthesis gestures recognition a 3D character etc Supporting their interpretation in the interaction system the Ontology deals with the set of concepts and the relations among them which are referenced in the communication process The Dialogue Model processes the user communicative acts and develops its own acts and the Discourse Generator manages the system tasks during the interaction process and produces the conte
25. good when performing the task the system is comfortable 11 2X 311 4L 5l 111 211 30 41 5X I found the functionality which I required easily the system is intuitive 1X 21 30 4L 5f iL 201 3L 4X 5 re aad the task successfully and without errors the system is 1K 20 30 40 50 10 20 30 4 50 a the task did not take too much time to be completed the system is IK 20 30 40 50 1020 30 40 58 gile 2 What do you think are the advantages of System A 3 What do you think are the disadvantages of System A It s very easy to make mistakes setting the coordinates It s needed a more precisely rule with more than one decimal position You have to spend a lot of time taking the measures because usually you get wrong and you have to repeat the measure You have to write the pl sql code and in this moment you can get wrong again transcribing the measures 4 What do you think are the advantages of System B It s easy to create the network in a few minutes with less errors than the manual system You don t have to take the measures manually and only moving the mouse You can modify the network easily 5 What do you think are the disadvantages of System B You cannnot set a node in a known coordenates 6 In what situations do you think System A would be convenient When you don t have access to an informatic system when you hate computers 7 In what situations do you think System B would be convenient I that is convenient allways
26. has been intentionally left blank Chapter 6 Validation and Evaluation The main purpose of this chapter is to check that the system successfully does what it is required to do given the software requirements specified in section 3 2 This phase is called validation and it is essential to guarantee the quality of a software product and to accept it Section 6 1 shows the results for the validation tests In the cases where a test reveals that some of the requirements cannot be validated corrective actions will be proposed to solve this fact Additionally this chapter proposes metrics for a system evaluation in section 6 2 The proposed evaluation is a comparative one and given that there are no other systems of similar purposes to which this work could be compared the tool is finally compared over the manual knowledge edition This evaluation will consider both a quantitative and a qualitative approach by means of measurements of quality and performance as well as a subjective questionnaire finally concluding whether the developed tool presents advantages over the direct edition of the model knowledge 6 1 Validation The process of software validation will check that the requirements specified by the user are covered by the system To do so the validation tests from section 3 5 are executed In order to perform this execution correctly it must checked that the postconditions are met given that the test steps are executed over the syst
27. it is adequate to introduce the state of the art of the concepts tech nologies and techniques related to the aim of this project such as ubiquitous computing context aware systems context models natural interaction and situation models as well as presenting some applications Chapter 2 contains this information Chapter 3 provides the description for the project The main aim of that chapter is to formalize the system functionality through its requirements specification Once the user requirements are formalized a feasibility study is performed to ascertain if the goals and requirements are realistic or not given some previous work on the field and known the technical and environmental background The chapter also includes the definition of CHAPTER 1 INTRODUCTION 17 validation tests to cover all the user requirements Finally a planning is elaborated which describes the development methodology as well as an schedule and a cost projection for the whole development process Chapter 4 provides a high level design of the system both in terms of the physical and functional architecture and 5 delves into this design describing the system architecture with enough detail to serve as a guide for its implementation After the implementation phases are carried out chapter 6 provides the results of the execution of the validation tests in order to determine whether the developed system meets the user requirements and proposes and performs a
28. it is only for the personnel However you can use the restaurant restroom when you get there ARTHUR Ok I ll wait ARTHUR arrives to the end of the corridor and calls the elevator HAL Remember go to the first floor now ARTHUR Thanks HAL ARTHUR steps out of the elevator in the first floor HAL Ok turn right and you ll find the restaurant ARTHUR Ok thank s a lot 118 Circumstantial Knowledge Management for Human Like Interaction A 3 Assisting Disabled People This scenario shows how the situation model could take into account information about the user in order to improve its experience Alan is working on his University entrance exams these days He has to do these exams in the campus of Legan s of Universidad Carlos III and he does not know the college He is a nice guy but unfortunately he is using a wheelchair since he was five Although he is really used to his wheelchair he is still having some mobility issues Alan has downloaded in his smartphone a mobile application developed by the Uni versity wich contains the map of the campus and includes a simple navigation feature to guide the user to a searched location Furthermore from the last update a simple situation model was implemented which takes into account among others some information about the user For instance he could select in a profile that he is a disabled person and the system will provide alternative routes In his case the system wil
29. network item has that feature 18 The save details button saves the changes that the user does over the details fields 15 16 17 and 18 19 The reference node creation button allows to create a new reference node by clicking over the map area 4 after selecting the button This reference node is considered as an usual node and the same operations can be applied over them but is represented as a yellow triangle in the amp area 4 At most two reference nodes can be created 20 The update references button will update the coordinates of all the nodes in the network automatically For this to happen the plane must contain two reference nodes whose coordinates must be set 21 The commit button will persistently store all the changes done since the last commit or rollback operation in the database It must be noticed that if there are some uncommited changes and the user tries to perform any operation that would discard these changes such as disconnecting from the database loading another network or even closing the application a warning will appear such as the one shown in figure E 4 22 The rollback button will discard all the changes done since the last commit opera tion 23 The status bar will show the status of the executed operations as well as any error which could have occurred in red A Some changes over the open network are not committed and will be lost Are you sure you want to load a new network Fig
30. nowadays mainly serve as a supplement for standard GUIs in the future those interfaces will eventually replace them as well as current GUIs have replaced the already obsolete CLIs 2 5 4 Natural Interaction The previous concepts will serve as the proper basis to define what natural interaction is According to Bernsen 2000 natural interaction is the interaction between humans and computer systems which takes place in the ways in which humans normally ex change information with one another Based on this definition it can be realized that natural interaction is the kind of interaction used with a NUI As previously stated an important advantage of these interaction techniques is that the user does not require any technical knowledge or ability to deal with the system but only the capability of interacting with other humans It involves that these kind of systems must be able both to interpret and understand what the user is expressing as well as to ellaborate its own interventions in a way that can be understood by the user A natural interaction system combines techniques from very different fields Research in natural interaction not only is strongly influenced by areas from the computer science such as the human computer interaction the knowledge engineering or the natural lan guage processing but also from other human sciences such as the psycholinguistics and the sociolinguistics A key issue for a natural interaction system is
31. of the main screen corresponding to the UI_CognosS sub system This screen allows to start a connection to a knowledge base and serves as an entry point for the other application modules through the tab panel selector at the top Circumstantial Knowledge Management for Human Like Interaction _ xX Start Situations Management Edition Simulation Database Connection Host Host Port Port User User Password 8806 Instance Instance Disconnect Circumstantial Knowledge Management for Human Like Interaction Status Figure 4 8 Mockup for the application start screen CHAPTER 4 ANALYSIS AND DESIGN 73 In brief this subsystem will contain one screen which must show the next components e A way to go into the other interface components A good approach for this mechanism could be a tab selector where the different application functionalities are placed in different tabs This provides an intuitive way for navigating between the different subsystems Additionally some tabs may be disabled if they are not accessible at any given moment in time so the user is able to recognize anytime the operations that can be performed e A database connection form which would contain all the required fields to estab lish the connection to the knowledge base and also some mechanism to close the connection As these components require little spa
32. some objects which are deleted on cascade e Update an existing object in the database This function is not provided for some objects which cannot be updated e Select an object or a set of objects from the database This selection may be filtered by some fields to return some results filtered by an identifier to return only one result or not filtered to return all the results 5 4 2 Edition Subsystem The Edition subsystem is composed by only one class containing the functions required by the edition functionality and an auxiliar class Actually most of this functionality is provided by the DataGateway subsystem as well as the classes in the presentation tier It must be taken into account that the use of the interface for edition purposes all the components of the GUI belongs to the presentation tier carries out in most of the cases direct modifications over the database Due to this fact most of the times direct com munication is performed between the UI_Edition subsystem and the corresponding data gateway Figure 5 4 shows the class diagram for the Edition subsystem 84 Circumstantial Knowledge Management for Human Like Interaction Edition EditionFunctions ReferenceNode calcEuclideanDistance nodel Node node2 Node double calcReferenceCoordinates refNodel ReferenceNode refNode2 ReferenceNode node VisualNode double Figure 5 4 Classes for the Edition subsystem 5 4 2 1 ReferenceNode The
33. the data which is required by the situation model This data contains all the network related information nodes and links costs features and situations This subsystem is intended to be able to work separately from the EditionKnowledge subsystem 4 2 1 2 EditionKnowledge Subsystem The EditionKnowledge subsystem contains additional information which complemented with the model knowledge supports the model edition functionality Most of the data entities within this subsystem have the purpose of graphically representing the model data such as nodes and links For this reason this subsystem considers a plane as a bidimensional entity to represent a subset of a network in the application 4 2 2 Logic Tier The logic tier gathers the business logic that provides the system functionality It also acts as an intermediate layer between the data model and the user interface The data tier contains three different subsystems the edition subsystem the simulation subsystem and the data gateway Figure 4 4 shows this structure graphically 4 2 2 1 Edition Subsystem The aim of the edition subsystem is to support the functionality for situations and fea tures management and network edition The software requirements for this functionality CHAPTER 4 ANALYSIS AND DESIGN 67 Logic Tier subsystem subsystem subsystem Edition Simulation DataGateway Figure 4 4 Subsystems for the logic tier are detailed in section 3 2 1 1 In
34. the user for authentication information user name and password before he can access to the model and select a network Cyce First Priority Eso al Stability Sabe Table 3 47 Requirement NFR SA 01 Authorized access CHAPTER 3 PROJECT DESCRIPTION 47 3 2 2 3 Interoperability Requirements This subsection describes the user requirements referring to how the application shall be have over different platforms as AAA Name Name Cross platform A Description The system shall operate in Microsoft Windows XP or higher and Linux systems either of 32 or 64 bits Cyce Fist Priority Essential Stability Stable Table 3 48 Requirement NFR IO 01 Cross platform 3 3 Previous Work Some previous work on the database storing the situational information already exists Fig ure 3 1 shows the architecture for the previously existing situational database as described in Rivero et al 2007 Link IDLink CommonType Search gt lt Formed of StartNodelD CommonType id Place 4 Represents EndNodelD CommonType IDPlace CommonType p 1 CostLink CommonType PlaceForm setOf TGeometry LinkGeom setOf TGeometry Bidirection CommonType Tr 2 User Search gt Object lt 4 Represents d Node earc IDNode CommonT IDUser CommonType IDObject CommonType gt CostNode E COMMON ObjectForm
35. they are not essential The support for planes management is in cluded as a future line for the short term Table 6 5 Results for validation test 05 Plane selection CHAPTER 6 VALIDATION AND EVALUATION Validation Test Test 06 table 3 54 97 Not required Table 6 6 Results for validation test 06 Plane visualization Validation Test Test 07 table 3 55 Result X Wrong The test does not pass because cost factors cannot be assigned to a network item Corrective action Network items can be edited to update their description co ordinates basic cost activeness and features However they cannot be assigned cost factors directly and for this reason requirement FR MA 27 cannot be validated So far cost fac tors can be assigned indirectly by means of features As this requirement is defined as conditional the support for the as signment of cost factors is included as a future line with the highest priority Table 6 7 Results for validation test 07 Network items management Validation Test Test 08 table 3 56 Test Name Result Corrective action Not required Table 6 8 Results for validation test 08 Agent simulation Validation Test Test 09 table 3 57 Test Name Result Corrective action Not required Table 6 9 Results for validation test 09 Simulation services Validation Test Test 10 table 3 58 Test Name Routing and navigation services Result v Ok Corrective
36. we have considered Java to be the most appropriate as it was stated in the feasibility study section 3 4 Besides the advantages listed in that section many others can be found First of all it is highly extended and the development team is already ex perienced in Java programming Secondly it executes over a virtual machine Java Virtual Machine and it provides multiplatform features Thirdly there are many libraries for Java which could already provide some functionality required by this system and could be reused Finally there are many Integrated Development Environments IDEs which allows a What You See Is What You Get WYSIWYG creation and edition of user inter faces which will be useful for the design of the application screens On the other hand Java is less efficient that other object oriented languages however this is not a big handicap as the big load of the application is intended to be carried out in the server side For the application development Oracle NetBeans 7 will be used This IDE for Java allows an easy to use WYSIWYG editor for user interfaces Moreover it also supports collaboration tools such as software versionining and revision control systems For our case Subversion will be used as a system for revision control and versioning which also serves for storing a backup of each different version of the application source code Regarding the data storage it has been decided to use a relational database mana
37. while the second specifies the cost as a positive real number When a new cost factor is added a new empty row is included in the table The selector in the first column lists all the situations in the system Notice that the system will not allow two different cost factors referring to the same situation 132 Circumstantial Knowledge Management for Human Like Interaction 9 The cost factor removal button deletes the row with the selected cost factor from table 8 10 The cost factor save button confirms the changes in the cost factors shown 8 and saves them in the database 11 The status bar shows the status of the last performed operation or whether an error occurred E 3 Network Management This screen allows to manage the networks of the situation model Figure E 3 shows this interface lola Cognos s Database Connection Situations Management Network Edition Simulation Network Selection SABaTiNI_2 Sabatini Floor 2 Network O jons 8 0 O 0 ote 072 o Details Description Laboratorio DEI Location X 4 21143102707348 83 7284983728498 Cost 0 0 Features Feature Outside Stairs Ne AAA AAA A VEE Y ca Y PI ee
38. will play the role of the customer while the student will act as a requirements engineer analyst designer and programmer An introductory meeting was carried out the first week of December 2011 This meeting took the format of an open interview between the customer and the engineer where the problem of how to improve a natural interaction system in order to manage situational knowledge was discussed Later that month a second meeting was arranged also with the format of an open interview to discuss in more detail the theoretical background of the project as well as 39 36 Circumstantial Knowledge Management for Human Like Interaction some existing applications of situation models in natural interaction systems A semi structured interview took place in February with the purpose of settling the re port structure and to choose the most appropiate development methodology for the project section 3 6 Additionally three semi structured interviews were carried out for the beginning of each development cycle in March April and May to obtain specify and update the software requirements In the last two interviews the user could refine the software requirements given its own experience with the generated prototypes from the previous cycles 3 2 Software Requirements Specification For the requirements specifications task the IEEE recommended practices IEEE 1998 will be followed According to these guidelines a good specific
39. 012 B squeda R pida de Caminos en Grafos de Alta Cardi nalidad Est ticos y Din micos PhD thesis Universidad Carlos III de Madrid Rivero et al 2007 Rivero J Cuadra D del Valle D and Calle F J 2007 Incor porating Circumstantial Knowledge Influence over Natural Interaction In EEE Inter national Conference on Pervasive Services pages 415 420 Ronzani 2009 Ronzani D 2009 The Battle of Concepts Ubiquitous Computing Pervasive Computing and Ambient Intelligence in Mass Media Ubiquitous Computing and Communication Journal 4 9 19 Schmidt 2000 Schmidt A 2000 Implicit Human Computer Interaction Through Con text Personal and Ubiquitous Computing 4 191 199 Schofield 1994 Schofield J 1994 Analysis Can Novell Make the Connections The Guardian London page T8 Smith 1985 Smith H C 1985 Database Design Composing Fully Normalized Tables from a Rigorous Dependency Diagram Communications of the ACM 28 8 BIBLIOGRAPHY 145 S nchez Pi 2011 S nchez Pi N 2011 Intelligent Techniques for Context Aware Sys tems Adaptation PhD thesis Universidad Carlos III de Madrid Strang and Linnhoff Popien 2004 Strang T and Linnhoff Popien C 2004 A Context Modeling Survey In Workshop on Advanced Context Modelling Reasoning and Man agement pages 1 8 Turing 1950 Turing A M 1950 Computer Machinery and Intelligence Mind 59 433 460 Wahlster 2006 Wahlster
40. A 07 FR MA 08 FR MA 09 FR MA 12 FR MA 16 FR MA 17 FR MA 18 FR MA 19 FR MA 20 FR MA 21 FR MA 22 FR MA 27 Table 3 55 Validation test 07 Network items management CHAPTER 3 PROJECT DESCRIPTION 55 Test 08 Agent simulation The system simulates the motion of an agent over the network e The user has authenticated in the application The user has selected a network and a network plane The user selects the screen for network simulation The user may perform the next operations a place the agent over the network b move the agent across the network using the mouse or keyboard or c use a physical layer simulation GPS RFID to move the agent to a new position over the net work The system places the agent in the new position Post Conditions The agent is placed in a new position of the network The closest node is selected as the source node for the simulation services which the user may request FR 03 FR SI 03 FR SI 04 FR SI 05 FR SI 10 Table 3 56 Validation test 08 Agent simulation TOT The system shows a screen to simulate a situation model service Oo Pre Conditions The user has authenticated in the application The user has selected a network and a network plane The user selects the screen for network simulation The user has placed the agent over the network Steps The user may perform the next operations a select the de scription service b select the routing se
41. Aspects DNA UNA DC UNA name aspect description features costsFeatures situation feature costFactors situation item costFactors itemsFeatures feature networks networksItems DC UNA DC UNA id radius parentNode DC UNA DC UNA DC UNA DC UNA Figure 5 8 Relational graph for the model knowledge The relational graph does not show the field types although it can be deduced from the data types shown in the class diagrams from the previous sections It may be noticed that all the referential integrity keys are On Delete Cascade On Update No Action but for the foreign key from the situations relation to the situationalAspects relation which is On Delete No action On Update No Action Actually this last relation is intended to be a validation table where its values are fixed and are not supposed to be deleted This fact may also require a mechanism such as a trigger to prevent modifications in this table The rule for deletions allows the removal of elements on cascade which makes sense for all the cases in our design No action is considered for updates as in the futures the names will refer to concepts in an Ontology and it is not desired that an editor alters the semantical value introduced by another editor CHAPTER 5 IMPLEMENTATION 93 Regarding additional semantics not shown in the graph the cost factors must be greater or equal to zero as negative cost factors would probably make it impos
42. Calle F J Albacete E S nchez E Olaziregi G del Valle D Rivero J and Cuadra D 2011 Cognos A Pragmatic Annotation Toolkit for the Ac quisition of Natural Interaction Knowledge Procesamiento del Lenguaje Natural pages 269 276 Calle et al 2008 Calle F J Mart nez P del Valle D and Cuadra D 2008 To wards the Achievement of Natural Interaction In Engineering the User Interface From Research to Practice pages 63 80 Springer Chomsky 1957 Chomsky N 1957 Syntactic Structures Mouton amp Co Clements et al 2003 Clements P Bachmann F Bass L Garlan D Ivers J Little R Merson P Nord R and Stafford J 2003 Documenting Software Architectures Views and Beyond Addison Wesley Connolly 2001 Connolly J 2001 Context in the Study of Human Languages and Computer Programming Languages A Comparison In Modeling and Using Contezt volume 2116 pages 116 128 Springer Cook et al 2007 Cook D J Augusto J C and Jakkula V R 2007 Ambient Intelligence Technologies Applications and Opportunities Cuadra et al 2008 Cuadra D Calle F J Rivero J and del Valle D 2008 Apply ing Spatio Temporal Databases to Interaction Agents In International Symposium on Distributed Computing and Artificial Intelligence pages 536 540 Springer Dey and Abowd 1999 Dey A K and Abowd G D 1999 Towards a Better Under standing of Context and C
43. Cycle Fist Priority Essential Stability Stable Table 3 13 Requirement FR MA 08 Network items cost FR MA 09 Network items coordinates Networks items are located in a geometric coordinates system ayas Fist Priority Fatal Stability Stable Table 3 14 Requirement FR MA 09 Network items coordinates D FMM Features The a shall provide means to create and delete user defined features Cycle Firs Priority sential Stability Stable Table 3 15 Requirement FR MA 10 Features 39 40 Circumstantial Knowledge Management for Human Like Interaction FRMATE Features properties Features are identified by their name Cycle Fist Priority Essential Stability Stable Table 3 16 Requirement FR MA 11 Features properties FR FR MA 12 eee 12 E a a a sasas items features Description The system shall provide means to the user to assign features to the network items Cyce Fist Priority Fssential Stability Sabe Table 3 17 Requirement FR MA 12 Network items features D MMR Name Network Network planes management ssi Network planes management ssi PE The system should provide means for the user to create and delete planes for a previously selected network Cyce Fist Priority Optional Stability Sadie Table 3 18 Requirement FR MA 13 Network planes FRMATI Network planes properties Description Network planes a
44. D Test 11 Name Big networks Description The system is able to work with a big network Pre Conditions e The user has authenticated in the application e The user has selected a network and a network plane a Steps 1 The user may perform the next operations manage a net work with more than 500 network items or b perform a simu lation over a network with more than 500 network items 2 The system carries out the requested operation Post Conditions e The management or simulation services are executed over the big network Covered Regs NFR SC 01 Table 3 59 Validation test 11 Big networks CHAPTER 3 PROJECT DESCRIPTION 97 Steps 1 The user starts the application either in Windows XP or higher or in Linux systems Post Conditions e The application starts and is operable Covered Regs NFR IO 01 Table 3 60 Validation test 12 Interoperability Table 3 61 shows the traceability matrix which graphically plots how the requirements are covered by each validation test 58 Circumstantial Knowledge Management for Human Like Interaction NFR IO 01 NFR SA 01 gt NFR SC 01 FR SI 12 FR SI 11 FR SI 10 FR SI 09 FR SI 08 FR SI 07 FR SI 06 FR SI 05 FR SI 04 FR SI 03 FR SI 02 FR SI 01 FR MA 28 FR MA 27 FR MA 26 FR MA 25 FR MA 24 FR MA 23 FR MA 22 FR MA 21 Table 3 61 Traceability matrix CHAPTER 3 PROJECT DESCRIPTION 59 3 6 Development Methodology The software development pro
45. FW 15 Full integration in Cognos 112 Future work FW 16 Acquisition of knowledge of a real scenario 112 Future work FW 17 Model installation 112 Future work FW 18 Integration with the Task Model 112 Future work FW 19 Integration with the Dialogue Model 113 Future work FW 20 Full integration in the natural interaction system 113 Future work FW 21 Physical layer is i ES ee A eve ewe 113 Future work FW 22 Full evaluation of the situation model 113 Time measurements for the evaluation 2 20004 121 LIST OF TABLES C 2 Accuracy measurements for the evaluation with direct edition C 3 Accuracy measurements for the evaluation with the edition tool This page has been intentionally left blank Chapter 1 Introduction In recent years personal computers have evolved so much that we can now find them in many different forms such as smartphones tablets or navigation systems As most of the people are gaining access to those devices new interfaces are required so that the users can interact with systems in a simple way thus not requiring any previous training or specific knowledge Moreover these interfaces must be useful and accesible to all the users regardless of their condition or their technical capabilities The aim of human like interaction is to provide interfaces which imitate the way in which humans interact with each other Wit
46. I which groups a set of computational tech niques for dealing with natural language 25 26 139 ontology system that formally represents the knowledge of a domain as a set of con cepts and the relationships among them supporting the other models of the natural interaction architecture 10 15 111 pervasive computing term to refer to the fact that computers are located everywhere so the user could connect to the Internet at any time 20 21 situation model computational model that aims to formalize and store the circumstantial knowledge regarding any interaction in order to process it 10 14 19 28 36 69 106 107 111 113 spatio temporal database database which manages space and time information 15 ubiquitous computing term coined by Mark Weiser in the late 1980s to refer to a com puting paradigm in which computers vanish in the background and the user can focus on the purpose of the interaction rather than in its form 19 22 25 33 User Model model of the natural interaction system that formalizes the knowledge about the user preferences in order to provide a customized interaction to the user 15 Acronyms ACO Ant Colony Optimization 87 110 AI Artificial Intelligence 21 22 25 26 28 119 137 138 CLI Command Line Interface 25 27 CRUD Create Read Update and Delete 68 78 ER Entity Relationship 5 24 69 71 GPS Global Positioning System 23 44 55 76 113 GUI Graphical User Interface 25 27 30 68
47. Ine Onl De ni oyl E 12 Finally the results for the relative error are shown in table 6 15 Edition Tool 90 8195 28 3089 38 107 10 73 Table 6 15 Relative error over 1 meter for modeling tasks 6 2 3 Subjective Evaluation The experiment subjects were also asked for whether the system is comfortable intuitive reliable and agile these questions are shown in figure 6 2 and they had to answer in a scale from 1 to 5 Figures 6 3 to 6 6 graph the results for this subjective evaluation Comfortable Direct Edition Comfortable Edition Tool ee Agree Strongly disagree Neither agree nor disagree Disagree Strongly agree oe a Direct edition b Edition tool Figure 6 3 Results for the subjective evaluation comfortable CHAPTER 6 VALIDATION AND EVALUATION 103 Intuitive Direct Edition a disagree Strongly agree A a Neither agree nor disagree de Disagree Intuitive Edition Tool Neither agree nor i disagree y Strongly agree a Direct edition b Edition tool Figure 6 4 Results for the subjective evaluation intuitive Reliable Direct Edition a Strongly disagree Strongly agree Disagree Neither agree nor isagree Reliable Edition Tool 7 A Neither agree nor disagree a Direct edition Aypee b Edition tool Figure 6 5 Results for the sub
48. LL op A gp AA ATA a REE ee eS 47 34 Feasibility Study le a ri lr lr ri Bee on 48 3 Validation Testine pepa dr aby de e o A Mine a ap e 50 3 6 Development Methodology xo rs de A ge 59 on Project Planning pa e A A A ES 60 Analysis and Design 63 4 1 Physical Architecture ida A A Se bea od 63 4 2 Functional Architecture aa a AA GR 64 4 3 Data Storage A a 69 4 4 Interaction Design a bas Sako ete Buda ale hale ta Bada ote 72 Implementation 77 5 1 Implementation Issues sto we al ac ne at Oke eA eae ane ee whee we TT 5 2 Physical Architecture ooa A E lhe od 78 5 3 Functional Architecture Data Tier oc bsg arse oh bo ao 79 5 4 Functional Architecture Logic Tier ooa aa 81 5 5 Functional Architecture Presentation Tier 89 o A ete ee hice re tele etn en etn ae le erties Ilene fae 92 Validation and Evaluation 95 6L Validation s a iste panda ga es the Sn Baar fen Eo he Fs Ee the JE we Eire ir 95 69 EvaliatiOns 2 Es eii ee Ga ee ee EE ee A ee 98 CONTENTS 7 Conclusions and Future Work 7 1 Conclusions T2 Fatre Works ring rates a ts tacos aos ea ts A Scenarios A 1 Improving a Navigation System 4 4 463 kk A ee a a Ae A 2 Enhancing a Natural Interaction System 04 A 3 Assisting Disabled People eri E EA Be a B Ethical Considerations C Evaluation Data D Installation Guide DE dnstallino the JVM 200 de ol E de Ge Rd ae GN ee he E eae tt D 2 Connecting tO the VEN Los a
49. ReferenceNode class is a subclass of VisualNode Actually it does not provide any additional field or method to this class however it has a different graphical representation Reference nodes are used for calculating the coordinates of other nodes in the network The procedure for this operation is explained in the description of the EditionFunctions class 5 4 2 2 EditionFunctions This class only contains two auxiliary methods The first one stands for calculating the euclidean distance between two nodes Given the coordinates of the first node z and y1 and the coordinates of the second node x2 and y2 the euclidean distance is defined as d y 2 21 y2 y1 For our application the euclidean distance is calculated in order to set an initial basic cost to a link By default the cost for a node is 0 while the cost for a link is the euclidean distance between his initial and ending nodes The second method calculates the position of one node given the position of two different nodes The purpose of this operation is to ease the process of assigning coordinates as now the user only has to define two reference nodes in the network and assign coordinates to both of them Each reference node contains both real coordinates x and y where i 1 2 and graphical coordinates over the edition plane x and y where i 1 2 Given that the two reference nodes are located in different places and the segment they genera
50. Subsystem The data gateway is an independent module which performs logic operations which deals with the data tier directly For this case the data gateway will provide routines to perform Create Read Update and Delete CRUD operations over a database The inclusion of a data gateway as a separate module provides several advantages as it increases modularity and decreases the coupling Actually the modules in the simulation subsystem always perform the database communication by means of the data gateway using the methods provided by this subsystem For this reason any change over the database design will require only the data gateway to be updated in order to comply with the new design while other modules can still rely on the gateway subsystem to perform their data CRUD operations 4 2 3 Presentation Tier The presentation tier contains the modules in charge of handling the user interfaces which are required to interact with the system In this case the presentation tier includes three different subsystems as shown in figure 4 5 This section describes each of them pro viding an approach of the GUI controls that they must provide to support the required functionality Presentation Tier subsystem subsystem subsystem UL Edition UL Simulation UL _CognosS Figure 4 5 Subsystems for the presentation tier CHAPTER 4 ANALYSIS AND DESIGN 69 4 2 3 1 UI_CognosS Subsystem This subsystem contains the main frame of the
51. The system shall provide means to the user to select a network in order to perform a simulation on it Gyele Second Priority Essential Stability Stable Table 3 34 Requirement FR SI 01 Network selection 44 Circumstantial Knowledge Management for Human Like Interaction PESTO Network display Description When a network is selected the system shall display the network items and the network background image Cyce Second Priority Essential Stability Table 3 35 Requirement FR SI 02 Network display D n Name Person Person placement 2 DE A the user clicks on the network the system shall place a person on the network over the location that the user has clicked Cycle Second Priority Esential Stability Stable Table 3 36 Requirement FR SI 03 Person placement FRESCO Description Once the person is placed over the network the system shall allow its motion using the keyboard direction arrows and the mouse Cyce Second Priority Essential Stability Stable Table 3 37 Requirement FR SI 04 Person motion PISE Physical layer simulation Description Additionally the system should provide some simulators for a physical layer for localization such as GPS or Radio Frequency IDentification RFID which will move the per son across the network Cycle Second Priority Conditional Stability Stable Table 3 38 Requirement FR SI 05 Physical layer simu
52. UNIVERSIDAD CARLOS III DE MADRID ESCUELA POLITECNICA SUPERIOR BACHELOR IN COMPUTER SCIENCE AND ENGINEERING BACHELOR THESIS CIRCUMSTANTIAL K NowLEDGE MANAGEMENT FOR Human Lixe INTERACTION Author ALEJANDRO BALDOMINOS G MEZ Tutor FRANCISCO JAVIER CALLE G MEZ LEGAN S JUNE 2012 This work is distributed under the Creative Commons 3 0 license You are free to copy distribute and transmit the work under the following condi tions i you must attribute the work in the manner specified by the author or licensor but not in any way that suggests that they endorse you or your use of the work ii you may not use this work for commercial purposes and iii you may not alter transform or build upon this work Any of the above conditions can be waived if you get permission from the copyright holder See http creativecommons org licenses by nc nd 3 0 for further details a zZ S Z gt a Email abaldomi inf uc3m es Phone 34 91 624 9114 Address Advanced Databases Group LaBDA Computer Science Department Universidad Carlos III de Madrid Avda de la Universidad 30 28911 Legan s Madrid Spain Please cite this thesis as Baldominos A 2012 Circumstantial Knowledge Management for Human Like Interaction Bachelor thesis Universidad Carlos II de Madrid 111 BACHELOR THESIS CIRCUMSTANTIAL KNOWLEDGE MANAGEMENT FOR HUMAN LIKE INTERACTION Author ALEJANDRO BALDOMINOS G MEZ Tut
53. a ver realmente mi progreso con el trabajo Un agradecimiento muy especial se lo dirijo a Javi por todas esas reuniones de diez minutos que en realidad acababan durando horas reuniones en las que de alg n modo me ha conseguido introducir la idea de que todo es tremendamente sencillo sin importar el n mero de dimensiones a considerar Tambi n aunque hayamos tenido menos contacto debo agradecer a Loli su disponibilidad para atenderme con cualquier duda que haya podido tener Durante cuatro a os de carrera a uno le da tiempo a conocer a muchos profesores pero no todos consiguen que te cuestiones tu propia forma de pensar Gracias Gonzalo por hacerme ver que detr s de cada idea por evidente que hoy nos resulte alguien en alg n momento de la Historia tuvo que luchar para poder establecerla Tambi n quiero agradecer a mis compa eros del Laboratorio de Bases de Datos los buenos ratos que me hacen pasar a diario Aunque cada vez dedico m s horas en el labo ratorio hab is conseguido que cada vez la jornada se me haga m s corta Aprovecho para agradecer a Mar a y Chumo el tiempo que dedicaron para participar en mi evaluaci n Espero que pas rais un buen rato con ella Y por ltimo me gustar a dedicar un agrade cimiento muy especial a Espe y a Gara no s lo por participar en la evaluaci n sino por dejar lo que estuvieran haciendo para ayudarme con cualquier duda que he podido tener en cualquier momento Muchas gracias chica
54. ach Pervasive and Mobile Computing pages 37 64 Herzog and Reithinger 2006 Herzog G and Reithinger N 2006 The SmartKom Ar chitecture A Framework for Multimodal Dialogue Systems In SmartKom Foundations of Multimodal Dialogue Systems chapter 4 pages 55 70 Springer IDC 1996 IDC 1995 1996 Transition to the Information Highway Era Information Industry and Technology Update page 2 IEEE 1998 IEEE 1998 830 1998 IEEE Recommended Practice for Software Require ments Specifications Institute for Electrical and Electronics Engineers Izquierdo 2011 Izquierdo P J 2011 Entorno Multiagente Implementado sobre Oracle 11g Bachelor thesis Universidad Carlos III de Madrid James Allen and Stent 2001 James Allen g F and Stent A 2001 An Architecture for More Realistic Conversational Systems In Proceedings of Intelligent User Interfaces 2001 pages 17 17 Jones 1999 Jones S 1999 Digital Era will Make Prime Time Television Obsolete Financial Times page 10 Korf 1985 Korf R E 1985 Depth First Iterative Deepening An Optimal Admissible Tree Search Artificial Intelligence 27 97 109 Liddy 2003 Liddy E D 2003 Natural Language Processing Encyclopedia of Library and Information Services Lohr and Markoff 1998 Lohr S and Markoff J 1998 Computing s Next Wave is Nearly at Hand Imagining the Future in a Post PC World The New York Times page Cl 144 Circumstantial K
55. agement over a specific situation model In order to work on this project solid understanding on mathematics graph theory geometry vectorial algebra software engineering object oriented programming graph ical interfaces design and development computer graphics artificial intelligence search algorithms knowledge engineering and databases was required However this is far from being as important as the knowledge obtained while working on the project Not only an improvement of the expertise level on the previously mentioned areas but also a better understanding on the components of the interaction between humans and a comprehension CHAPTER 7 CONCLUSIONS AND FUTURE WORK 107 on the state of the art of related research areas And again this is still far from the fun of working on this project 7 2 Future Work The purpose of this section is to propose future lines of work for this project The first aim of these works is to complete the project in order to satisfy some user requirements which did not validate in the previous chapter Moreover this project is a first step for going into the research of situation modeling for a natural interaction system and for this reason other future works are intended to improve the management tool and the situation model already implemented This section describes the future work for the project Each line of work is defined in a table and contains a name a description and a priority which c
56. aluation pages 34 37 Bill N Schilit and Want 1994 Bill N Schilit N A and Want R 1994 Context Aware Computing Applications In Proceedings of the Workshop on Mobile Computing Systems and Applications pages 85 90 Blocher 2006 Blocher A 2006 Facts and Figures About the SmartKom Project In SmartKom Foundations of Multimodal Dialogue Systems chapter 2 pages 29 39 Springer BOE 2008 BOE January 19th 2008 Real Decreto 1720 2007 de 21 de diciembre por el que se aprueba el Reglamento de desarrollo de la Ley Org nica 15 1999 de 12 de diciembre de protecci n de datos de car cter personal Bolet n Oficial del Estado 17 4103 4136 Boehm 1986 Boehm B 1986 A Spiral Model of Software Development and Enhance ment ACM SIGSOFT Software Engineering Notes 11 14 24 Bolt 1980 Bolt R A 1980 Put That There Voice and Gesture at the Graphics Interface ACM Computer Graphics 14 262 270 141 142 Circumstantial Knowledge Management for Human Like Interaction Calle 2004 Calle F J 2004 Interacci n Natural mediante Procesamiento Intencional Modelo de Hilos en Didlogos PhD thesis Universidad Polit cnica de Madrid Calle et al 2010 Calle F J Albacete E S nchez E del Valle D Rivero J and Cuadra D 2010 Cognos A Natural Interaction Knowledge Management Toolkit In Natural Language Processing and Information Systems pages 303 304 Springer Calle et al 2011
57. an be either high normal or low depending on whether the future work is intended to be carried out in the short term within the next few months the mid term to be completed in the next year or the long term respectively Regarding the taxonomy for the future works they will be classified attending to whether they are related to the edition tool to the situation model to the Cognos toolkit or to the natural interaction system Actually this work will be applied in a research project funded by the Spanish Ministry of Industry CADOOH TSI 020302 2011 21 and some of the future works will be taken into practice in the next few months 7 2 1 Management Tool Future works in this section propose some improvements on the edition tool in order to provide new functionalities or enhance the user experience From all of those future works FW 01 and FW 02 are essential in order to validate the user requirements EWIT Costs factors assignment Description A direct way of assigning cost factors to network items must be imple mented so that features are not strictly required for this purpose This future work has the highest priority as the development of this function ality is required to validate requirement FR MA 27 Table 7 1 Future work FW 01 Costs factors assignment 108 Circumstantial Knowledge Management for Human Like Interaction FW02 Planes support Description So far the management tool only supports one
58. ane and its network items Pre Conditions e The user has authenticated in the application e The user has selected a network and a network plane Steps 1 The system loads the plane and shows its background image and network items nodes and links 2 The user may perform the next operations a load a new back ground image for the plane or b move and zoom through the plane 3 The system performs the user operation Post Conditions e The plane is loaded and shown in the interface e A new background image is loaded for the plane if the user se lected that option FR 02 FR MA 14 FR MA 23 Table 3 54 Validation test 06 Plane visualization Text 07 Network items management The system shows a screen to manage network items e Pre Conditions The user has authenticated in the application The user has selected a network and a network plane The user selects the screen for network management The user may perform the next operations a create a new node or link b remove an existing node or link or c edit the properties description coordinates basic cost activeness features and cost factors of an existing node or link The system performs the user operation Post Conditions Either the new item is created the removed item no longer ex ists in the network or the selected item properties are updated according to the user actions Covered Reqs FR 02 FR 04 FR MA 04 FR MA 05 FR MA 06 FR M
59. ary can be found in http jxlayer java net 5 5 4 2 Java Universal Network Graph Framework JUNG The JUNG library provides many interesting tools for dealing with graphs Its functionality goes from rendering and drawing the graph to executing algorithms over it In this work the renderers of the JUNG library are used to draw the graph Many of the library classes will require to be overriden to change the default functionality in order to alter the nodes and links appearance to print labels etc Moreover the library also provides the Dijkstra algorithm to find the shortest path between two nodes of the graph This algorithm can be used for the project unless more sophisticated heuristic algorithms are implemented later Downloads documentation and examples for this library can be found in http jung sourceforge net 92 Circumstantial Knowledge Management for Human Like Interaction 5 6 Data Storage This section describes the implementation of the data stores of the project As it was stated in section 4 3 two independent data stores are required the first one containing the model knowledge base and the second one storing the edition knowledge database 5 6 1 Model Knowledge Base Figure 5 8 shows the relational graph Smith 1985 for implementing the model knowledge base This graph provides a detailed description of the database architecture which may be used for modeling the database using the SQL language situational
60. ation must address the software functionality the external interfaces performance issues other non functional features and design or implementation constraints Moreover the requirements specification should be e Correct every requirement is one that the software shall meet according to the user needs e Unambiguous every requirement has one single interpretation e Complete the document reflects all significant software requirements e Consistent requirements must not generate any conflict with each other e Ranked for importance and or stability every requirement must indicate either its importance or its stability e Verifiable every requirement must be verifiable that is there exists some process to verify that the software complies with every single requirement e Modifiable the structure of the specification allows performing changes to the re quirements in a simple complete and consistent way e Traceable the origin for every requirement is clear and it can be easily referenced in further stages Table 3 1 provides a template for requirements specification CHAPTER 3 PROJECT DESCRIPTION ID Requirement ID Requirement description The development cycle in which the requirement was defined Priority Requirement priority according to its importance A requirement can be identified either as essential conditional or optional Stability Requirement stability defined either as stable or unstable Table 3
61. be observed the time is almost four times lower when the edition tool is used Direct Edition Edition Tool 2720 Std Deviation 1 137 Table 6 13 Times for modeling tasks 6 2 2 Accuracy The accuracy is checked by observing whether the nodes positions in the network modeled by the user correspond with the coordinates of the nodes from the original network Specif ically the error is calculated as the sum of the Manhattan distances for each of the nodes This means if o is the original node and n Vi 1 6 are the nodes placed by the user in the modeled network the error is calculated using the next equation 6 E SN Ini os Inj 0y l Table 6 14 shows the average error in meters in the modeling process for each of the edition methods as well as the standard deviation 102 Circumstantial Knowledge Management for Human Like Interaction Direct Edition Edition Tool 10 9019 meters 3 3970 meters Std Deviation 4 6152 meters 1 2877 meters Table 6 14 Absolute error for modeling tasks However given that the total absolute error may not be significant an alternative relative error can be considered For instance supposing that the technology of the physical layer used has a maximum margin of error of 1 meter it will be considered 1 meter as the maximum acceptable error 100 Then the relative error can be calculated for each of the edition tasks using the next formula ae
62. bject 2 0 0 0 0 000 cee eee eee 123 Survey completed by subjechte s ane ot ose ao ee aw ae ee we 124 Survey completed by subject 4 2 6 2 e aw we ee wR a a we 124 Survey completed by subject De ooo ee eR a we 125 Survey completed by subject 6 ooa See oe Pe EPS SPS 125 Screenshot for the main sereen ca a wR ee ae 130 Screenshot for the situations management screen 131 Screenshot for the network management screen o 132 Sample warning message for uncommited changes 134 Screenshot for the simulation screen 135 List of Tables 3 1 3 2 3 3 3 4 3 9 3 6 3 7 3 8 3 9 3 10 3 11 3 12 3 13 3 14 3 15 3 16 3 17 3 18 3 19 3 20 3 21 Template for software requirements specification 37 Requirement FR 01 Situation model selection 37 Requirement FR 02 Situation model management 37 Requirement FR 03 Situation model simulation 37 Requirement FR 04 Persistent storage o o 38 Requirement FR MA 01 Networks management 38 Requirement FR MA 02 Network selection 38 Requirement FR MA 03 Networks properties 38 Requirement FR MA 04 Network items 38 Requirement FR MA 05 Network items properties 39 Requirement FR MA 06 Network items activeness 39 Requirement FR MA 07 Network items hierarch
63. brief this subsystem must provide logic for the next operations e Management of situations this is creating and removing situations which could apply for the simulation e Management of item features and cost factors by creating and removing features as well as assigning cost factors to the network items These cost factors could be applied directly or by means of features e Network edition which includes the selection of a network the selection and man agement for planes and the management of network items nodes and links and its properties 4 2 2 2 Simulation Subsystem The simulation subsystem is intended to support the simulation functionality whose soft ware requirements are detailed in section 3 2 1 2 Thus this subsystem must provide logic for the next operations e Selection of a network as well as the network plane where the simulation will start e Selection of applying situations for the simulation which will dynamically change the cost for each network item e Motion of the agent over the network through direct motion or by means of a physical layer e Selection of a simulation service description routing or navigation as well as addi tional requirements for the service execution e g a target node for the routing or navigation service 68 Circumstantial Knowledge Management for Human Like Interaction e Execution of the selected service and generation of the service output 4 2 2 3 DataGateway
64. cal circumstantial aspects of the situation Additionally the implemented situation model supports the selection of situations which may belong to any of the context aspects 105 106 Circumstantial Knowledge Management for Human Like Interaction in order to provide partial support to these aspects Apart from the basic situation model a management and simulation tool has been also implemented within this project The edition functionality of the tool allows to edit the spatial network by adding removing or updating its nodes and links as well as it allows the management of context situations When the edition tasks are completed the tool allows to simulate the model s working to check the knowledge validity and finally it allows to upload the result into a given knowledge base for a compatible situation model The simulation functionality of the tool runs over the edited network which is stored in the situation model and provides different services to the user such as the description of his current position and the generation of the optimal route to his goal given the situations that apply during the simulation A physical layer is emulated in order to determine the user location and its motion across the network After the development of the tool an evaluation was carried out to compare the advan tages of the tool over the manual edition of the situation model knowledge The evaluation showed that the edition using the tool improve
65. cation time etc For that reason context aware systems appear as a field of ubiquitous and pervasive computing which refers to systems that adapt their behavior to the current context A more elaborated definition is provided in S nchez Pi 2011 A system is context aware if it can extract interpret and use context informa tion and adapt its functionality to the current context of use The challenge for such systems lies in the complexity of capturing representing and processing contextual data The term context aware applied to the field of computer science was first used in 1994 in Bill N Schilit and Want 1994 Two years before in 1992 the active badge location system made its appearance Want et al 1992 This system which is considered to be the first context aware one serves for the location of people in an office environment Those people were wearing badges which provided information of their location to a central location service CHAPTER 2 STATE OF THE ART 23 To better understand the concept of context aware systems it is important to previously realize what is the meaning of context in the area of computer science Although many definitions have been proposed for the term one of the most accurate meanings is given by Dey and Abowd 1999 Context is any information that can be used to characterize the situation of an entity An entity is a person place or object that is considered relevant to the intera
66. ce it has been found appropriate to place in this screen the about us form which shows the logo and the name of the application Finally a status bar is shown at the bottom providing some feedback to the user on whether the connection to the database was correctly established or an error occurred Figure 4 9 shows the mockup for the first screen of the UL Edition subsystem which allows situations and features management This screen would contain different panels to add and remove situations and features as well as to assign cost factors to the features As this screen will provide means to manage situations and features it requires the next components e A field to list the existing situations in the system The contents of this list may be filtered by a selected situational aspect When a situation from the list is selected its description will be shown so that the user can check that the desired network is the selected one Moreover a button will allow to remove a selected situation A new situation form which will contain the required fields to add a new situation to the system e A field to list the existing features in the system A button will allow to remode a selected feature A new feature form which will allow the required fields to add a new feature to the system A feature edition form which will contain a field to assign cost factors to a selected feature On the other hand figure 4 10 shows a mockup for the s
67. cess for the project will obey the spiral lifecycle model Boehm 1986 This development model combines the simplicity and formality of the wa terfall model with the flexibility of prototyping approaches and turns out to be a good approach for research projects where a complete set of requirements is hardly known from the very beginning and thus may require several prototypes before achieving a satisfactory result like the tool which is object of this project Figure 3 2 shows the concept of the spiral model graphically The spiral model provides four main phases which are repeated for each iteration These phases are the next ones 1 Planning during this phase the objectives for the current iteration are determined and a requirements analysis is performed Each iteration will refine the requirements eventually leading to the complete requirements set for the software application 2 Analysis and Design this phase performs a risk analysis and a feasibility study During this stage a feasible prototype will be generated 3 Development during this phase the prototype is designed implemented and tested 4 Evaluation this phase evaluates the already generated prototype The main pur pose of this evaluation is to detect the defects of the prototype in order to be used when planning planning the objectives for the next iteration The software development will be divided in three cycles each of them providing as a result a new prototy
68. cific way the existence of three areas of the AI contributing to the devel opment of ambient intelligence Cook et al 2007 can be stated The first area is sensing which is required as ambient intelligence systems rely on information about the environ ment The perception task can be accomplished using a variety of wireless sensors which are usually quite small These sensors may optionally perform some local processing of the data before sending it to some servers which gather all the data from the different sensors to produce more accurate and complete information The second area is acting which provides some means to the ambient intelligence system to execute some actions on the environment and affect its users Some systems act through robots which implement some humanoid appearance and behavior thus providing a human like way of communication while other systems rely on other output interfaces such as panels Finally a reasoning process sits between the sensing and acting tasks in order to achieve the intelligent behavior of the system 2 3 Context Aware Systems Ubiquitous computing and ambient intelligence introduce some challenges that are not present in traditional computing since systems must now be capable of working in highly dynamic environments Henricksen and Indulska 2005 For instance in some mo bile devices it may be useful that applications react to the rapidly changing context S nchez Pi 2011 such as the lo
69. comparative system evaluation over manual edition Finally chapter 7 provides conclusions on the project and also provides some lines of future work in the research and development of the project After the main content the document includes five appendices Appendix A presents different scenarios where the situation model could be useful whereas appendix B describes some ethical considerations in the project Appendix C presents the original data gathered during the evaluation process Finally appendices D and E contains an installation guide and a user manual respectively This page has been intentionally left blank Chapter 2 State of the Art The purpose of this chapter is to introduce the state of the art for the project Sections 2 1 to 2 6 aim to introduce key concepts which are related to the topic of this project as well as to provide a historical overview on how these concepts evolved eventually leading to the appearance of natural interaction and situation models Although there are few practical applications of situation models section 2 7 describes systems where situation models have been successfully applied Finally some conclusions are provided on the state of the art 2 1 Ubiquitous Computing In the late 1980s the term ubiquitous computing was coined to refer to a computing paradigm in which computers are everywhere and they turn out to be invisible for hu mans meaning that the user can focus on the task he is pe
70. computer so that a context aware system can use this knowledge to suuport its work Baldauf et al 2007 According to Strang and Linnhoff Popien 2004 most context aware systems highly demand six main requirements on context models First they must have a distributed composition as pervasive devices lack of a central system and the composition of a context model varies notably in terms of time source and network topology Secondly it is desirable to be able to partially validate contextual knowledge as the complexity of contextual interrelationships make the modeling intention error prone In the third place given that the quality richness and accuracy of the information varies over time depending on the technology used a context model should support quality and richness indication Moreover as contextual information uses to be incomplete or ambiguous mostly due to technology 24 Circumstantial Knowledge Management for Human Like Interaction limitations the context model must be able to deal with this ambiguity In the fifth place contextual facts and interrelationships must be described in a precise and traceable manner so that each participant in an ubiquitous interaction shares the same interpretation of the data exchanged and its semantics Finally a context model must be applicable within existing ubiquitous computing infrastructures 2 4 1 Context Modeling Approaches Strang and Linnhoff Popien 2004 summarizes the most rele
71. ction between a user and an application including the user and applications themselves Moreover Dey and Abowd present different categories of context types which can be classified as primary and secondary The primary context types are location identity time and activity which answer the questions where who when and what respectively The secondary context types contains other pieces of contextual information such as details about places people or objects e g a person email address A simple approach to context aware systems is known as location aware systems which is a subset of context aware systems only dealing with location information These systems are rapidly growing due to the spread of mobile devices and applications Baldauf et al 2007 and as a consequence of it it can be found many systems and devices that provide some functionality based on their current position or proximity to other de vices Some examples of this kind of systems are navigators smartphones cameras with geotagging functionality social networks game consoles etc Many of these systems uses technologies such as Global Positioning System GPS and wireless networks for retrieving their current location 2 4 Context Models For a context aware system to work it is required that the contextual information is some how formalized and stored A context model formalizes stores and shares contextual in formation in a way that can be processed by a
72. ctor and the sign of the coefficient of the component of the third dimension will determine the angle direction The sign of the result will be added to the angle to indicate its direction In our example the value of the cross product is i gk 7 Vuser X Vtarget 1 1 0 3k 2 10 where k represents the direction of the third dimension and so the angle 0 113 So far the system is able to provide six different instructions each one for a given interval of angles These instructions are shown in table 5 1 It can be seen that for the previous example the system provides the instruction to turn right in order to achieve the next node 30 to 60 Turn slightly right 60 to 120 Turn right S120 oF lt 100 Table 5 1 Routing instructions The RoutingServices class provides a service quite similar to the navigation service The only difference is that while the navigation service only inform the user about the next step he must take to achieve its desired goal the routing service provides a list of instructions to eventually achieve the final goal 5 5 Functional Architecture Presentation Tier For this tier we have decided to implement one class for each screen in the GUI Although some additional classes and external libraries are required these will not be detailed here as most of them act as black boxes and serves for specific purposes 90 Circumstantial Knowledge Management for Human Like Interactio
73. d networks will be associated to a corpus and the user will be required to select a corpus in the management tool Table 7 13 Future work FW 13 Integration in Cognos Knowledge base FW 14 D A S S O Integration in Cognos Ontology Description In the future features and situations could be integrated with the Ontology so that their identifiers are no longer strings but concepts with an associated semantical value Table 7 14 Future work FW 14 Integration in Cognos Ontology 112 Circumstantial Knowledge Management for Human Like Interaction WE Full integration in Cognos Description A full integration in the Cognos toolkit would enrich the annotation in other models by supplying circumstantial metadata to them Table 7 15 Future work FW 15 Full integration in Cognos PWT Acquisition of knowledge of a real scenario Description The Cognos toolkit may be used to acquire knowledge over real scenar ios so that this knowledge can feed a natural interaction system with a physical layer Table 7 16 Future work FW 16 Acquisition of knowledge of a real scenario 7 2 4 Natural Interaction System Finally some future works are proposed which have to do with the integration of the situation model in the natural interaction system this is in the cognitive architecture presented in figure 1 1 FWA Model installation Description A new function will be included in the tool in order to install
74. d VisualNodeByNode nodes Set lt VisualNode gt id int VisualNode getClosestNode nodes Set lt VisualNode gt x double y double VisualNode Figure 5 2 Classes for the EditionKnowledge subsystem two nodes in different planes This explains why the VisualNetwork class contains a set of visual links while the Plane class contains a set of visual nodes e The VisualNode class serves as the graphical representation of a node It contains two position attributes x and y which represents the visual node position over the plane Moreover this class also contains an attribute which references to a node in the ModelKnowledge subsystem This class presents several static methods which may be useful for some purposes The first method allow to find a visual node in a set given a node from the ModelKnowledge subsystem Another static method allows to find the visual node in a set which is closest to a given coordinates e The VisualLink class serves as a graphical representation of a link It contains two attributes which reference the start and end visual nodes as well as another attribute to refer the actual link in the ModelKnowledge subsystem 5 4 Functional Architecture Logic Tier This section will detail the different classes of the logic tier Moreover it will discuss the choice of an algorithm over others when it is required 82 Circumstantial Knowledge Management for Human Like Interaction 5 4 1 DataGateway Sub
75. d requirements no technical constraint in the client side compromises the project feasibility CHAPTER 3 PROJECT DESCRIPTION 49 3 4 1 2 Server Requirements As most of the application computations fall on the server side the server must fulfill some specific technical requirements Particularly the server will run an Oracle Database to store all the model knowledge in a persistent way Moreover Oracle Spatial will load the network in main memory to run graph algorithms over it which may require more RAM memory to be available The actual server specifications will depend on how the application will be used being some key factors the size of the knowledge to be stored and the complexity of the networks The next requirements should suffice for a common use of the application e Operating system Microsoft Windows Server 2003 Windows Server 2008 or Win dows Server 2011 e Processor AMD Opteron Series 2300 Quad Core 2GHz or higher or Intel Xeon Series X3200 2GHz or higher e RAM Memory at least 4096 MB e Storage HDD of 300GB at 15000 rpm e Network an Internet connection must be available e Software Oracle Database 11g Enterprise Edition The use of Oracle Database for persistent storage is justified for several reasons First the Advanced Databases Group already owns software licenses for Oracle Database and the developers are trained in the use of this database management system Moreover the previous work describ
76. d the performance and increased the accu racy of the process and was also preferred by all the experiment subjects when they had to consider the comfort intuitiveness reliability and agility of the two methods Moreover the simulation functionality provides a way of checking the validity of annotations when uploading the knowledge into the base of the production system This work will be part of a research project in the LaBDA group which will improve the situation model and the tool will study the implementation of many of the future works described in the next section and will integrate the model within the cognitive architecture for a natural interaction system presented in figure 1 1 A copy of the Cognos toolkit including the edition tool developed within this project is available in the next site http labda inf uc3m es doku php id es labda_lineas cognos Moreover comercial applications of the developed model can be found in the area of software for smartphones where a location aware system can be enriched with the situation model While most of the situation models existing in the state of the art are implemented for a specific system the situation model of this project is generic and could be applied to any system Moreover the management tool allowing the knowledge edition of the situation model is a contribution to the state of the art as no similar tools exist even if the developed tool only supports the knowledge man
77. developed in order to ease this task by avoding the use of specific SQL commands The second way will edit the model by means of the management tool developed in this project Figure 6 1 shows the instructions sheet which was distributed among the subjects which details the tasks which have to be executed in the experiment This experiment will enable to gather both qualitative and quantitative information e Through direct observation the experiment manager will be able to measure the times for each of the tasks It is expected that the edition tool increases the performance of the operation e After the tasks are completed the experiment manager will be able to measure the accuracy of both systems by observing the deviation of each node to its original location CHAPTER 6 VALIDATION AND EVALUATION 99 Cognos S Circumstantial Knowledge Management Instructions Please start by reading the instructions carefully In order to evaluate the developed management tool you will be required to edit the next network for a situation model re Le ee a EE 3 1 NV Id INLLVaVS mI Reference points Al and A2 are physically located in the coordinates 0 0 and 100 50 The coordinates for each node must be extracted from this information as checking the accuracy is part of the evaluation You are asked to perform the next tasks A Model the network directly over the model knowledge base For this purpos
78. e you will be provided with access to a database and will be able to execute SQL instructions You must follow the next steps 1 Establish a connection to the database the access information is attached in a different document 2 Create a new network in the table Networks named Ev your_name Man You can use the next procedure insert_network network_name returns network_name 3 Model the network provided in the figure The next procedures are provided to ease this task insert_node network_name x y returns node_id insert_link network_name nodel_id node2_id returns link_id B Model the network using the management tool For this purpose you will be given a copy of the edition tool the plane background image and the user manual You must follow the next steps 1 Start the application and establish the connection to the default database 2 Go to the network management tab create a new network called Ev your_name Tool Load the network 3 Select the provided image as background image 4 Model the network provided in the figure You can use the reference nodes feature to avoid calculating the position for each node You can change the node scale if you wish After finishing with the two operations please fill the attached survey Figure 6 1 Instructions for the experiment e By filling a survey the experiment could also gather some subjective information such as whether the user considers that the sys
79. e shortest route which passes by the entry of the library and enters the Sabatini park with the fountain just before going into the building by gate A This route is shown in the figure as M A However there are some issues for which this route is not convenient In the first place entry A to Sabatini is closed after 5 00 PM so Martha will not be able to access the building by this gate Secondly Martha did forget her umbrella and she really hates walking under the rain something that wants to avoid as long as possible Martha who is a very clever girl starts her navigation system empowered by the situ ation model and asks for the best route The system is perfectly aware of the context it knows that entry A is closed after 5 00 PM and it also knows that the weather is really annoying for Martha Finally the system proposes an alternative route M B which goes from Martha location 115 116 Circumstantial Knowledge Management for Human Like Interaction Figure A 1 Map of the campus illustrating the first scenario to entry B which still opens at 9 00 PM Moreover this route is covered by the buildings so Martha will get to her goal dry and happy A 2 Enhancing a Natural Interaction System This scenario shows a sample dialog between a user and a natural interaction system which is enriched by a situation model Arthur is in the building of CafreSoft Corporation a young and promising software company He has just finish
80. e ubiquitous computing era e The ubiquitous computing era in which each person owns many computers which are interconnected A few years later in 1994 another term closely related to ubiquitous computing ap pears this term is pervasive computing It was first used to refer to Novell strategies for connecting people to people and to information any place at any time Schofield 1994 After some years of decline the term reappered in 1998 in IBM s post PC world Lohr and Markoff 1998 this time meaning that computers were everywhere The main objective during that time was that computer devices could be everywhere and that they allowed their owners to be connected to the Internet in any place at any moment Despite the fact that the terms ubiquitous computing and pervasive computing are mostly used as synonyms today their original meanings were significantly different Ronzani 2009 while the term ubiquitous computing referred to an Human Computer Interaction HCI style which turns out the system to be invisible to the user which is obtained through miniaturising some computer devices and reducing CHAPTER 2 STATE OF THE ART 21 their costs the term pervasive computing focused on the spread of portable and handheld systems which allowed the user to be connected to the Internet everywhere Nowadays the appearance of many handheld computer devices such as smartphones and tablets are taking us to a progressive achievement of ubiqui
81. econd screen of the UL Edition subsystem which allows networks management This screen contains panels for networks 74 Circumstantial Knowledge Management for Human Like Interaction Circumstantial Knowledge Management for Human Like Interaction _ x Start Situations Management Edition Simulation m Situations Features Aspect Situational Aspect N Situations Feature 1 Foe Feature 2 Situations Situation 1 Situation 2 Text showing the selected situation New Feature description Name Name m New Situation Add Name Name Costs Factors Descr Description Situation Cost Factor Situation 1 v 2 Situation 2 v 3 5 Add Remove Save Status Figure 4 9 Mockup for the situations and features management screen and planes creation deletion and selection Once a network and a plane are selected the plane image is shown and the network edition can be performed over it adding or removing nodes and links as well as editing its properties The second screen would contain all the controls to allow the networks management A chooser element to select an existing network in order to load it as well as a text to show the description for the chosen network Additionally a button must allow the removal of the selected network A new network form with all the required
82. ed a job interview and is really satisfied It is 11 30 AM and it takes a few hours for him to go home so he has decided to have his lunch in the building He starts a dialogue with his personal assistant Hal ARTHUR Hi HAL Look ld like to eat something here What do you say APPENDIX A SCENARIOS 117 T he system realizes that it is 11 30 in the morning It knows that ARTHUR usually have breakfast at 9 00 AM and lunch at 1 00 PM The system asks ARTHUR for more information HAL Nice It s eleven thirty would you rather have breakfast or have lunch ARTHUR Today Ill have something for lunch The system looks for places of interest near ARTHUR HAL Ok There is a snack bar in this floor but it is only serving breakfast now You can have a sandwich there If you prefer you can go to the restaurant to have a meal and you can eat from the menu until 3 o clock in the afternoon ARTHUR Fine let s go there The system tries to disambiguate the anaphora there by asking the user HAL To the restaurant right ARTHUR Right HAL Ok Go ahead until the end of the corridor Then you ll have to go to the first floor You can take the elevator ARTHUR Nice By the way do you know where can I find a bathroom T he system looks for bathrooms He finds a bathroom in the current floor but it is only for the company staff He looks for alternatives and assists ARTHUR HAL There is a bathroom in this floor but
83. ed in section 3 3 was implemented over Oracle Database thus the use of this software would allow code reusability and provides a better intuituion that the current project is feasible The Advanced Databases Group provides a server which fulfills these requirements for this project so there are no technical constraints on the server side that could threaten the project feasibility 3 4 2 Environmental Constraints Environmental constraints will be classified attending to their nature Thus it can be found constraints that have to do with the application social environment and the project stakeholders sociocultural constraints others which are founded over economic aspects economic constraints and those who are related to the legal framework of the project legal constraints Finally ethical considerations are discussed in appendix B 50 Circumstantial Knowledge Management for Human Like Interaction 3 4 2 1 Sociocultural Constraints The edition tool is aimed at members of the academic community The users of the appli cation may not be familiar with this kind of advanced edition tools and for that reason the application developed in this project must have an easy to use friendly interface This fact is not an obstacle as many modern programming languages allow the creation of attractive user interfaces 3 4 2 2 Economic Constraints The work in this project could eventually be included in a research project which may provide fina
84. el could then turn into the Big Brother that knows absolutely everything about everybody For this reason it is important that the system does not store more information that the one that is absolutely required to provide the service Furthermore it is essential that the user is notified about the data that the system is storing about him and the reasons why this data is required so that he can freely decide whether he accepts to use the system or not Appendix C Evaluation Data This chapter shows the original data gathered during the evaluation process and it can be used to get a better idea on how the edition tool compares to the direct edition in terms of performance and accuracy as well as what are the subjects opinions regarding the two systems From now on System A is the name given to the system which allows direct edition of the model knowledge and System B is called to the edition tool developed in this project Table C 1 shows the execution time of both tasks for each of the experiment subjects Table C 1 Time measurements for the evaluation Tables C 2 and C 3 show the position of the nodes in the networks modeled by each of the subjects for both the direct edition system and the edition tool respectively This information has been used in the evaluation as a measure for the accuracy The first row contains the original position for each of the nodes while the others display the coordinates of the nodes in t
85. em which already satisfied the preconditions If that happens then the test is successful and the requirements covered by the validation test are satisfied and this fact is indicated with a green checkmark Otherwise it is important to analyze why the test failed and to propose a corrective action In this case the requirements which are not validated can be essential or not and the failure is indicated with a red or orange cross respectively Tables 6 1 to 6 12 details the results for all the validation tests 95 96 Circumstantial Knowledge Management for Human Like Interaction Validation Test Test 01 table 3 49 Not required Table 6 1 Results for validation test 01 Database connection Validation Test Test 02 table 3 50 Not required Table 6 2 Results for validation test 02 Situations management Validation Test Test 03 table 3 51 7 Ok Not required Table 6 3 Results for validation test 03 Features management Table 6 4 Results for validation test 04 Network selection Validation Test Test 05 table 3 53 Result X Wrong not essential Planes management is not imple mented Corrective action Due to its cost planes management feature has not been im plemented So far only one default plane per network its supported For this reason the requirements FR MA 13 FR MA 14 and FR MA 15 are not validated However all these requirements have their priority defined as optional and
86. etwork items cost factors 43 Requirement FR MA 28 Features cost factors o o 43 Requirement FR SI 01 Network selection 43 Requirement FR SI 02 Network display 44 Requirement FR SI 03 Person placement 44 Requirement FR SI 04 Person motion 04 4 44 Requirement FR SI 05 Physical layer simulation 44 Requirement FR SI 06 Simulation services 020004 44 Requirement FR SI 07 Description yong eases ee oneness we a 45 Requirement FR SI 08 AQUILES AAA Se it 45 Requirement FR SI 09 Navigation li oe Bee be ee 45 Requirement FR SI 10 Source and goal location 45 Requirement FR SI 11 Minimum cost path 45 Requirement FR SI 12 Situations selection 46 Requirement NFR SC 01 Big networks o o 46 Requirement NFR SA 01 Authorized access o o 46 Requirement NFR IO 01 Cross platform o o o 47 Validation test 01 Database connection e o 51 LIST OF TABLES 9 3 50 3 51 3 02 3 93 3 54 3 99 3 56 3 57 3 58 3 59 3 60 3 61 3 62 3 63 3 64 5 1 6 1 6 2 6 3 6 4 6 5 6 6 6 7 6 8 6 9 6 10 6 11 Validation test 02 Situations management 52 Validation test 03 Features management oa a 52 Validation test 04 Network selection o o o
87. ever the study of the imple mentation of other algorithms is proposed as a future work It must be noticed that performance can also be improved if the system is not required to calculate the path every time the user moves to a new node For these cases the system may check whether the new node is included in the previously calculated path and take advantage of this to save computational effort 5 4 3 3 ServiceProvider The ServiceProvider interface in motivated as new services can be easily integrated into the system as long as they comply with the interface This interface contains only the method executeService which must provide all the logic for the service execution and returns its output So far three classes implementing the ServiceProvider interface will provide the re quired services the DescriptionServiceProvider class provides the description of the user location the RoutingServiceProvider and NavigationServiceProvider classes ob tain the shortest path between the user location and its desired goal While the routing service prints all the route to the goal the navigation service provides the instructions one by one while the user is moving across the network These classes require to store the target node in order to calculate the path The DescriptionService class provides the description of the node which is closest to the user location The NavigationService class generates an instruction which the user must follow t
88. fears are unfounded or not it is true that interest in AI is accompained by some distrust The purpose of this appendix is to evaluate the consequences of the development of this project over the society When scientists try to build computers that emulate the human behavior we first have to model in some way our own knowledge in order to introduce it into the computer However most of the times we are not aware of our own knowledge as many things seems to be natural for us In the last years with the expansion of the social networks we can realize that in some cases the Internet actually stores more information about us that the one that we have Even if all this information is spread in different computers some data mining techniques can be applied to integrate it all and generate an impressively huge material about ourselves Although the situation model developed in this project is relatively simple future works are intended to improve it in order to allow the support of several agents for example the 119 120 Circumstantial Knowledge Management for Human Like Interaction current user and also other people and objects and the consideration of all the context aspects in the network graph apart from the spatial dimensions In this case the situation model could be storing a lot of confidential data about the users Not only its current position but also the evolution of its location its role or its tasks over time The mod
89. for the DataGateway subsystem 5 4 1 1 DBConnection The DBConnection class allows to execute basic operations over the database It uses the Oracle Database 11g JDBC driver library which provides Java classes to establish a communication with an Oracle database This class provides methods to e start a connection with the database given the information required for the connec tion to be established e close an open connection to the database e inform whether a connection is established or not CHAPTER 5 IMPLEMENTATION 83 e execute a query over the database and return a set of results e execute any other SQL instruction such as insertions deletions or updates e commit or rollback the changes if applicable All the methods regarding the execution of SQL instructions must be designed to prevent SQL injections This can be easily achieved using the tools provided by the Oracle Database JDBC library The application may take advantage of the use of a transactional database management system in order to allow transactions for the network edition process Thus the user is given the option to commit or rollback the changes at any time 5 4 1 2 DataGateway Classes DataGateway classes provides basic CRUD functionality for each of the classes in the data tier The next methods are provided by these classes e Insert a new object in the database e Remove an existing object from the database This function is not available for
90. gement system This decision was taken as the developers had previous experience on the design and development of relational databases The chosen database management system is Oracle Database 11g as it was already stated in the feasibility study as licenses were provided by the LaBDA research group and it can be easily integrated with Java to execute CRUD operations from the application 5 2 Physical Architecture Section 4 1 specified the physical architecture to be implemented for the system which was a client server model Actually this architecture have been implemented using only one server which was provided by the LaBDA group This server contains two databases the situation model CHAPTER 5 IMPLEMENTATION 79 knowledge and the edition knowledge Cognos Moreover one desktop computer is being used as the development computer and also as a client running the edition and simulation application The technical specifications for these machines were specified in section 3 4 5 3 Functional Architecture Data Tier This section will detail the different classes of the data tier Additionally it will include a relational graph which can be used for the implementation of the database model 5 3 1 ModelKnowledge Subsystem Figure 5 1 displays the class diagram for the classes within the ModelKnowledge subsystem It is important to notice that from now on the diagrams will not show the getters and setters methods as all fields in a
91. h this purpose in mind some particular knowledge about the context may assist the system to provide a more elaborated interaction thus achieving a more natural and realistic way of communication with a computer system 1 1 Theoretical Background In the first place a theoretical background for this project is provided We will focus on briefly describing what the circumstantial knowledge is and how a situation model can enhance a human like interaction system in order to provide a more complete and realistic experience Circumstances are defined as facts or conditions which are relevant to a particular event or action Regarding an interaction process we can define the circumstantial knowledge as the set of knowledge and skills regarding the context and therefore the circumstances of the interaction The circumstantial knowledge can be classified in different aspects According to Gee taxonomy Gee 1999 the context in a communicative social interaction comprises the fol lowing components a semiotic aspect regarding the language used in the communication process an activity or operative aspect concerning the transactions the major activi 13 14 Circumstantial Knowledge Management for Human Like Interaction ties or tasks underlying the communication a material aspect respecting to the physical situation of the action that is mainly the spatio temporal context including people and objects taking part on the interaction yet als
92. he networks modeled by each of the subject Each cell is composed of two rows containing the z and y coordinates of the node respectively 121 122 Circumstantial Knowledge Management for Human Like Interaction Direct Edition Diret Edition OO eee 6 26 11 35 16 19 11 35 16 08 53 66 6 26 11 12 15 99 11 12 15 99 54 21 6 58 11 84 15 79 11 84 15 79 53 29 7 87 12 12 16 36 12 12 16 36 49 69 6 90 11 73 14 75 9 99 14 65 54 51 6 60 9 90 14 52 9 90 14 52 50 82 6 09 11 06 15 76 11 13 15 76 53 31 Table C 2 Accuracy measurements for the evaluation with direct edition Edition Tool 6 26 11 35 16 19 11 35 16 08 53 66 6 39 11 85 16 11 11 85 15 76 53 79 6 28 10 78 15 88 10 90 15 76 53 67 6 58 11 75 16 09 11 75 15 98 54 52 6 47 12 12 16 24 12 12 16 24 54 94 6 57 11 62 16 19 11 50 15 85 53 17 6 38 11 58 16 31 11 70 16 19 53 78 Table C 3 Accuracy measurements for the evaluation with the edition tool Finally figures C 1 to C 6 contains the results of the surveys which were filled by the subjects after completing the experiment All the personal information has been removed from the survey in order to preserve the subjects confidentiality APPENDIX C EVALUATION DATA 123 1 Evaluate the next statements in a scale from 1 to 5 being 1 strongly disagree and 5 strongly agree for each of the two systems Question System A System B I felt good when performing the task the system is
93. his section is to discuss some natural interaction systems which have either contributed to the research in situation models or where these models have been successfully applied 2 7 1 TRIPS TRIPS The Rochester Interactive Planning System Ferguson and Allen 1998 is a system which integrates several AI techniques in order to provide the user with an interactive problem solving assistant in a logistic domain This research project was carried out in the University of Rochester s Department of Computer Science in the late 1990s TRIPS is built over the TRAINS system Allen et al 1994 although it allows planning on more complicated domains The main purpose of TRIPS system is to assist the user to build plans in crisis situations The generated plan may have to meet some specific constraints such as time cost or weather conditions Thus the system may require some circumstantial knowledge to generate an adequate plan The TRIPS architecture shown in figure 2 1 is divided in three groups The first one has to do with modality processing which mostly means input recognition and output generation and the third group contains the specialized reasoners such as the planner or the scheduler each one being able to solve some particular problems Meanwhile the second CHAPTER 2 STATE OF THE ART 29 Map Keyboard Microphone Mouse Menu Displays Speakers Spelling Speech Correction Synthesis Natural Language Realization Parsing Generation
94. ibing the components for each of the screens 4 1 Physical Architecture The physical architecture for the software is based on a client server model The server or servers will provide two differect roles the storage of the situation model and the storage of the edition information for the cognitive architecture Cognos The clients will connect to the servers both for the use and the management of the situation model It must be pointed out that this architecture requires the servers to be accessible by all the clients Actually the server which stores the situation model is replicating a subset of the edition data in the Cognos server This fact provides a natural redundancy for the model data However external mechanisms must guarantee the consistency of this replica to ensure that the modifications performed using the management tool are also carried out in the situation model Some advantages of this physical architecture is that it supports concurrent edition and operation over the situation model thus the clients can take advantage of this parallelism high scalability confidentiality and reliability as long as the database management system 63 64 Circumstantial Knowledge Management for Human Like Interaction guarantees these principles Figure 4 1 shows the physical architecture for the system graphically Client p A Edition Client 1 Edition Client 2 Situation Model Knowledge Base Edition Cognos Kno
95. ical CLIs where the commands were required to be typed in the screen and GUIs where the user interacts with the computer through images which usually served as metaphors for daily objects A key concept related to NUIs has to do with multimodality In a typical human CHAPTER 2 STATE OF THE ART 21 human conversation each individual does not only communicate through spoken language but also body language takes an important role Multimodal interfaces allow the user to communicate through different modalities such as speech written language gestures etc leading to a more natural and transparent interactive experience The use of multimodal NUIs provides several advantages Oviatt and Cohen 2000 In the first place they have the potential for increasing the system accesibility usability and comfortability These systems usually have a very simple learning curve and can often be used by users of different ages and skills and even by impaired users Moreover multimodal interfaces usually allow the user to switch between different modalities to adapt to the user necessities Secondly multimodal interfaces usually improve the robustness and accuracy of the system a property that can be achieved by the mutual disambiguation of the different inputs Finally these interfaces provide great expressive power to deal with the exchange of information given that each interface complements the other ones While most of the natural interfaces
96. iental political sociocultural operational and semiotical Cycle Third Priority Essential Stability Sabhe Table 3 29 Requirement FR MA 24 Situational knowledge taxonomy D MAA Situations management Description The system shall provide means to the user to create and delete situations within the situational aspects specified in FR MA 01 Cyce Third Priority Essential Stability Stabe Table 3 30 Requirement FR MA 25 Situations management CHAPTER 3 PROJECT DESCRIPTION 43 FIRMA Description Situations are identified by a name and can have an optional de scription Cycle Third Priority Esc al Stability Stabla Table 3 31 Requirement FR MA 26 Situations properties PRAT Network items cost factors Description The system shall provide means to the user to assign cost factors cost multipliers to network items for each defined situation Cycle Third Priority Conditional Stability Stable Table 3 32 Requirement FR MA 27 Network items cost factors FERMA Description The system shall provide means to the user to assign cost factors to defined features for each situation Cycle Thixd Priority Essential Stability Stable Table 3 33 Requirement FR MA 28 Features cost factors 3 2 1 2 Simulation This subsection describes in further detail the user requirements for the simulation func tionality FR SEOL Description
97. ies Moreover it provides some features such as encapsulation inheritance and polymorphism which can be really useful for our purposes e The principle of encapsulation allows a class to hide its own attributes Instead public methods describes the functions which can be executed over an instance of the class also known as an object This provides modularity and allow to use classes as black boxes where nothing but its functionality is required to be known e The principle of inheritance allows to create a class hierarchy where some classes are childs of other classes Often these child classes provides some specialization TT 78 Circumstantial Knowledge Management for Human Like Interaction over their parents or some additional features Moreover abstract classes can be created which cannot be instantiated directly Thus a hierarchy with several levels of abstraction can be built by means of inheritance e The principle of polymorphism allows to hide the actual type of the object The variable type can be known for being an interface or a parent class in a hierarchy If the actual type of the instance complies with the known type then its methods can be called Once the object oriented paradigm has been chosen as the preferred programming paradigm for this application the next step consists on choosing the programming language for the system implementation Although there are several alternatives of object oriented languages
98. igence technolo gies requires them to be sensitive responsive adaptive transparent ubiquitous and intelligent Cook et al 2007 Thus it can be realized how closely re lated is ambient intelligence with other concepts such as ubiquitous computing pervasive computing and context aware systems In order to be able to provide these features the IST Advisory Group Ducatel et al 2001 described five key technological requirements the need of very unobstrusive hardware which can be obtained through miniaturisation a seamless web based communications infrastructure to allow communi cation and interoperability of different devices and networks dynamic and massively dis tributed device networks natural feeling human interfaces to support the intuitive use of the system and dependability and security Although the concept of ambient intelligence may seem identical to that of ubiquitous computing the main difference lies in the fact that ambient intelligence incorpo rates a strong component of Artificial Intelligence AI encompassing contributions from machine learning multiagent platforms and robotics For instance ambient intelligence 22 Circumstantial Knowledge Management for Human Like Interaction may include works about voice technologies artificial vision natural language and knowl edge management Eisaku and Yasuhiru 2006 so that they can show a more sensitive responsive adaptive and ubiquitous behavior In a more spe
99. ility Study The purpose of this section is to discuss the project feasibility given the requirements specified by the user as well as technical and environmental constraints 3 4 1 Technical Constraints The application will follow a client server model thus the client and the server technical specifications for the application to work correctly have to be described separately 3 4 1 1 Client Requirements The client will execute the edition and the simulation software which does not require too much computing power given that expensive operations such as optimal paths calculations are intended to be performed in the server However some technical specifications are recommended for the user to obtain the best experience from the application e Operating system Microsoft Windows XP Windows Vista or Windows 7 e Processor Intel Core 2 Duo 2GHz and higher or AMD equivalent e RAM Memory 1024 MB e Storage 2 GB of free space in HDD e Network an Internet connection must be available e Screen screen resolution of 12801024 or 1440x900 in widescreen mode e Software Java Virtual Machine 1 6 or higher The use of Java Virtual Machine is justified as it provides cross platform capabilities as it was stated in requirement NFR IO 01 Additionally the development team is already trained in the development of applications and GUIs in the Java programming language As most commercial personal computers exceed by far these recommende
100. im to improve the functionality of the situation model 110 Circumstantial Knowledge Management for Human Like Interaction PIS Native support for the temporal aspect Description While so far the situation model only provides native support for the spa tial aspect through a three dimensional network the temporal aspect is only supported by means of situations Native support for this aspect could be implemented directly in the situation model by using temporal triggers or cron jobs Table 7 8 Future work FW 08 Native support for the temporal aspect FW Native support for all the context aspects Description The situation model could be improve to natively support all the context aspects instead of just the material While these aspects are currently being supported by means of situations native support would provide a better integration among them thus increasing the power and reliability of the system as well as going a step further in the research of situation modeling The result would be a multidimensional network where each aspect of the context would contribute by adding one or more dimensions to the graph On the other hand the implementation of this future work can turn out to be really hard and for this reason this future work is proposed with the least priority Table 7 9 Future work FW 09 Native support for all the context aspects Description In order to calculate the shortest path the system i
101. ion instruc tions D 2 Connecting to the VPN Although the VPN connection is not required to load the application it will be required in order to connect to the default database if the application is running outside the university network The next page contains a manual for establishing the VPN connection from different operating systems https asyc uc3m es index php Id 168 127 128 Circumstantial Knowledge Management for Human Like Interaction D 3 Loading the Application After the Java installation process concludes you will be able to open the jar file con taining the management and simulation tool In some systems double clicking on the file must be enough to open the application However other operating systems may require to open a command window place the bash directory where the jar file is contained and execute the next command java jar cognos_s jar Appendix E User Manual This section will explain the use of the tool developed for this project The design and structure for this application was discussed in chapter 4 As that chapter explained the application contains four different screens each of them in a different tab This manual describes the functionality provided in each of the screens E 1 Main Screen Figure E 1 shows an screenshot of the main screen which provides functionality for estab lishing and closing a connection to a database The components of the figure are explained here 1
102. is project is available in the next site http labda inf uc3m es doku php id es labda_lineas cognos ix This page has been intentionally left blank Keywords Circumstantial Knowledge Situation Model Natural Interaction Context Context Aware System Knowledge Base Knowledge Management Management Tool xi This page has been intentionally left blank Contents Agradecimientos vii Abstract ix Keywords xi Contents 1 List of Figures 5 List of Tables 7 1 Introduction 13 1 1 Theoretical Background aos a E Seo A dd 13 LD Project ODJeCtHIVeS acis t se A t a Oe A a oe a ia 14 1 3 Project Cont xt bea dal a a E eM E OE ER Re N OME ES 15 1 4 Document Structure aoaaa oe Be Be E AE eT YB 16 2 State of the Art 19 al U big ito s COMPp tino ap os s Ae ld By Ae ees ey A ews ee A 19 2 2 Ambient Intelligence zos sis eS By ces ERS Y Ge EE Y Es Ne Re n 21 29 Context Aware Systems 314 Dev ai Aap EE SY a ER STARE BE A 22 24 Context Models si pos ai dea a ee fal de EAn ae e D aS al de AS 23 2 5 Natural Interaction OE Ho SY Bo A 24 Circumstantial Knowledge Management for Human Like Interaction 20 Pitation Models 10950 es Hood Bente Bord A Herd Sark Hook Geos Seok 28 2 Applications AAA AS OE ae AH AS 28 Dies Gonel sions y SS ok PG TI REI ESA E RR 33 Project Description 35 3 1 Requirements Elicitation sa 3 054 hia es ita le A ee a 35 3 2 Software Requirements Specification o 36 397 Previous WO L
103. jective evaluation reliable Agile Direct Edition Strongly agree Strongly disagree Neither agree nor lisagree A Disagree Agile Edition Tool Agree Strongly agree a Direct edition b Edition tool Figure 6 6 Results for the subjective evaluation agile 6 2 4 Conclusions After the experiment is concluded some conclusions are extracted regarding both the quan titative measurements performed during the evaluation time and accuracy and the sub 104 Circumstantial Knowledge Management for Human Like Interaction jective evaluation completed by all the subjects In the first place the information in table 6 13 shows that the tool improves the efficiency of the situation modeling task in a significant factor Regarding the quality of the annotations by observing table 6 15 it can be concluded that the tool increases significantly the accuracy of the network modeling as the relative error is remarkably lower when the edition tool was used The results in figures 6 3 to 6 6 clearly show that the experiment subjects prefer the tool edition over direct edition of the model knowledge In average users think that the edition tool is more comfortable 4 83 vs 2 33 more intuitive 4 33 vs 2 67 more reliable 2 83 vs 4 and more agile 4 83 vs 2 17 Finally some other conclusions can be extracted from subjective evaluation supported by the open questions in the survey
104. kup for the simulator screen This screen contains all the elements described above which are required to perform a simulation over the situation model The simulation screen would require the next components e A chooser element to select an existing network in order to load it as well as a text to show the description for the chosen network e A chooser element to select a plane for the selected network 76 Circumstantial Knowledge Management for Human Like Interaction Circumstantial Knowledge Management for Human Like Interaction ala x Start Situations Management Edition Simulation m Networks Planes Network Network y Plane Plane v The selected network description r Network Simulation a Situations gt f yO l Situation 1 VY Situation 2 GPS X X coordinate Y Y coordinate Send RFID ID Node ID l Send Services r Output GomjiCehi x The output text generated by the running service Load Image Zoom 7 Status Figure 4 11 Mockup for the simulator screen A panel to display selected network plane A slider to zoom over the image e An image representing the agent which can be placed over the plane image e Mouse and keyboard events required to place and move the agent over the network e A panel for each physical layer emulator For the m
105. l find a route that goes through ramps and elevators rather than doing through stairs This route will also prefer the automatic doors Even if the route is longer and requires more time to get to the goal the system will prefer it as it has less cost for Alan Appendix B Ethical Considerations In 1867 Alfred Nobel patented the dynamite which was obtained as a result of adding kieselguhr to nitroglycerine taking the form of a moldable paste The invention increased significantly the safety and reduced the costs of mining works such as rock blasting tunnels drilling etc However dynamite also got an important role in the military industry Nobel which was very interested in social and peace issues and made a great fortune with his inventions started the Nobel Foundation and when he wrote his final will he included a Prize for those persons or organizations which promoted science and peace A summary of Nobel s life and work can be found in NobelPrize org 2012 Research in Artificial Intelligence AI often generates expectation as it reveals new discoveries and inventions which show us realities which we were not able to imagine a few years ago However research in these areas is not controversy free On one side advancements in AI systems helps to build a more modern and comfortable society On the other these intelligent systems are perceived as machines which will eventually replace humans at work Regardless of whether these
106. lation D i Description The simulation shall provide the next services description route and navigation Cycle Second Priority Esential Stability Sabe Table 3 39 Requirement FR SI 06 Simulation services CHAPTER 3 PROJECT DESCRIPTION FRSIOT Description A service for a context description based on the descriptions of the network items which are closest to the person shall be available as an eventual service Cycle Second Priority Esnal Stability Sabe Table 3 40 Requirement FR SI 07 Description D me Name Routing eee DE ET service for describing a route between two points shall be available as an eventual service Cycle Second Priority Esential Stability Stable Table 3 41 Requirement FR SI 08 Routing D mn Name Navigation ees ot A S E Ts Abis KEE ASA i e navigation service shall be available which describes the steps for achieving a goal as the person moves through the network Cyde Second Priority Essential Stability Stable Table 3 42 Requirement FR SI 09 Navigation ESED Source and goal location Description The source location will be based on the person location while the goal location will be selected by the user in the network Cycle Second Priority Essential Stability Stable Table 3 43 Requirement FR SI 10 Source and goal location FR FR S o 11 Minimum cost A a a a a aeaee Descripti
107. ld of NLP was achieved in the late 1980s when machine learning algorithms started to be applied for language processing Previously most NLP systems were based on complex rules which were manually coded and the application of machine learning and statistical models instead of pure linguistical ones produced more reliable systems In the last years the field has been growing rapidly mostly due to the increase of the computing power and the availability of large amounts of data in the Internet Liddy 2003 Some of the tasks of NLP includes natural language recognition understanding and gener ation Optical Character Recognition OCR and speech recognition machine translation discourse analysis sentiment analysis question answering machine translation and many others 2 5 3 Natural User Interfaces In 1980 Richard Bolt described a computer system in which the user could place some shapes in certain places of a large display by using both voice commands and pointing gestures Bolt 1980 Despite the simplicity of this system it pointed out a new way of human computer interaction which did not require the use of computer input devices such as a keyboard or a mouse This system was one of the earliest ones to include what was later known as a Natural User Interface NUI A NUI can be defined as a computer interface in which the interaction is performed without requiring specific input devices These kind of interfaces follow the class
108. llows to choose one situational aspect 2 The situations list enumerates all the situations for the selected situational aspect in 1 When a situation is selected its description is shown just under the list 3 The situation removal button deletes the situation selected in 2 from the system 4 The new situation panel allows to add a new situation with the specified name and description In order to make the insertion effective the button must be pressed APPENDIX E USER MANUAL 131 Network Edition Simulation New Feature a se O Cost Factors O Situation Cost Factor ee O This situation represents a rainy environment o Remove Figure E 2 Screenshot for the situations management screen The new situation will be added in the situational aspect selected in 1 and shown in the list 2 5 The features list shows all the features stored in the system 6 The remove feature button deletes the feature selected in 2 from the system 7 The add feature panel allows the insertion of a new feature with the specified name in the system The new feature will be immediately shown in the list 5 8 The cost factors table shows the cost factors for the feature selected in 5 The first column contains the situation
109. me services such as the navigation service to give the most suitable response for example the orientation must be known to tell the agent whether he has to turn to a different direction in order to achieve his goal 86 Circumstantial Knowledge Management for Human Like Interaction 5 4 3 2 ShortestPath The ShortestPath class is used by the routing and navigation services to calculate the shortest path between the user location and its desired goal The method getShortest Path receives a network and source and target nodes Moreover a class field contains a set of applying situations which is used to calculate the cost factors for each network item Given that the network is represented as a directed weighted graph we can study different algorithms to find the shortest path between two nodes A first approach is found in the classic Dijkstra s algorithm Dijkstra 1959 This al gorithm is complete it finds a solution and admissible it provides the optimal solution Thus it may be a good solution for our problem However it can be really inefficient for big networks for which further approaches should be observed The use of heuristic search techniques can improve the performance as the search is no longer blind and instead is guided by a heuristic function For instance the A A Star algorithm could be used Hart et al 1968 This algorithm is complete and it is also admissible as long as the heuristic function is admis
110. n Figure 5 7 shows the classes for the presentation tier The attributes for the graphical components of these classes which are automatically generated by the NetBeans GUI builder are not shown in the figure UL CognosS UL Edition UL Simulation UL SituationManagement UI_NetworkManagement UL Simulation selectedNetwork VisualNetwork selectedNetwork VisualNetwork selectedPlane Plane selectedPlane Plane pickedNode VisualNode userNode UserNode pickedLink VisualLink motionSpeed int simulationSituations Set lt Situation gt UserInterface Figure 5 7 Classes for the presentation tier The description of the required graphical components was already included in section 4 2 3 The purpose of this section is to detail the logical components of the different classes as well as to introduce the external libraries used 5 5 1 UI_CognosS Subsystem The UI_CognosS subsystem only contains the UserInterface class which shows the initial screen with the database connection form and the about panel This class does not require any attribute apart from all the graphical components 5 5 2 UI_Edition Subsystem The UI_Edition subsystem contains two classes one class SituationManagement con tains the screen for situations management while the other one NetworkManagement contains the screen for network edition The SituationManagement class requires some attributes which will store the currently
111. ncial support for the research and development of a situation model and its associated management tools However in the short term the situation model developed for this project could be integrated with any other application in order to increase its funcionality As smartphones and other intelligent handheld devices are rapidly gaining in importance a situation model could provide a new way of interaction from which the user could take an advantage as now the interaction would depend not only on what the user asks but also on what the context is 3 4 2 3 Legal Constraints While the management and simulation application neither stores nor transmits confidential information about the users so far future applications which work over the situation model will communicate sensible information such as the user profile and his location over time Many countries require that the transmission of this information is performed in such a way that third parties cannot gain access to this information for example by encrypting the transmitted information In Spain this requirement is specified in the article 104 of the RD 1720 2007 BOE 2008 Therefore security requirements must be taken into account for further developments 3 5 Validation Testing The validation process aims to check that the development of the system results in a product that meets the requirements specified by the user which are defined in section 3 2 For this reason validati
112. nd situations are the only entities which do not require any other entity to exist A network is composed by network items which could be either nodes or links between two nodes It may be also noticed that the mechanism to assign costs to the network items is redundant as these costs can be directly assigned to the network items or assigned by means of items features 70 Circumstantial Knowledge Management for Human Like Interaction ModelKnowledge oo oo E Ss E el 335 a 8 Situation 2 E 3 Network 4 description 2 Feature Figure 4 6 ER model for the model knowledge 4 3 2 Edition Knowledge Base Figure 4 7 shows an ER model which serves as a conceptual representation of the data which is required by the system to operate The diagram distinguishes between the model knowl egde and the edition knowledge entities Note that the data store of the edition knowledge requires both of them while the situation model knowledge can be stored and operated independently the edition knowledge requires the existence of the model knowledge entities Then the edition database is actually a superset of the model knowledge base Keeping the same structure in the two data stores provides some advantages In the first place extracting the knowledge from the edition database is a trivial task as no translation between different architectures is required Sec
113. nexperienced users or for defining a network quickly 8 Please express here any other observation Figure C 4 Survey completed by subject 4 APPENDIX C EVALUATION DATA 125 1 Evaluate the next statements in a scale from 1 to 5 being 1 strongly disagree and 5 strongly agree for each of the two systems Question System A System B I felt good when performing the task the system is comfortable 1X 211 3L 4L 5 1L 21 3L 40X 5l I found the functionality which I required easily the system is intuitive 11 2K 30 4 5 11 21 304 41 5 DS a the task successfully and without errors the system is 1020 3 40 5 10 20 38 40 a i ea the task did not take too much time to be completed the system is 1028 3040 5 1 20 30 4 sl gile 2 What do you think are the advantages of System A I do not find advantanges in this system 3 What do you think are the disadvantages of System A Ithink as more important the inprecision and lack of scalability 4 What do you think are the advantages of System B More intuitive usable and faster 5 What do you think are the disadvantages of System B Some elements are confuse The reference points triangle and circles are treated the similar way to move The zoomis different for points and the image or plane I think the zoom should be integrated for both 6 In what situations do you think System A would be conve
114. ng requirements elicitation and specification problem analysis design of a solution system implementation and validation In more specific terms an spiral lifecyle composed of three phases was followed which is a convenient approach in research projects as all the system requirements might not be known from the very beginning After the implementation was completed an evaluation was carried out for observing the advantages of the edition tool over the manual edition of the model knowledge The results of this evaluation showed that the tool provides a mechanism for the model management which is significantly faster and more accurate than the manual edition Moreover a subjective survey also revealed that the experiment subjects preferred the edition tool as they considered it to be more comfortable more intuitive more reliable and more agile The resulting general purpose situation model and the edition tool are a significant contribution to the state of the art as the previously existing situation models were ad hoc models i e models implemented for supporting an specific system and their knowledge bases were edited manually Finally this work will be applied in a research project funded by the Spanish Ministry of Industry CADOOH TSI 020302 2011 21 and for this reason some of the future works observed in this document will be executed in the coming months A copy of the Cognos toolkit including the edition tool developed within th
115. nient I dont know 7 In what situations do you think System B would be convenient I think the system B is convenient for the purpose that the tooal has been created 8 Please express here any other observation Isuggest changing the zoom Figure C 5 Survey completed by subject 5 1 Evaluate the next statements in a scale from 1 to 5 being 1 strongly disagree and 5 strongly agree for each of the two systems Question System A System B I felt good when performing the task the system is comfortable iL 2X 3L 4L 5 11 21 3L 4 5X I found the functionality which 1 required easily the system is intuitive 101201 30 41 5N 1O 201 30 401 sX DS aa the task successfully and without errors the system is 1020 3 40 5 10 20 30 40 5 i a the task did not take too much time to be completed the system is K 20 3040 5 1 20 30 40 5 gile 2 What do you think are the advantages of System A Low level edition close to tech often enables greater datail and process shortcuts Copy paste is also powerful and I can keep the generated code and run it as many times as desired 3 What do you think are the disadvantages of System A Easy mismatch hard doing 4 What do you think are the advantages of System B Automate powerful enables edition of larger scenarios enables minor granularities greater level of detail 5 What do you think are the disadvan
116. nowledge Management for Human Like Interaction Malaka and Zipf 2000 Malaka R and Zipf A 2000 Challenging IT Research in the Framework of a Tourist Information System In Information and Communication Tech nologies in Tourism 2000 pages 15 27 Springer McCarthy and Buvac 1997 McCarthy J and Buvac S 1997 Formalizing Context Expanded Notes In Working Papers of the AAAI Fall Symposium on Context in Knowledge Representation and Natural Language pages 99 136 American Association for Artificial Intelligence NobelPrize org 2012 NobelPrize org 2012 Alfred Nobel His Life and Work http www nobelprize org alfred_nobel biographical articles life work Accessed February 16th 2012 Oracle 2012 Oracle 2012 Oracle Spatial Java API Reference 10g Documentation for class JGeometry http docs oracle com cd B19306_01 appdev 102 b14373 oracle spatial geometry JGeometry html Accessed May 30th 2012 Oviatt and Cohen 2000 Oviatt S and Cohen P 2000 Multimodal Interfaces that Process what Comes Naturally Communications of the ACM 43 45 53 Porzel et al 2006 Porzel R Gurevych I and Malaka R 2006 In Context In tegrating Domain and Situation Specific Knowledge In SmartKom Foundations of Multimodal Dialogue Systems chapter 18 pages 269 284 Springer Potts 1987 Potts M 1987 Computer Industry Wary of Jobs Perot Alliance The Washington Post page H2 Rivero 2012 Rivero J 2
117. nteraction in which the user communicates with a computer as if he were interacting with other humans Additionally this way of interaction would also increase the accesibility of the system to disabled users 2 5 1 Implicit Interaction In the first place it is important to understand the concept of implicit interaction Schmidt 2000 When humans interact with themselves a lot of information is exchanged implicitly through the participants behavior gestures the context etc which supports complements or disambiguates the information explicitly exchanged Nowadays when a user interacts with a computer he mostly inputs some commands in an explicit form through a Command Line Interface CLI a Graphical User Interface GUI or through the speech On the other hand implicit HCI does not require from the user explicit in structions or commands instead the system understands an action carried out by the user as input Schmidt 2000 provides the next definition Implicit human computer interaction is an action performed by the user that is not primarily aimed to interact with a computerized system but which such a system understands as input To attain such approach the system must be able to infer what the user is seeking with his behavior and for this reason it may need some understanding of the context and how the user can behave in it thus requiring some mechanisms for both perception of the environment and interpretation
118. nts of the interaction Meanwhile the Emotional Model operates the emotional state the User Model manages the interlocutor features the Session Model stores the evolution of the interaction and the Self Model manages the system goals Finally the Situation Model which is the aim of this project manages the contextual information of the interaction All these models work over a multi agent architecture thus acting in an independent but collaborative way The advantages of this architecture are presented in Izquierdo 2011 The presented cognitive architecture is currently being applied in the Cognos project 16 Circumstantial Knowledge Management for Human Like Interaction Color Code I Static Information in r z 1 q ES Dynamic Information Discourse Interpretation d B Emotional Information s 9 Domain Knowledge Linguistic Intentional Mle Interface Structures Management N l G Generation LP 8 BG o ULL Prosodic 3D Character n component Interface Agents Interaction Agent Figure 1 1 Cognitive architecture for a natural interaction system Calle et al 2010 Calle et al 2011 which is carried out at Universidad Carlos III de Madrid by the Advanced Databases Group LaBDA group http labda inf uc3m es of the Computer Science Department The current work could eventually be embodied into the Cognos toolkit 1 4 Document Structure In the first place
119. o go a step closer to his desired goal This instruction tells the user whether he must either turn to a different direction go back or go straight given its current position and orientation Moreover the system provides the description of the next node to be visited if it is available in order to provide the user with helpful and reliable instructions Figure 5 6 shows a cartesian plane representing a possible distribution of the user the 88 Circumstantial Knowledge Management for Human Like Interaction source node which is shown translucent in the figure and the target node which in the figure is shown with an incoming link 1 2 3 4 5 Figure 5 6 Geometric state of the plane In order to get the next instruction the system must calculate the angle 0 which is shown in the figure Actually this angle can be calculated using the dot product of the two vectors Utarget 2 1 Duser 1 1 We can calculate the angle 0 using the dot product and the vector lengths Vtarget Vuser arccos 0 SS SS S Utarget j User In our sample 0 113 Moreover not only the angle is required but also the angle direction in order to know whether the user must turn left or right The angle direction can be calculated using the cross product between the two vectors CHAPTER 5 IMPLEMENTATION 89 i j k a le a S Vuser Utarget Vuser Vuser up ue 0 target target The result of the cross product will be a ve
120. o other environmental eventualities such as noise weather etc a political aspect with regard to the roles or status of each person tak ing part in the communication process and a sociocultural aspect observing the influence on the interaction of the specific social and cultural conditions underlying the interaction Once we have introduced what the circumstantial knowledge and its components are we can delve into the concept of situation model In a first approach it should be stated that a situation model formalizes stores and processes the circumstantial knowledge for the communication process The inclusion of a situation model in a human like interaction system Rivero et al 2007 provides some advantages e Supports more knowledge influence in the produced interaction thus allowing the system to adapt its operations to the current context and enhancing the naturality of the communication e Enables the use of circumstantial information for filtering other components knowl edge bases thus reducing ambiguity and resolving indexicals and consequently in creasing the efficiency and efficacy of the system e Provides mechanisms for situation triggering i e to perform a certain action when some spatio temporal conditions are fulfilled e Provides new circumstances related functionalities such as guidance to a certain sit uation from the current one or predictions on an object situation in the future However the situation m
121. ockup construction GPS and RFID as physical layer emulators will be used e A button for each service provided by the simulator Some services may require some special controls e A list of situations which can be selected to apply during the simulation e A text box to generate the services output Chapter 5 Implementation The aim of this chapter is to delve into the design of the system thus providing enough detail as required to serve as a guide for its implementation This implementation must satisfy the physical and functional needs listed in the previous chapter Section 5 1 describes some considerations regarding the system implementation which may be helpful to understand some implementation decisions of the later sections Moreover sections 5 2 and 5 3 detail the implementation of the physical architecture for the system and provide a low level design of the functional architecture respectively The contents of the latest one can be used directly for the system coding Finally section 5 6 introduces the implementation of the data storages of the system 5 1 Implementation Issues This section describes some implementation considerations which may be helpful to under stand some decisions carried out during system implementation The system implementation will follow an object oriented approach This turns out to be a good option as the system is structured into classes each of them having well defined responsibilit
122. ode if any is shown in white Clicking on this area will move the user node 3 to the clicked position Also the direction keys of the keyboard can be used in order to move the user location 3 The user node shows the position of the user over the map This node is rotated according to the user orientation eight different values are possible 136 10 11 12 13 14 15 16 Circumstantial Knowledge Management for Human Like Interaction The zooming slider allows to zoom in and zoom out the map area 2 The number under the slider is the zoom percentage over the image real size The motion keys are an alternative for the user 3 motion which substitutes the direction keys in the keyboard The speed spinner allows to increase or decrease the motion speed of the user 3 over the map area 2 The GPS emulator panel allows to input two coordinates X and Y and by click ing int the send button send the user 3 to the location in the map 2 corresponding to those coordinates The RFID emulator panel allows to input a node ID and by clicking in the send button move the user to the corresponding node The RFID show IDs button shows the ID over each node in the map area 3 This button can be pressed again to hide those IDs The description service button executes the description service as long as the button is pressed The routing service button executes the routing service as long as the button is p
123. odel is interesting even when not included within a human like interaction system At the very least it provides context awareness capabil ities to any system thus empowering it For instance we can imagine a navigation system which not only plans a route to the user destination but it does it taking into account ambiental factors time conditions and the user preferences thus enriching its functionality by means of a situation model 1 2 Project Objectives In order to achieve human like interaction the human behavior in the communication pro cess must be studied formalized and stored in a form that can be processed by a computer system This task includes the formalization and storage of knowledge regarding the cir cumstantial aspects of the interaction The purpose of this project is to ease the implementation of specific situation models for a human like interaction system by providing edition tools able of feeding its knowledge CHAPTER 1 INTRODUCTION 15 bases From the five different components of the circumstantial knowledge which we have previously described the scope of the project is restricted only to the material aspect i e the spatio temporal component of the contextual information However a partial support of the remaining aspects will be provided by means of user defined situations This situation model will be implemented over a spatio temporal database Cuadra et al 2008 as it is able to store the location and
124. of which is perceived The fact that implicit interaction does not require from the user to explicitly emit commands to the computer establishes the link between this concept and ambient intelligence which provides some response to user actions proactively and consequently with context aware systems 26 Circumstantial Knowledge Management for Human Like Interaction 2 5 2 Natural Language Processing The Natural Language Processing NLP is a field of computer science which groups a set of computational techniques for dealing with natural language with the main purpose of achieving human like language processing capabilities For this reason NLP is often considered as a discipline of Al and its problems are usually allocated as Al hard problems Moreover this field is also strongly influenced by other disciplines such as linguistics which focuses on formal models of language and cognitive psychology Liddy 2003 Starting in the late 1940s the research on NLP during the first years was mostly focused on machine translation In the beginnings machine translation was basically performed by direct word by word translation and reordering thus leading to very poor results In 1957 Chomsky introduced the idea of generative grammars Chomsky 1957 resulting in one of the earlier contributions from linguistics to NLP In those years some other areas of NLP emerge such as speech recognition and synthesis Another important progress in the fie
125. om the model knowledge with a visual node from the edition knowl edge e The Plane VisualNode relationship cannot be removed as there are not alternative ways of specifying which nodes belong to a given plane e The VisualNetwork Plane relationship could be removed if it could known which planes belong to a network given the nodes belonging to that plane Although it can be actually knownn it is required that the plane contains at least one node which will not always happen Thus the relationship must exist 72 Circumstantial Knowledge Management for Human Like Interaction e A similar reasoning can be applied to the Network VisualNetwork relationship Given that the model could theoretically lead to some inconsistencies the edition tool must provide mechanisms for controlling the redundance thus preventing these inconsisten cies and avoiding incoherence Moreover some additional mechanisms may be implemented in the database with the same purpose such as triggers or check constraints The Corpus entity which is emphasized in the edition knowledge data serves as a connection point with the existing Cognos architecture and serves for integration purposes 4 4 Interaction Design This section goes deeper by providing some mockups of the different screens of the GUI which describe a high level design of the interfaces that will be shown to the user and which will be used to operate the application Figure 4 8 shows a mockup
126. on For the routing and navigation services the path to the goal will have minimum cost Cycle Second Priority Esential Stability Sabe Table 3 44 Requirement FR SI 11 Minimum cost path 45 46 Circumstantial Knowledge Management for Human Like Interaction PRESETS Situations selection Description The system shall provide means to the user to select several user defined situations which will apply during the simulation Cycle Third Priority Essential Stability Stable Table 3 45 Requirement FR SI 12 Situations selection 3 2 2 Non Functional Requirements This subsection describes the user requirements which are not related directly to what the system shall do but instead how the system shall be that is describes system features rather than system functionality 3 2 2 1 Scalability Requirements This subsection describes the user requirements referring to how the application shall scale when the size of the knowledge increases ERA Name Big networks eee a Sao re situation model and the management and simulation tool must be able to support big networks Cyde Fist Priority Essential Stability Stable Table 3 46 Requirement NFR SC 01 Big networks 3 2 2 2 Safety Requirements This subsection describes the user requirements referring to how the application shall re strict access to some features only to authorized users D RRS Description The system shall ak
127. on is an important task for guaranteeing the quality of the software product and thus a validation plan containing complete and correct tests must be carefully designed CHAPTER 3 PROJECT DESCRIPTION 51 The aim of this section is to describe the validation testing plan which is composed by a set of validation tests each of them covering at least one but probably more than one user requirement It can also happen that a requirement is covered by more than one validation test In this case all the tests covering the requirements must be successful in order to consider that the requirement is conveniently validated Each validation test is shown in a table containing a unique identifier a name and a description The test is defined by means of preconditions that the system must fulfill before the test is executed steps that must be executed over the system and postconditions which must be fulfilled after the mentioned steps are carried out A validation test is considered to be successful if the postconditions hold when the test is over In other cases the validation test fails and the requirements it covers cannot be validated In this case some corrective actions are required so that the system can be validated and accepted by the user Finally a traceability matrix is included in table 3 61 which shows the correspondence between user requirements and validation tests this information is also shown in the tests tables As it can be ob
128. ondly the edition database would support the simulation over it Regarding the edition knowledge each network entity in the model has an associated entity in the edition database which serves to draw a graphical representation of that network item However the entity Plane which does not exist in the model knowledge is required in order to draw a multidimensional network in a two dimensional grid for the edition purposes VisualNode and VisualLink entities are used for drawing the network The reader may have noticed that a loop exists between the entities VisualNetwork Network Item Node VisualNode Plane VisualNetwork Although this could lead to eventual inconsistencies it can be easily proved that any of these relations can be actually removed As the model knowledge may exist independently the relationships within that model must be kept Regarding the other relations CHAPTER 4 ANALYSIS AND DESIGN 71 ModelKnowledge g 3 al 2 E E 2 3 s El 3 Network relates to 1 1 Feature a 1 1 Visual Network gt ys E Si Figure 4 7 ER model for the edition knowledge e The VisualNode Node relationship must exist as there are not any other way of relating a node fr
129. onsideration contextual information in order to adapt its behavior to the current context 21 23 25 Dialogue Model model of the natural interaction system that processes the interlocutor communicative acts and develops its own acts and conversational strategies in order to generate a discourse 10 15 113 human like interaction style of Human Computer Interaction which tries to resemble the interaction between humans 13 15 33 138 implicit interaction set of actions that are not primarily aimed to be part of an interac tion but which another system understands in order to generate a response 25 knowledge base database which serves for knowledge management 69 70 72 73 92 93 98 knowledge model model that stores some kind of formalized knowledge expressed in some knowledge representation language so that it can be interpreted by a computer 69 137 138 Circumstantial Knowledge Management for Human Like Interaction location aware system context aware system only dealing with the spatial aspect of the context 23 mockup high level design of user interfaces which shows the user what the software will look like before starting its implementation 72 73 75 multimodality fact of using several modalities for the communication such as speech gestures written language etc 26 33 natural interaction see human like interaction 10 25 27 28 33 35 36 105 107 112 113 137 138 Natural Language Processing field of A
130. ontext Awareness In Proceedings of the 1st international Symposium on Handheld and Ubiquitous Computing pages 304 307 Springer Dijkstra 1959 Dijkstra E W 1959 A Note on Two Problems in Connexion with Graphs Numerische Mathematik 1 269 271 Ducatel et al 2001 Ducatel K Bogdanowicz M Scapolo F Leijten J and Burgel man J C 2001 Scenarios for Ambient Intelligence in 2010 Technical report IST Advisory Group Eckerson 1995 Eckerson W W 1995 Three Tier Client Server Architecture Achiev ing Scalability Performance and Efficiency in Client Server Applications Open Infor mation Systems 10 1 BIBLIOGRAPHY 143 Eisaku and Yasuhiru 2006 Eisaku M and Yasuhiru M 2006 Steps Towards Ambient Intelligence NIT Technical Review pages 50 55 Ferguson and Allen 1998 Ferguson G and Allen J F 1998 TRIPS An Integrated Intelligent Problem Solving Assistant In Proceedings of the 15th National Conference on Artificial Intelligence pages 567 572 AAAI Press Gee 1999 Gee J P 1999 Introduction to Discourse Analysis Routledge Hart et al 1968 Hart P Nilsson N and Raphael B 1968 A Formal Basis for the Heuristic Determination of Minimum Cost Paths IEEE Transactions on Systems Science and Cybernetics 4 2 100 107 Henricksen and Indulska 2005 Henricksen K and Indulska J 2005 Developing Context Aware Pervasive Computing Applications Models and Appro
131. or FRANCISCO JAVIER CALLE G MEZ THE EXAMINING BOARD President RICARDO COLOMO PALACIOS Secretary RA L ARRABALES MORENO Member MANUEL CARRETERO CERRAJERO After the defense of the Bachelor Thesis taking place in Escuela Polit cnica Superior of Universidad Carlos 111 de Madrid Legan s on July 3rd 2012 the examining board agrees to confer the next GRADE This page has been intentionally left blank It s difficult to be rigorous about whether a machine re ally knows thinks etc because we re hard put to define these things We understand human mental pro cesses only slightly better than a fish understands swim ming John McCarthy 1927 2011 We can only see a short distance ahead but we can see plenty there that needs to be done Alan Turing 1912 1954 This page has been intentionally left blank Agradecimientos Durante la realizaci n de un proyecto como ste uno se siente constantemente apoyado El inter s y la colaboraci n de las personas que me rodean son sin lugar a dudas un incentivo para la realizaci n del trabajo y es ahora mi intenci n devolver una peque a parte de aquellos en forma de agradecimiento En primer lugar debo agradecer a mis padres su inter s permanente en mi trabajo a pesar de que por cuestiones t cnicas y de idioma s que ser n incapaces de entender una sola palabra del mismo A veces la pregunta recurrente de cu ntas p ginas hab a escrito ese d a me hac
132. or network simulation 1 The system shows a list with all the existing networks The user may perform three different actions a create a new network identified by a name and a description b remove an existing network or c select an existing network The system performs the user operation Post Conditions Either the new network is created the removed network no longer exists in the system or the system loads the selected net work FR 02 FR MA 01 FR MA 02 FR MA 03 FR SI 01 Table 3 52 Validation test 04 Network selection Test Plane selection The system shows a screen to select a network plane Pre Conditions e The user has authenticated in the application e The user has selected a network e The user selects the screen for network management or network simulation 1 The system shows a list of the existing planes for the selected network 2 The user may perform three different actions a create a new plane identified by a name b remove an existing plane or c load an existing plane 3 The system performs the user operation Post Conditions e Either the new plane is created the removed plane no longer exists in the network or the system loads the selected plane FR 02 FR MA 13 FR MA 14 FR MA 15 FR SI 02 Table 3 53 Validation test 05 Plane selection 54 Circumstantial Knowledge Management for Human Like Interaction Te Plane visualization The system shows a network pl
133. ork by triangulation although other ones such as accelerometers estimate the changes on the location Table 7 21 Future work FW 21 Physical layer FW Full evaluation of the situation model Description A complete evaluation of the situation model could be performed af ter it has been integrated within the natural interaction system This evaluation would test the system with and without the support of the situation model and will observe the actual advantages that this model is providing to the interaction Table 7 22 Future work FW 22 Full evaluation of the situation model This page has been intentionally left blank Appendix A Scenarios This appendix will illustrate some scenarios where the situation model can be used to assist the user or to improve the user experience with other application A 1 Improving a Navigation System This scenario shows how a pedestrian navigation system which usually only considers the spatial aspect of the situation can be enhanced where a broader view of the context is taken into account For this case we will consider that the system also considers the temporal and ambiental conditions of the material aspect of the situation Figure A 1 illustrates this scenario Martha M wants to go from the Torres Quevedo building to the Sabatini building in the campus of Legan s of Universidad Carlos III It is 9 00 PM and there is a heavy rain Usually Martha would take th
134. orrow simply say What will the weather be like tomorrow Or Does it look like rain tomorrow Or even Will I need an umbrella tomorrow No matter how you ask Siri will tell you the forecast More information about Siri can be found in the next website http www apple com iphone features siri html 2 8 Conclusions The concept of ubiquitous computing presents a future where computers not only are lo cated everywhere but also the interaction with them vanishes in the background allowing humans to focus on the topic of the communication instead of the communication itself When humans talk each other in a daily context they can put all their efforts on what they are saying not paying attention to how they are moving their lips or how they are gesturing Unfortunately current computer interfaces are not really contributing for the achievement of this ubiquity as people still have to focus on whether they pressed the correct key or how they are moving the mouse Natural User Interfaces NUIs are a step forward in achieving that goal as far as they provide a way of interacting with the system which resembles human human interaction Two main features of these interfaces are multimodality and context awareness The first one involves using more than one way of interaction in the communication process whereas the second allows a more realistic dialogue which takes into account the context of the interaction The purpo
135. pe of the application Prior to the development the tasks which will be performed during each phase will be estimated The purpose of the first phase is to develop a prototype which provides basic network management features During the second phase a new prototype will be generated which will provide simulation features for the situation model Finally in the third cycle the last prototype will support the management of situations and the simulation over them This last cycle is also intended to perform a final refinement of the application by detecting and implementing features that were not specified in the previous phases 60 Circumstantial Knowledge Management for Human Like Interaction Cumulative cost 1 Determine objectives Progress 2 Identify and resolve risks Operational Prototype Review Prototype 2 Requirements Plan Prototype 1 Concept of Concept of operation requirements Requirements Development Verification amp plan f Validation Detailed Design Integration Verification amp Validation Implementation 4 Plan the Release next iteration 3 Development and Test Figure 3 2 Spiral lifecycle development model 3 7 Project Planning This section details the projected planning both in terms of schedule temporal planning and budget costs planning 3 7 1 Time Estimation Figure 3 3 shows a Gantt chart with the estimated planning
136. plane per network al though the situation model knowledge base actually has multiplane sup port Support for several planes is required to validate requirements FR MA 13 FR MA 14 and FR MA 15 and for this reason the future work is given the highest priority even if these requirements are not de fined as essential The support for several planes requires a new design of the user interface which allows to work with three dimensions rather than two in a comfortable and usable way Table 7 2 Future work FW 02 Planes support Description So far in order to edit several network items the edition must be per formed for each item one by one An alternative method for items selec tion could be implemented in order to be able to select several elements at once For instance the user could press the key Ctrl to select more than one item or even a selection window could be included where all the items within a region are selected Table 7 3 Future work FW 03 Multiple selection for network items FWO XML schema Description To increase the model portability and interoperability an XML schema could be specified The networks could be saved as XML documents thus the applications would not require a connection to a database to operate with the situation model In fact two schemas could be defined the first schema containing the model knowledge and the second one formalizing the edition knowledge Table 7 4 Future
137. poral or historical information The third dimension is required to store knowledge about the visibility of objects from a given location while the fourth dimension is required to store temporal evolution of such objects Regarding the interaction with the system it requires to be intuitive and usable and as it is intended to be used in a situation where the user may be walking or driving it also has to be unintrusive Therefore the interaction by means of natural language both for the input and the output is clearly adequate Actually Deep Map presents different interfaces such as a classical GUI 3D models natural language etc 2 7 3 SmartKom The importance of multimodality in computer interfaces and the advantages it provides for the interaction between humans and computers have been pointed out in previous sections The SmartKom system aims for an access to its functionalities which is efficient and intuitive for the user by providing symmetric multimodality Wahlster 2006 that is not only understanding a multimodal input but also generating its own multimodal output using the same modalities that those considered for the input SmartKom interaction management is based on models to represent and reason about a certain type of knowledge CHAPTER 2 STATE OF THE ART 31 such as user model task model domain model etc One of the major scientific goals of SmartKom was to understand imprecise or ambiguous input through the fussion of
138. re identified by their name and must have an associated background picture Cycle First Priority Optional Stability Unstable Table 3 19 Requirement FR MA 14 Network planes properties FR FR MA 15 eee 15 Network A selection Description The system shall provide means to the user to select a plane within the selected previously selected network Cycle Fist Priority Optional Stability Stable Table 3 20 Requirement FR MA 15 Network planes selection CHAPTER 3 PROJECT DESCRIPTION FRMATG Description The system shall provide means to the user to create edit and delete nodes within a previously selected network and plane Cyce First Priority Essential Stability Stable Table 3 21 Requirement FR MA 16 Nodes management FRMATE OOOO S EA Description The user shall E able to create nodes with a graphic interface E EE ising thm oe fhe a ago of he seated pl Cycle Firs Priority Essential Stability Stable Table 3 22 Requirement FR MA 17 Nodes creation Cycle Fist Priority Essential Stability Stable Table 3 23 Requirement FR MA 18 Nodes edition D FA Nodes properties Description Besides the properties common to all network items specified in requirements FR MA 08 to FR MA 13 a node may have an action radius which would be dependent on the localization technology Cycle First Priority Conditional Stability Unstable Table
139. rea 4 The node selection button allows the user to select a node from the map area 4 When the user selects a node its details are shown in the details panel 15 16 17 and 18 The link creation button will create a link between two nodes The user must click first in the source node and later in the target node in the map area 4 in order to create the link The link removal button will delete a link between two nodes In order to select the link to be removed the user must first click in the source node and later in the target node in the map area 4 The link selection button allows the user to select a link from the map area 4 by clicking first in the source node and later in the target node When the link is selected its details are shown in the details panel 15 17 and 18 The location details 16 are not shown for links The description details are shown when either a node or a link is selected and contains the item description 134 Circumstantial Knowledge Management for Human Like Interaction 15 The location details are shown when a node is selected and contains the coordinates of the node position 16 The cost details are shown when either a node or a link is selected and contains the basic cost for the network item 17 The features details are shown when either a node or a link is selected This element contains a table with features each of them with a checkbox that indicates whether the
140. ressed The select target button must be used along with the routing service 11 to select a target node in the map area 2 The service output textbox shows the output of the service in execution The situational aspect selector allows to choose a situational aspect When one is selected the situations within that aspect are shown in the situations chooser 16 The situations chooser shows the situations corresponding to the situational aspect selected in 15 A checkbox is available for each situation so that checking or unchecking it determines whether that situation applies or not for the simulation The status bar shows the status of the last operation performed or a message in red if an error occurred Glossary Al hard the set of the most difficult AI problems whose complexity is equivalent to making computer as intelligent as people 26 ambient intelligence term to refer to the idea that a computer system is incorporated to an environment in order to sense it and act proactively to provide some benefits to the users 21 22 25 circumstance piece of contextual information 13 circumstantial knowledge knowledge regarding the context 13 context set of circumstances applying in an environment which surround and condition an event 23 context model computational model that aims to formalize and store the context of a specific system in order to ease its processing 23 31 context aware system system that takes into c
141. retation Prosodic Analysis Intention Analysis Discourse Modeling Modeling System Lexicon Management Shared Knowledge Services a 1 sa Action Planning Character Display Animation nn Presentation Planning Audio Speech Language Output Synthesis A Generation Service B Figure 2 2 SmartKom architecture or just for entertaining Siri takes advantage of the iPhone localization mechanisms to know the user location and online information can provide much more details about the material aspect of the context in that place such as weather or traffic information This situational knowledge enables Siri to provide some context aware functionalities such as looking for places near your location and starting a navigation to that place creating new reminders or alarms at a given time or asking for the weather at a certain place The assistant may refer to additional information in order to increase the naturality of the interaction for example the user can ask Siri to create a reminder when he leaves work which requires previous CHAPTER 2 STATE OF THE ART 33 knowledge on the user schedule A reference to Siri as a human like interaction system is included in the Siri FAQs Apple com 2012 in the Apple website You can speak to Siri as you would to a person in a natural voice with con versational tone If you want to know what the weather will be like tom
142. rforming instead of focusing on the way he interacts with the system Anecdotically this term was first used by Steve Jobs in 1987 with a different meaning that the one that was later assigned to it he referred to the Apple II computer as an ubiq uitous computing resource that could be found everywhere on campus due to its reliability and low cost Potts 1987 One year later in 1988 Mark Weiser envisioned the concept of ubiquitous computing that has reached our days however it was not until 1991 that an article about it was published Weiser 1991 The main idea beyond ubiquitous computing as described in this article is to conceive a new way of thinking about computers that focuses in the human world and vanishes computers into the background that is the computer technology is so spread out that people use it without thinking about the technology itself but focusing in their own goals In order to achieve a real ubiquitous computing Weiser identifies two key issues loca tion and scale Weiser 1991 Regarding the first one a computer that knows where it is 19 20 Circumstantial Knowledge Management for Human Like Interaction located can adapt its behavior to the circumstances in significant ways Meanwhile the issue of scale is related to the fact that computers would have different sizes in order to suit a particular task Delving into the issue of scale Weiser proposed three different types of components for ubiquito
143. rvice or c select the navigation service The system executes the requested service Post Conditions Either the system describes the user location description ser vice or starts the routing or navigation service FR 03 FR SI 06 FR SI 07 FR SI 08 FR SI 09 Table 3 57 Validation test 09 Simulation services 56 Circumstantial Knowledge Management for Human Like Interaction ID Test 10 Name Routing and navigation services Description The system shows a screen to simulate a situation model service Pre Conditions e The user has authenticated in the application The user has selected a network and a network plane The user selects the screen for network simulation The user has placed the agent over the network The user has requested the routing or navigation services The user may perform the next operations a pick the tar get location destination for the routing or b choose applying situations for the simulation The system carries out the requested operation Post Conditions Either the system fixes the selected goal or updates the applying situations The system shows the most adequate route from the source to the goal The system produces the output for the service the route for the routing service and the next step for the navigation Covered Regs FR 03 FR SI 06 FR SI 08 FR SI 09 FR SI 10 FR SI 11 FR SI 12 Table 3 58 Validation test 10 Routing and navigation services I
144. s Entre mis compa eros y amigos de la Universidad me gustar a dar las gracias a aquellos que de un modo u otro han colaborado conmigo en la realizaci n de este trabajo a Jos Luis por sus visitas tan inspiradoras al laboratorio a Carlos por su constante inter s en el trabajo y por compartir conmigo sus fant sticas ideas a Sara por preguntarme por vil viii AGRADECIMIENTOS mi trabajo constantemente a Adela por soportar mis explicaciones entretenidas espero sobre mi trabajo a Josito por compartir conmigo su estr s diario con el TFG a pesar de que finalmente lo has acabado antes que yo y a Irene por su util sima contribuci n como soci loga Y por ltimo agradezco de un modo especial a Nerea por la excelente labor que ha realizado report ndome erratas en el trabajo y compartiendo su punto de vista Pero por encima de todo por haber sido capaz de aguantar estoicamente durante estos a os mi fingida prepotencia Gracias Abstract This project focuses on the circumstantial aspects of human like interaction Its purpose is in the first place to design and develop a general purpose situation model which would enhance any natural interaction system by providing knowledge on the interaction context and secondly to implement an intuitive management tool to edit the knowledge base of this situation model The development of both the model and the edition tool has followed the usual processes of software engineeri
145. s are hill climbing which takes the most promising node in each step but considering only ony path which most of the times will lead to a local optimum and beam search which behaves like hill climbing but considering more than one path at a time In fact an hybrid solution may provide a good behavior for our system While Dijkstra or A will provide good solutions for small networks greedy algorithms may be used for bigger networks In this case a greedy algorithm can be used and when the user is closer CHAPTER 5 IMPLEMENTATION 87 to the target node it is substituted by an admissible algorithm Finally some non conventional techniques shall also be taken into account such as Ant Colony Optimization ACO techniques Some advances in ACO are promising re garding hugh dynamic networks Rivero 2012 such as the ones that may be used within this project where the situations may change dynamically the costs of the different paths thus providing a convenient and efficient way to constantly update the route to adapt the new situations For this project Dijkstra s algorithm will be used for calculating the optimal path Although more interesting approaches in terms of efficiency have been observed Dijk stra is the only one which guarantees optimality given that no heuristic functions can be implemented It must be noticed that optimality was required in requirement FR SI 11 minimum cost path which is described in table 3 44 How
146. s currently us ing the Dijkstra algorithm However better approaches were pro posed in section 5 4 3 such as heuristic search algorithms A and IDA greedy algorithms hill climbing or beam search or even Ant Colony Optimization ACO techniques Different alternatives could be implemented in the mid term in order to evaluate their response time and their optimality Table 7 10 Future work FW 10 Efficient algorithms for shortest path CHAPTER 7 CONCLUSIONS AND FUTURE WORK 111 FWT Description In order to analyze the reasoning processes of the situation model a trace tool could be implemented Table 7 11 Future work FW 11 Trace tool 7 2 3 Cognos Toolkit The tool is now isolated as it is the result of an independent development project but it will be integrated into a major toolkit for the complete edition of interaction knowledge Cognos Calle et al 2010 Calle et al 2011 in the short term Future works proposed in this section have the purpose of integrating the situation model tool in the Cognos toolkit thus enriching the other models W Inlaying in Cognos toolkit Description The edition tool will be incorporated to the Cognos toolkit by making the application interface consistent with the already existing tools Table 7 12 Future work FW 12 Inlaying in Cognos toolkit Description The edition knowledge will be integrated in the Cognos knowledge base When this integration is complete
147. se of this project as it was stated in section 1 2 is to develop a tool to manage circumstantial knowledge thus providing context awareness capabilites to a natural interaction system This page has been intentionally left blank Chapter 3 Project Description The main purpose of this chapter is to describe the project by obtaining and specifying the requirements for the software product sections 3 1 and 3 2 which will provide enough information to be able of generating an analysis and a detailed system design in a further stage see chapter 4 which satisfies those requirements Sections 3 3 and 3 4 include a review of the previous existing work and a feasibility study respectively to conclude whether the system can be developed and which environmental constraints must be satisfied for this to happen Section 3 5 defines a set of tests to be performed after the software is developed to vali date the software against the specified requirements Section 3 6 describes the development methodology and lifecycle which will be followed for the current project Finally section 3 7 documents the project planning both in terms of schedule and costs 3 1 Requirements Elicitation In order to perform a requirements specification a previous work on requirements elicitation must be carried out The aim of this process is that the requirements engineer gathers from the users or customers the requirements of a system For this project the tutor
148. selected network and plane as well as the selected node or link if any It is required to store these items in order to enable their edition CHAPTER 5 IMPLEMENTATION 91 5 5 3 UI_Simulation Subsystem The UI_Simulation subsystem contains the UI_Simulation class which contains the screen for performing a simulation over a network This class requires apart from the attributes storing the graphical components some additional attributes to work properly First the selected network and plane must be stored as it contains the set of nodes and links for the simulation The user node must also be stored and its purpose is to keep the agent position and orientation over the network The motion speed attribute is included to determine the speed by which the agent moves across the map although it is not required Finally the set of selected simulations is stored and is passed to the service providers when a service is requested 5 5 4 External Libraries For dealing with some advanced features of the graphical interface a couple of third party libraries are used in this project These libraries are briefed in this section 5 5 4 1 JXLayer The JXLayer library provides some useful functionalities for graphical interfaces For ex ample the JX Layer class acts a JPanel which can be zoomed or rotated This project takes advantage of this class to be able to provide a zooming feature to change the size of the loaded plane Downloads for this libr
149. served in the figure the traceability matrix is complete thus all the requirements are covered by at least one validation test and the validation plan is correct The validation tests will be executed after the system implementation is completed and the results will be analyzed in chapter 6 TOT Database connection Description The system shows a screen for the user to establish a database connection with a username and a password Pre Conditions e The user has started the application Steps 1 The system shows the database authentication screen 2 The user inputs the database connection information and clicks on Connect 3 The system validates the information and if it is correct it es tablishes the connection Post Conditions e A database connection is established e The interface is enabled for the user to use the management and simulation tools FROL NFR SAOL Table 3 49 Validation test 01 Database connection 52 Circumstantial Knowledge Management for Human Like Interaction Situations management The system shows a screen to manage add or remove situations for some fixed situational aspects Pre Conditions e The user has authenticated in the application Ce AAA 1 Steps The system shows a list with the different situational aspects The user selects a situational aspect for the list The system shows a list with the situations for the selected sit uational aspect The user may perform t
150. sible as well for an heuristic function to be admissible it must never overestimate the real cost Although A would lead to a performance gain over Dijkstra both in the temporal and spatial aspects of the computational complexity in order to obtain the shortest path we need to design an admissible heuristic and that will not be always possible A good approach for an admissible heuristic would be the euclidean distance between the source and the target nodes However although this may be admissible most of the times it could overestimate the real cost if the user have edited the links costs so that they are smaller than the euclidean distance or if there are cost factors which are smaller than 1 Despite that A would improve the performance it would still require exponential space to operate On the other hand there are some alternatives which are heuristic guided and can also provide an optimal solution if the heuristic is admissible such as IDA Iterative Deepening A Korf 1985 This algorithm would require more expansions than A and therefore may take slightly more time to find a solution but it will require much less memory While the previous algorithms are intended to calculate the optimal solution it may not be required Given that time is an important constraint in the case of big networks a greedy algorithm may be used which may not provide an optimal solution but a suboptimal one instead An example of these algorithm
151. sible to calculate the shortest path between two nodes These semantics can be implemented by means of a table constraint Moreover the network items numeric identifier will be autoincremental This mechanism can be implemented through a sequence and a trigger 5 6 2 Edition Knowledge Base Figure 5 9 shows the relational graph for the edition knowledge base This relational graph duplicates the structure of the model knowledge and includes the new relations for sup porting the edition operations The need for the replication of the model knowledge was justified in section 4 3 situationalAspects DNA UNA DC UNA situations name aspect description features costsFeatures situation feature costFactors DC UNA costs situation item costFactors DC UNA itemsFeatures networks name description DC UNA id radius parentNode DC UNA DC UNA DC UNA DC UNA DC UNA network name backgroundImage DC UNA Figure 5 9 Relational graph for the edition knowledge The redundancy which was mentioned in section 4 3 still remains in the database im plementation For this reason we will implement also a trigger which will prevent possible inconsistencies in the data This trigger will ensure that when a visual node is inserted in the database it belongs to the same network that its associated real node Moreover the application will be developed in order to prevent this inconsistency This page
152. system Figure 5 3 displays the class diagram for the classes within the DataGateway subsystem This section describes each of these classes in higher detail DataGateway FeatureDataGateway SituationDataGateway CostDataGateway NetworkDataGateway insertFeature insertSituation getCostsByNetworkItem Set lt Cost gt insertNetwork removeFeature removeSituation removeNetwork updateFeature getSituations Situation I getAllNetworks Set lt Network gt getFeature Set lt Cost gt getSituationsByAspect Situatio getNetworkByName Network l getCostFactors Situation getSituationByName Situation DBConnection VisualNetworkDataGateway host string port string C instance string insertVisualNetwork user string getLinksByNetwork Set lt Link gt getVisualNetworkByName VisualNetwork password string getLinkFeatures Set lt Feature gt connection Connection connect VisualNodeDataGateway disconnect TOTO fscomestt poo Finsen VisualNode0 X executeQuery ResultSet updateVisualNode executeSQL getVisualNodesByNetwork Set lt VisualNode gt commit getNodesByNetwork Set lt Node gt trollback getNodeFeatures Set lt Feature gt VisualLinkDataGateway insertVisualLink update VisualLink getVisualLinksByNetwork Set lt VisualLink gt getPlanesByNetwork Set lt Plane gt Figure 5 3 Classes
153. tages of System B Copy paste of nodes alignment of nodes easy by manual edition vertical horizontal distribution eas y by manual edition 6 In what situations do you think System A would be convenient small scale problems with regular grids 7 In what situations do you think System B would be convenient any problem of this kind 8 Please express here any other observation please support the author and also his tutor Bank notes are accepted preferable the purple ones Figure C 6 Survey completed by subject 6 This page has been intentionally left blank Appendix D Installation Guide The management and simulation tool is provided as a jar a Java archive which can be directly executed in a system with the Java Virtual Machine JVM installed Moreover as the database is located in a server within the network of Universidad Carlos III a connection to the university Virtual Private Network VPN must be established as well This chapter explains the steps in order to install the JVM establish the VPN connec tion and start the application D 1 Installing the JVM The Java environment can be downloaded from the next site http www java com Follow the instructions in order to download the installation files The website will detect automatically your operating system and will provide the last version of the JVM Once the installation package is downloaded load it and follow the installat
154. te networks Cyce Fist Priority Eso al Stability Sabe Table 3 6 Requirement FR MA 01 Networks management PIMA Networks selection Description The system shall provide means to the user to select a network in order to work with it Cycle Fist Priority Essential Stability Stable Table 3 7 Requirement FR MA 02 Network selection D FRA Networks properties Description Networks are identified by a name and can have an optional de scription Cycle Fist Priority Esnal Stability Stable Table 3 8 Requirement FR MA 03 Networks properties mD PRA Networks may oa nodes and links between nodes Osee Fist Prionty seal Stability Stable Table 3 9 Requirement FR MA 04 Network items CHAPTER 3 PROJECT DESCRIPTION FR MAUs Network items properties Description Networks items are identified by a numeric identifier and may have an optional description Cyce First Priority Essential Stability Stable Table 3 10 Requirement FR MA 05 Network items properties Cycle Fist Priority Essential Stability Sabe Table 3 11 Requirement FR MA 06 Network items activeness maano S SA Name Network items hierarchy 2222 items oN sete _ ae The system shall support a future extension to structure the net work items in a hierarchy Cycle Fist Priority Conditional Stability Table 3 12 Requirement FR MA 07 Network items hierarchy
155. te is not contained neither in the x axis nor in the y axis i e 11 2 Y1 F Ya Zol Ty2 and Yu Yv2 the scale between the visual plane and the real network can be calculated for two dimensions as follows La T scale LyQ Ly Ya Y scale Yvu2 Yvi CHAPTER 5 IMPLEMENTATION 85 Given a node n for which we know its visual coordinates 2 and Yun but we have no information about its real coordinates 1 and yn these can be calculated given the next formula Ln 11 Ly1 Lun X scale Un Y1 Yo Yun x scaley Then we can take advantage of the two reference nodes to calculate the coordinates of any other node in the network 5 4 3 Simulation Subsystem The Simulation subsystem which contains the logic to run a simulation over the situation model is formed by the classes which are shown in figure 5 5 Simulation RoutingServiceProvider ShortestPath targetNode VisualNode situations Set lt Situation gt int DescriptionServiceProvider Po getShortestPath network Network start Node end Node Set lt Link gt Figure 5 5 Classes for the Simulation subsystem 5 4 3 1 UserNode The UserNode class is a subclass of VisualNode and represents the user or agent location in the network Apart from the fields of the VisualNode class it also includes a field to represent the agent orientation This orientation may be required by so
156. tem is comfortable reliable efficient and 100 Circumstantial Knowledge Management for Human Like Interaction intuitive The survey also asks for the advantages and disadvantages of each of these systems and allows the user to express other considerations Figure 6 2 shows the questionnaire that was distributed among the subjects and which they filled after completing the two tasks Cognos S Circumstantial Knowledge Management Survey Please take some minutes to fill this survey after performing the two tasks Survey ID 0 Name 1 Evaluate the next statements in a scale from 1 to 5 being 1 strongly disagree and 5 strongly agree for each of the two systems Question System A System B I felt good when performing the task the system is comfortable 11 211 31 4L 5 111211 31 41 5 I found the functionality which I required easily the system is intuitive 101 201 30 4U 5U 10 20O 3U 4H SE TE completed the task successfully and without errors the system is Brasa daa 10 20 30 40 50 in the task did not take too much time to be completed the system is 10 20 30 40 50 1020 30 40 50 gile 2 What do you think are the advantages of System A 3 What do you think are the disadvantages of System A 4 What do you think are the advantages of System B 5 What do you think are the disadvantages of System B 6 In what situations do you think System A would be con
157. that knowledge supporting the interaction has different natures Besides the purely interactive the system may require additional types of knowledge such as the operational which has to do with tasks the circumstantial the emotional knowledge about the user or even knowledge about its own status and goals The management of some of this knowledge may be highly complex and computationally expensive However removing some of this knowledge will reduce the naturality of the 28 Circumstantial Knowledge Management for Human Like Interaction system leading to a more mechanical behavior for the aspect regarding the discarded knowledge Calle et al 2008 2 6 Situation Models As stated in chapter 1 an important component of a natural interaction system has to do with the management of circumstantial knowledge i e the context of the interaction The term situation was used by Connolly 2001 to refer to the non linguistic context It has also been used in previous projects based in the cognitive architecture proposed in Calle 2004 Throughout this work term situation model will be used to refer to a context model which stores and processes the circumstantial knowledge A situation model turns out to be a key component in those cases where the context is not fixed for the interaction The advantages of the inclusion of a situation model in a natural interaction system have been listed in section 1 1 2 7 Applications The purpose of t
158. the simula tion model This function will install the database schema over an empty instance so that it can be used to create and manage new networks Table 7 17 Future work FW 17 Model installation FAS Integration with the Task Model Description The situation model could enrich the Task Model by providing new avail able tasks for it to carry out High Table 7 18 Future work FW 18 Integration with the Task Model CHAPTER 7 CONCLUSIONS AND FUTURE WORK 113 FW Integration with the Dialogue Model Description The situation model could be enriched by the Dialogue Model as the last one could provide interactive strategies supporting the tasks of the first one Table 7 19 Future work FW 19 Integration with the Dialogue Model Fa Full integration in the natural interaction system Description The situation model could enhance the performance of any other model in the natural interaction system as it can provide functionality to filter its knowledge using the circumstantial one Table 7 20 Future work FW 20 Full integration in the natural interaction system FW Physical layer Description So far the simulation over the situation model provides emulator for a physical layer Actually this physical layer could be implemented to provide the system with the user location Some interesting positioning technologies are GPS RFID GSM UMTS Wi Fi accelerometers or ultrasounds Most of these methods w
159. tion subsystem subsystem subsystem ModelKnowledge Simulation UL Simulation I a a a ae cc ad subsystem subsystem DataGateway UL_CognosS Figure 4 2 3 tier architecture design Throughout this section it is provided a more detailed description for each of the tiers and its components 4 2 1 Data Tier The data tier contains the resources which serve for persistent storage i e the mechanisms for data storage themselves for example a database management system as well as the structures which may be required for handling the data For this project a relational database management system will be used to satisfy the application persistent storage necessities The choice of a relational database over other possible alternative was motivated by the predominance of these systems and their facility for the design and implementation as well as for the ease when executing queries and updates over them The data tier will be accesible by a data controller in the logic tier which will be described in the next section As it is shown in figure 4 3 the data tier contains two subsystems the model knowledge subsystem and the edition knowledge subsystem 66 Circumstantial Knowledge Management for Human Like Interaction Data Tier subsystem subsystem EditionKnowledge ModelKnowledge Figure 4 3 Subsystems for the data tier 4 2 1 1 ModelKnowledge Subsystem The ModelKnowledge subsystem contains all
160. tous computing However the current situation still falls short of the original vision introduced by Weiser in which computers are invisible to humans who can concentrate on the tasks they are performing Moreover the idea behind ubiquitous computing introduces new challenges particularly the computer devices can be located everywhere and consequently some knowledge about the environment might support their adaptation to circumstances 2 2 Ambient Intelligence In 1999 another key term related to the concept of ubiquitous computing arises this term is ambient intelligence and appears for the first time in a paper written by the director of Philips about the digital home Jones 1999 Two years later in 2001 the European Comission designs a roadmap for research in ambient intelligence Ducatel et al 2001 in which are also described four scenarios on how ambient intelligence could be applied to everyday life in 2010 from the current point of view those scenarios turned out to be too ambitious The basic idea behind ambient intelligence is that by incorporating some technology to an environment a system can be built for executing actions that will provide some benefits to the users in that environment The system which would be transparent to the user would decide by itself which actions to take by sensing the environment and the users in it and processing this information Some features which are commonly expected in ambient intell
161. tworkItem class It contains two fields which are references to the start and end node of the link Additionally an extra field indicates whether the link is bidirected or not This class also contains a static method to find a link within a set given its numeric identifier The Feature class is identified by a name It contain a set of costs The Cost class relates a cost factor to an specified situation 5 3 2 EditionKnowledge Subsystem Figure 5 2 displays the class diagram for the classes in the EditionKnowledge subsystem This section describes each of these classes in higher detail e The VisualNetwork class serves as an edition wrapper for the Network class in the ModelKnowledge subsystem This class also contains a set of planes and a set of visual links e The Plane class represents a bidimensional entity which is loaded into the edition tool A plane its identified by its name and contains a background image so that visual nodes are placed over this image It must be realized that while nodes must belong to a plane links may or may not belong to a single plane as a link can join CHAPTER 5 IMPLEMENTATION 81 EditionKnowledge name string VisualLink backgroundImage Image link Link network VisualNetwork nodes Set lt VisualNode gt planes Set lt Plane gt startNode VisualNode links Set lt VisualLink gt r endNode VisualNode VisualNode x double y double node Node fin
162. uations for 80 Circumstantial Knowledge Management for Human Like Interaction each situational aspect This decision makes it possible to use an enumerated type for the situational aspect e The Network class is identified by a name and may also contain a description More over it contain a set of network items e The NetworkItem abstract class stores the common fields for both nodes and links These items are identified by a positive integer and may also contain a description All the network items contains a geometry which is a spatial type which may be useful for operating with these items within a spatial database management system The documentation for this type can be found in Oracle 2012 The network items also contain a basic cost which can be complemented with additional cost factors Moreover the items can also be assigned features which also serves as an alternative way of assigning cost factors Additionally a boolean field represents whether the item is active or it is not Finally an integer field stores the level in the network item hierarchy which may be used for future developments e The Node class is a specialization of the NetworkItem class It provides two additional fields to represent the node radius and the parent node in the hierarchy So far these fields may not be required This class contains a static method to find a node within a set given its identifier The Link class is a specialization of the Ne
163. ure E 4 Sample warning message for uncommited changes APPENDIX E USER MANUAL 135 E 4 Simulation This screen allows to carry out a simulation over a network from the situation model Figure E 5 shows this part of the application r Q Cognoss z Database Connection Situations Management Network Edition Simulation loja 1 Situations Network i an O SABATINI 2 Ambiental w Sabatini Floor 2 Motion Situational Aspect Qj as JA aia Use the mouse and the keyboardto place and move the agent over the map Tum slightly right and go to Software Engineering Lab 22810 em m a E o a Y E en e Lc y Disconnected Figure E 5 Screenshot for the simulation screen The components of this screen are described here 1 The network selection panel allows to select a network from the selector box When a network is chosen its description is shown under the box The button must be clicked in order to load the network for the simulation 2 The map area shows the plane for the network with all its items nodes and links The items which are picked such as the closest node to the agent or the nodes and links of a route are highlighted Moreover the target n
164. ure the validation of all the user requirements in the first place but also to include new functionalities in order to improve the user experience and to go a step further in the research in situation modeling 7 1 Conclusions When humans take part in an interaction the surrounding context is influencing on many aspects of how this interaction is done The context is determining our roles in the commu nication process what we can say and what we cannot and even what our gestures mean The influence of the context is so strong that most of the times we do not realize about its existence It has become a natural part of the interaction Unfortunately this naturality is not extensible to computers If we want a machine to be able of simulating a human in order to communicate with other humans we must first identify all the components which take part in the human human interaction which is definitely not a simple task Later these components must be modeled and its knowledge must be formalized Again this task is hard but it is also essential in order to achieve natural interaction The purpose of this project was to design and implement a simple situation model which formalizes and stores the knowledge of a subset of the material aspect of the context Specifically the model stores an spatial network which is formalized as a graph The graph formalization was chosen because it can be easily extended for observing the rest of the theoreti
165. us systems each of them with different sizes and purposes The smallest compo nents are called tabs which are sized like a post it or ID card Slightly bigger devices would be pads sized like a sheet of paper A4 they do not need to be carried out with the user but instead they will stand as scrap computers The last and biggest devices are called boards which can serve as video screens digital whiteboards or bulletin boards etc Weiser specifies three technologies which are required for supporting ubiquitous computing cheap computers with convenient displays software for ubiquitous applications and a net work to allow the communication among the different systems Different computing eras have been identified according to the number of computers per person Weiser and Brown 1996 Within this taxonomy ubiquitous computing is his torically placed as a new computing paradigm which follows the personal computing The different computing paradigms or eras can be summarized as follows e The mainframe era in which many people usually experts shared one computer mainly for scientific purposes e The Personal Computer PC era in which each person has its own com puter In 1984 the number of people using a PC surpassed those using shared computers IDC 1996 Since then the appeareance of laptop computers Personal Digital Assistants PDAs and mobile phones have contributed to the es tablishment of the PC era and the evolution to th
166. vant context modeling ap proaches based on the data structures for the representation and exchange of contextual information e Key Value models they are the simplest data structure for context modeling and store the value of contextual information identified by keys Although these models are easily managed they lack of a more sophisticated structure e Markup scheme models they model the contextual information in a hierarchical data structure consisting of markup tags with attributes and contents e Graphical models there are several approaches Unified Modeling Language UML diagrams are used to model contextual information The use of UML is appropiate for context modeling due to its generic structure Moreover it is applicable to derive an ER model which is useful to elaborate the structure of a relational database e Object oriented models object orientation provides many powerful features for context modeling such as encapsulation reusability inheritance etc Existing ap proaches use objects to represent context types and encapsulates the details The context processing logic is provided through interfaces e Logic based models these approaches use facts expressions and rules to define a context model An inference process derives new facts from the existing knowl edge composed of the existing facts and rules The contextual information must be represented formally as facts One of the earliest approaches was published b
167. venient 7 In what situations do you think System B would be convenient 8 Please express here any other observation Thank you for your time Figure 6 2 Questionnaire for the experiment As the tool is aimed at a very specific audience the population chosen for the evaluation is small it is formed by six persons For this reason a test of significance will not be performed over the experiment From the evaluation population one half are experts in CHAPTER 6 VALIDATION AND EVALUATION 101 the annotation of spatio temporal knowledge and the other half are familiarized with the annotation of knowledge in other areas The next academic profiles can be found in the subjects of the experiment one mathematician one linguist one pharmacist and three computer engineers The targeted subjects are in the age range from 25 years to 40 years Each person was given the instructions sheet a copy of the survey and a copy of the edition tool user manual During the experiment the times were measured for each of the tasks which were carried out in different orders by the experiment subjects The next sections detail the results of the evaluation and elaborate some conclusions on them Meanwhile appendix C contains the raw data gathered in the experiment 6 2 1 Performance Table 6 13 shows the average times and the standard deviation for running the modeling tasks using the direct model edition method and the tool As it can
168. wledge Base Operation o Qs Operation Client 1 Operation Client 2 Situation Model Q Server Operation Client 3 Operation Client 4 Figure 4 1 Physical client server architecture 4 2 Functional Architecture Once the physical architecture is defined the next step proposes a design for the functional architecture of the system The system will follow a three tier architecture Eckerson 1995 This model develops the user interface presentation tier the process business logic logic tier and the data data tier as independent modules However although functionality is thus divided into three tiers this functional vision does not affect the design of storages or the interaction This modularity provides several advantages as any of these modules can be updated and maintained independently from each other reducing the coupling between the different components For example this model would allow the use of different user interfaces designed for different computer devices using the same application logic and the same data Additionally decoupling the data from the rest of the application also improves the system performance and scalability CHAPTER 4 ANALYSIS AND DESIGN 65 Figure 4 2 provides an overview of the application design showing the subsystems for each of the three tiers Data Tier Logic Tier Presentation Tier subsystem subsystem subsystem EditionKnowledge Edition UL Edi
169. wo different management operations a add a new situation for the selected situational aspect given its name and description or b delete an existing situation for the selected situational aspect The system performs the user operation Post Conditions Either the new situation is added to the list of situations or the removed situation no longer exists in the system FR 02 FR 04 FR MA 24 FR MA 25 FR MA 26 Table 3 50 Validation test 02 Situations management Features management The system shows a screen to manage add or remove features Pre Conditions e The user has authenticated in the application Steps 1 The system shows a list with all the stored features 2 The user may perform three different operations a add a new feature b delete an existing feature or c edit the cost factors for an existing feature 3 The system performs the user operation Post Conditions e Either the new feature is added to the list of features the re moved feature no longer exists in the system or the cost factors for the selected feature are updated FR 02 FR 04 FR MA 10 FR MA 11 FR MA 28 Table 3 51 Validation test 03 Features management CHAPTER 3 PROJECT DESCRIPTION 53 Network selection Description The system shows a screen to select an existing network or create a new one Pre Conditions e The user has authenticated in the application The user selects the screen for network management
170. wo systems Question System A System B I felt good when performing the task the system is comfortable 11 2L 3X 4 sU 10 211 3L 4101 5X I found the functionality which I required easily the system is intuitive 11 21 3k 41 5 11 21 311 41 5x eee the task successfully and without errors the system is 10 20 38 40 50 10 20 30 4 50 I think the task did not take too much time to be completed the system is 101 20 30 40 10 20 301 401 SX agile 2 What do you think are the advantages of System A Repetition makes it a simple and mechanical system 3 What do you think are the disadvantages of System A All the data has to be manually introduced 4 What do you think are the advantages of System B It is intuitive very graphical and easy to be used 5 What do you think are the disadvantages of System B The screen is too large It took some time to find some controls 6 In what situations do you think System A would be convenient Don t know 7 In what situations do you think System B would be convenient Don t know 8 Please express here any other observation Paa o Figure C 2 Survey completed by subject 2 124 1 Evaluate the next statements in a scale from 1 to 5 being 1 strongly disagree and 5 strongly agree for each of the two systems Question System A System B I felt
171. work FW 04 XML schema CHAPTER 7 CONCLUSIONS AND FUTURE WORK 109 FAS Node creation by coordinates Description Even if the tool graphical capabilities ease the task of modeling the net work by improving the performance and the accuracy of this process in some cases it might be interesting to be able to add a new node directly by inputting its coordinates instead of placing it over the plane Table 7 5 Future work FW 05 Node creation by coordinates Description So far the only supported dynamic agent is the user who moves across the network The dynamism for other elements can be partially simu lated by using situations which may change the item status However native support for dynamic agents such as other people is proposed as a future work This agents would be able to change their status on every dimension of the network i e they would be able to show some kind of evolution Table 7 6 Future work FW 06 Support for dynamic agents FW Undo and redo Description The application could include a functionality to revert the last changes So far this operation is partially implemented by taking advantage of a transactional system which allows to commit and rollback the changes However a more sophisticated system could be implemented using a stack of historic operations Table 7 7 Future work FW 07 Undo and redo 7 2 2 Situation Model The future works proposed in this section have as main a
172. y 39 Requirement FR MA 08 Network items cost o 39 Requirement FR MA 09 Network items coordinates 39 Requirement FR MA 10 Features o oo 2000004 39 Requirement FR MA 11 Features properties 40 Requirement FR MA 12 Network items features 40 Requirement FR MA 13 Network planes o o 40 Requirement FR MA 14 Network planes properties 40 Requirement FR MA 15 Network planes selection 40 Requirement FR MA 16 Nodes management 41 7 3 22 3 23 3 24 3 25 3 26 alt 3 28 3 29 3 30 3 31 3 32 3 33 3 34 3 99 3 36 3 37 3 38 3 39 3 40 3 41 3 42 3 43 3 44 3 45 3 46 3 47 3 48 3 49 Circumstantial Knowledge Management for Human Like Interaction Requirement FR MA 17 Nodes creation ooo 41 Requirement FR MA 18 Nodes edition 41 Requirement FR MA 19 Nodes properties 41 Requirement FR MA 20 Links management 41 Requirement FR MA 21 Links creation 42 Requirement FR MA 22 Links properties 42 Requirement FR MA 23 Edition interface zoom 42 Requirement FR MA 24 Situational knowledge taxonomy 42 Requirement FR MA 25 Situations management 42 Requirement FR MA 26 Situations properties 43 Requirement FR MA 27 N
173. y McCarthy and Buvac 1997 e Ontology based models ontologies are a very promising instrument for context modeling as they can represent a description of concepts and relationships in a very expressive and intuitive way 2 5 Natural Interaction In 1950 Alan Turing wrote his article Computing Machinery and Intelligence Turing 1950 in which he described a variation of the imitation game later known as CHAPTER 2 STATE OF THE ART 25 the Turing test The main idea behind this test is that a human is able to decide whether he is having a conversation with a machine or with another human For a machine to win this game it must be able to interpret and understand the conversation and to generate ap propiate human like answers The Turing test has been strongly influential over many sub fields of Artificial Intelligence AI such as the Natural Language Processing NLP Half a century later the concept of natural interaction was born A very important aspect of ubiquitous computing and ambient intelligence has to do with interaction The progressive establishment of computers in the society requires new ways of interaction and as many people with no technological training nor experience in interacting with computers gain access to these devices simple and natural interfaces must be implemented The goal as it was previously stated is that computers turn out to be invisible for humans thus requiring the achievement of new forms of i
174. you have an informatic system 8 Please express here any other observation Figure C 3 Survey completed by subject 3 1 Evaluate the next statements in a scale from 1 to 5 being 1 strongly disagree and 5 strongly agree for each of the two systems Question System A System B I felt good when performing the task the system is comfortable 11 2L 30 4 amp 1 SI iL 201 301 4 5 I found the functionality which I required easily the system is intuitive 101 201 301 4K 5U 10 201 3L 4 SE m completed the task successfully and without errors the system is 10 20 30 40 5K 1020 30 40 5 a the task did not take too much time to be completed the system is 1 20 32 40 sl 1020 30 40 sX gile 2 What do you think are the advantages of System A Circumstantial Knowledge Management for Human Like Interaction It provides an API an thus can be used by other systems or scripts 3 What do you think are the disadvantages of System A Too much typing users have to look up the API and must have basic programming skills 4 What do you think are the advantages of System B GUIs are easier to understand 5 What do you think are the disadvantages of System B No scripting capabilities 6 In what situations do you think System A would be convenient Whenever there is a repetitive and scriptable task 7 In what situations do you think System B would be convenient For u

circumstantial knowledge management - e

Contents

Download Pdf Manuals

Related Search

Related Contents