Home

Improving landfill monitoring programs

1. Multilingual Grammar Based Language Training Computational Methods and Tools Master of Science Thesis in Software Engineering and Technology ELNAZ ABOLAHRAR CHALMERS UNIVERSITY OF TECHNOLOGY Department of Computer Science and Engineering Division of Language Technology G teborg Sweden September 2011 The Author grants to Chalmers University of Technology and University of Gothenburg the non exclusive right to publish the Work electronically and in a non commercial purpose make it accessible on the Internet The Author warrants that he she is the author to the Work and warrants that the Work does not contain text pictures or other material that violates copyright law The Author shall when transferring the rights of the Work to a third party for example a publisher or a company acknowledge the third party about this agreement If the Author has signed a copyright agree ment with a third party regarding the Work the Author warrants hereby that he she has obtained any necessary permission from this third party to let Chalmers University of Technology and University of Gothenburg store the Work electronically and make it accessible on the Internet Multilingual Grammar Based Language Training Computa tional Methods and Tools ELNAZ ABOLAHRAR E
2. 56 T 56 6 3 2 Technical Concerasi 57 1 Introduction As we are in the communication era today human beings have realized the importance of communication more than ever Nowadays being able to communicate with other nations in far or near countries is not consid ered a luxury anymore as it might have been a few decades ago instead it is an essential demand which is growing at a rapid rate In this regards human languages as the major means of communication have gained an unconventionally high significance over time The actual problem with human languages is the fact that there are so many of them which are all unique regardless of probable similarities There exist roughly 6500 spoken languages in the world today of which 4500 have more than 1000 speakers I In order to be able to communicate we need to find some way to understand each other s languages Many efforts have been undertaken towards solving this problem and although great achieve ments have been gained the process is still ongoing and seems to be an everlasting challenge Among the works done in this area that is mutual between computer science and linguistics we have focused on GF Grammatical Framework and its unique and conceptual approach toward translation and multilinguality Thus GF is the major source of inspiration for this work and actually forms the basis that this thesis is built upon In brief the aim of this project
3. CSS stands for Cascading Style Sheets is a style sheet language used to describe the presentation semantics the look and formatting of a document written in a markup language Its most common application is to style web pages written in HTML and XHTML 21 but the language can also be applied to any kind of XML document 22 23 From the historical point HTML was never intended to contain tags for formatting a document but only to define the content of a document Therefore when tags like lt font gt and color attributes were added to the HTML 3 2 specification it started a nightmare for web developers Development of large web sites where fonts and color information were added to every single page became a long and expensive process To solve this problem World Wide Web Consortium W3C created and added CSS to HTML 4 0 24 Accordingly CSS is designed primarily to enable the separation of document content e g written in HTML or a similar markup language from document presentation including elements such as the layout colors and fonts This separation can improve content accessibility provide more flexibility and control in the specification of presentation characteristics enable multiple pages to share formatting and reduce complexity and repetition in the structural content Thus External Style Sheets stored in css files can save a lot of work both in the development and maintenance phases 23 31 2 1 6 JavaScript
4. 1 2 2 EngOnline What is EngOnline 6 Provided by the Center of Language and Communication at depart ment of Applied IT which is mutual between Chalmers University of Technology and Gothenburg University EngOnline is an on line 4 service that provides the students with an on line grammar book and also lots of exercises and quizzes exclusively designed for the English language learners Advantages The teaching material and the assessing procedures are smoothly combined Quizzes and exercise sets are generated automatically by drawing questions randomly from a question data base Thus various com binations are readily available for different purposes like homework exams and so on Drawbacks As stated above exercises are drawing from a question data base The same problems with Moodle about limited diversity and need for manual supervision hold here as well It is only targeting one language namely English 1 2 3 Rivstart What is Rivstart Rivstart is the name of a method for teaching Swedish language published by Natur amp Kultur Its main materials consist of a text book Textbok and a CD containing mp3 files accompanied by an exercise book Ovningsbok Rivstart comes in four levels A1 A2 B1 B2 according the CEFR standard levels 7 The part that we are most interested in about this method is its website which provides lots of on line exercises for each chapter in the b
5. 26 interface LexFoods open Syntax in 1 oper wine N N pizza N N cheese N N fish N N fresh A A warm A A italian A A expensive A A delicious A A boring A A Figure 10 A lexicon interface for Foods grammar 2 instance LexFoodsEng of LexFoods open SyntaxEng ParadigmsEng in oper wine N mkN wine pizza N mkN pizza cheese N mkN cheese fish N mkN fish fish fresh A mkA fresh warm mkA warm italian A mkA Italian expensive A mkA expensive delicious A mkA delicious boring A mkA boring Figure 11 An English instance of the lexicon interface for Foods grammar 2 Finally we see an instance of the lexicon interface for Italian in Figure 12 and the concrete syntax for Italian in Figure which has now become very straightforward to write 2 2T instance LexFoodsIta of LexFoods open SyntaxIta ParadigmsIta in oper wine_N mkN vino pizza_N mkN pizza cheese_N mkN formaggio fish_N mkN pesce fresh_A mkA fresco warm_A mkA caldo italian_A mkA italiano expensive_A mkA caro delicious_A mkA delizioso boring_A mkA noioso Figure 12 An Italian instance of the lexicon interface for Foods grammar 2 concrete Foodslta of Foods FoodsI with Syntax SyntaxIta LexFoods LexFoodsIta Figure 13 The Italian concrete syntax by a functor 2 See also Sect
6. Officially managed by Mozilla Foundation JavaScript is a scripting light weight programming language that was primarily designed to add in teractivity to HTML pages by providing enhanced user interfaces and dynamically JavaScript is usually embedded directly into HTML pages as it is an interpreted language means that scripts execute without pre liminary compilation 25 However as the amount of the JavaScript code is considerably large in the Quiz application about one kilo lines of code without taking the minibar application code into account we have organized it in modules which are in the from of separate external js files that accompany the html page These modules are available in detail in Appendix C Section 6 3 of this document In the Quiz application we have used JavaScript in the form of client side JavaScript as JavaScript s primary use which is known to work in all major browsers such as Internet Explorer 26 Firefox 27 Chrome 28 Opera 29 and Safari 30 25 Some other characteristics of JavaScript which are worth mentioning here are It is a prototype based object oriented scripting language which also supports the structured program ming paradigm It is dynamic weakly typed as types are associated with values and not with variables It has first class functions which are ob jects themselves So they have properties and methods and they can be assigned to variables passed as arguments returned
7. The command format is as shown in the following example where the from option determines the source language we want to translate from and the to option sets the target language Finally the cat option de termines the starting category for the random tree generation action for example A for adjective V for verb and N for noun gt translation quiz from DemoEng to DemoSwe cat N Here is an screen shot of the translation_quiz command in the GF shell 10 5 C WINDOWS system32 cmd_exe gf DemoSweEng pgf This is GF version 3 2 License see help license A Bug reports http code google com p grammatical framework issues list Languages DemoEng DemoSwe Demo translation quiz from DemoEng to DemoSwe cat N welcome to GF Translation Quiz The quiz is over when you have done at least 10 examples with at least 75 success You can interrupt the quiz by entering a line consisting of a dot Figure 5 translation quiz in GF shell However before the users can use this command they should perform some initialization tasks First of all they should have GF installed on their system to be able to run GF at the first place For practical installation instructions please refer to the GF official website at http www grammaticalframework org Then they have to get e g download the required GF grammar files with gf suffix and make the appropriate PGF file which contains all the from and to langu
8. Appendix B Sections and Configuration Variables which consist of minimum percentage of right answers score counter required to pass the Quiz minimum number of questions required to pass the Quiz number of questions in the Exam Mode which is a fixed number and finally maximum times the user may answer the question to increase his her score Mode Settings which makes it possible to modify inner settings for all mode properties including access to the Minibar magnet words Previous Question Hint and Check Answer buttons etc 39 The Server Configurations They define the PGF server the Quiz needs to communicate with and they are modifiable only from within the code It is possible to choose between a predefined server accessible over the Internet or a local server Below is an example code the variable in charge of this setting is called online_options More details about the GF Web Service API can be found in Section 2 1 3 Also in Appendix B Section 6 2 3 you can find information on how to launch this web service On your own server var online options 1 grammars url http www grammaticalframework org grammars grammars url http tournesol cs chalmers se 41296 grammars grammars url http localhost 41296 grammars grammar list Foods pgf gt 2 2 3 Functionalities Here is a list of the Quiz main functionalities that we will discuss its items individually afte
9. Very Quality gt Quality Fresh Warm Italian Expensive Delicious Boring Quality Figure 6 The abstract syntax module of Foods grammar 2 14 The first line is the module header which contains the module type ab stract as well as the module name Foods The module body consists of different judgment forms as we explain below 2 flags flag definitions e g the one in our example states that Comment is the start category by default cat category declarations telling what categories types of ab stract trees exist in the grammar fun function declarations telling what tree building functions there are in the grammar Now we discuss the concrete syntax modules of Foods grammar for En glish and Italian languages The English concrete syntax is given on page 17 Again the header indicates the module type which is a concrete syntax named FoodsEng for the abstract syntax of Foods grammar The judgments applied here are 2 lincat linearization type definitions indicating what are the type of produced objects resulted from linearizing trees of each category Note Linearization means the procedure of converting trees to strings lin linearization rules telling how trees are linearized In order to understand the the concrete module we need to explain some GF concepts first 2 Parameters By means of the param judgment in GF we can de fine new types exactly the same as
10. gt s table 1 Masc gt table Sg gt nero Pl gt neri Fem gt table Sg gt nera Pl gt nere ES regAdj Str gt Adjective nero gt let ner Str init nero in adjective nero ner a ner i ner e copula Number gt Str table Sg gt e Pl gt sono 19 As you might have noticed another difference from the English version is the introduction of a new form in the type of adjective operation which is called function types with variables The type nero nera neri nere Str gt Adjective is the same as Str gt Str gt Str gt Str gt Adjective just that the use of variables make it possible to share the argument type In this case we could have used wild cards as well because the the variables make no semantic difference However using mnemonic vari ables help the user of the function to give the proper forms of arguments 2 See other examples of this form in Section 2 1 2 under inflection paradigms needed for the Foods grammar on page 24 The last difference is that in ResIta we have used type synonyms for the linearization type of noun phrases nouns and adjectives This is all for the purpose of cleaner and more structured code as well as enhanced data abstraction 2 In the following subsection we describe the GF Resource Grammar Li brary and how it can be used to write application grammars Then we rewrite the same Foods example shown in this se
11. 10 November 2010 DOM Document Object Model http www w3 org DOM Ac cessed 11 November 2010 XHTML Extensible Hyper Text Markup Language http www w3 org TR xhtml1 Accessed 3 March 2011 XML Extensible Markup Language http www w3 org TR xml Accessed 7 March 2011 CSS Cascading Style Sheets http www w3 org TR CSS2 Ac cessed 3 December 2010 CSS Cascading Style Sheets web tutorial Available from http www w3schools com css css_intro asp Accessed 3 Decem ber 2010 w3schools com http www w3schools com js js intro asp Accessed 2 November 2010 Internet Explorer http www microsoft com windows internet explorer Accessed 2 November 2010 Mozilla Firefox http www mozilla com en US firefox personal html Accessed 2 November 2010 Google Chrome http www google com chrome intl en land ing chrome html hl en Accessed 2 November 2010 Opera http www opera com docs specs Accessed 15 November 2010 59 30 31 32 33 34 Safari http www apple com safari Accessed 15 November 2010 Scheme http www schemers org Documents Standards R5RS Accessed 20 March 2011 OCaml http dries ulyssis org apt packages ocaml ocaml spec html Accessed 20 March 2011 JavaScript https developer mozilla org en JavaScript Language Resources Accessed 1 November 2010 Marquez L and Gimenez J Automatic Evaluation in Machine Translation Towards Simi
12. 21 and XML documents 22 Aspects of the DOM such as its Elements may be addressed and manipulated within the syntax of the programming language in use which is JavaScript in our case DOM is likely to be best suited for applications where the docu ment must be accessed repeatedly or out of sequence order which is the exact case with our Quiz application DOM has been applied to make 30 it possible to inspect or modify a web page dynamically by JavaScript scripts In other words DOM exposes its containing HTML page and browser state to JavaScript 20 Getting to know the Minibar application as we are using it in the Easy Study Mode of the Quiz to make word magnets available for the user to be used in constructing answers and already facing problems with static elements in the Quiz the design applied in the Minibar was an inspiration to change the Quiz approach to use DOM as well For this purpose functions available in support js from the Minibar application 17 have been applied to access and modify the HTML elements from the JavaScript code dynamically See Section 2 2 1 for a detailed example on how applying DOM benefits the Quiz application 2 1 5 CSS Another change that took place in the Quiz development time was using a CSS external sheet to include the entire application s styling and layout instead of the inefficient style attributes in HTML tags for the purpose of controlling the layout and style of the web page
13. by other functions and manipulated like any other object JavaScript is also considered a functional programming language like Scheme BI and OCaml 32 be cause it has closure and supports higher order functions By closure JavaScript allows a combination of code that can be executed outside the scope in which it is defined with its own scope to be used during that execution 33 With all its diverse properties JavaScript as all other programming languages has many advantages and also some drawbacks Here we will briefly mention those we encountered Let s start with its advantages Advantages It is open source and therefore accessible to all without the need to purchase a license JavaScript supports regular expressions which provide a concise and powerful syntax for text manipulation that is more sophisti cated than the built in string functions 33 We have used regular expressions in preprocessing the raw user input as we will explain 32 in detail in Section 2 2 3 under Preprocessing and Evaluating the user s answer Check Answer Being very popular today JavaScript has abundant freely available supporting material and communities in the Internet Drawbacks The characteristic of being weakly typed makes the programming very easy indeed but some times has unexpected consequences which are difficult to avoid and costs more effort to be controlled eventually I encountered the same trad
14. gramming which is a significant rule in the functional programming paradigm functions are introduced under oper judgments Note This functions are called operations to avoid confusion with fun functions in abstract syntax modules The following example is an operation defining the regular inflection of nouns in English oper regNoun Str gt s Number gt Str word gt s table Sg gt word Pl gt word s Which is very similar to function definition in any func tional programming language Appending operator vs Gluing operator The operator combines two lists of tokens into one list while the operator combines two tokens into one token for example foo bar foobar but foo bar foo bar 16 concrete FoodsEng of Foods lincat Comment Quality s Str Kind s Number gt Str Item s Str n Number lin Pred item quality ls item s copula item n quality s det Sg this That det Sg that These det P1 these Those det P1 those Mod quality kind s Nn gt quality s kind s n Wine regNoun wine Cheese regNoun cheese Fish noun fish fish Pizza regNoun pizza Very a s very a s Fresh adj fresh Warm adj warm Italian adj Italian Expensive adj expensive Delicious adj delicious Boring adj boring This param Number Sg P1 ope
15. in the whole educating procedure It is difficult because it has to be correct and precise grammatically and new and creative at the same time and the main trouble is that this difficult process has to be repeated many times over and over as time passes by There are two specific reasons for this need being up to date and also creative rather than sticking to old clich s which may end up in a mere memorizing activity As computers were created to take the burden of exactly the same type of jobs dull full of repetitions and demanding high accuracy at the same time off human shoulders it seems that it is worth the effort to computerize the exercise generation process To achieve this goal we have focused on GF and its unique conceptual approach toward translation and multilinguality Here we only mention those GF s characteristics that the Quiz application has benefited from We will discuss the GF system itself in detail in Section 2 1 1 We can summarize a list of linguistic requirements of the GF Translation Quiz as follows High grammatical precision as the first priority Support for multilinguality Coverage of natural languages in small fragments GF addresses all these requirements First of all it provides a high level grammar formalism which allows writing multilingual domain specific application grammars also called semantic grammars at minimum ex pense Moreover having a resource grammar library along wit
16. is done in functional languages like Haskell Here is an example which defines the type Number which has two values Singular and Plural param Number Sg Pl Another example is the Gender type which is either Femi nine or Masculine param Gender Masc Fem Tables are applied in GF to formalize inflection tables where each value of a certain parameter type is assigned a string the corresponding inflection form Below is an example of a table type and its definition in GF 15 Number gt String table Sg gt mouse Pl gt mice Records and record types The same as their general meaning in computer science records in GF are data structures that gather objects of different types together Here are two examples in GF The first record contains a string with label s and a Number with label n and the second one contains a table from Numbers to strings labeled s and also a Gender with label g s mouse n Sg s table Sg gt mouse Pl gt mice g Fem Selection Operator is applied for accessing a value in a table for example table Sg gt mouse Pl gt mice Sg will give the string mouse as its result Projection Operator is used to get access to a value inside a record under given label For example the following projection will result in the number Sg s mouse n Sg 3 n Operation definition In order to avoid copy and paste pro
17. the pass condition occurs or an exam ends a Show Quiz History button will be displayed Clicking on this button will reveal all quiz questions together with user s answers and expla nations from the application in a separate window where the user has the possibility to review save a copy and print the History 53 6 1 3 Technical Concerns The Quiz application has been tested on different browsers like Google Chrome 28 Mozilla Firefox 27 and TE Windows Internet Explorer 26 Due to technical problems with TE and previous graphical issues with Chrome the author recommends Firefox over all tested browsers as the most consistent browser for the Quiz application However presently no problem has been detected with Chrome anymore 6 2 Appendix B Teachers Supervisor s Guide This section is mostly aimed at teachers or education supervisors who would like to use this application as a means of training and or evaluating their students language skills At this level you might want to run the Quiz with other GF grammars and so will need to run a server locally which will be discussed later on this chapter Moreover the quiz can be tailored to meet the specific requirements of different users with different perspectives For this purpose there exist some configurable parts in the JavaScript code which is explained as follows 6 2 1 Configuration Variables Here is the list of configuration variables available in the beginni
18. well as the structural words required for the Foods grammar 2 Moreover we need some inflection paradigms for each language The paradigms needed for English and Italian are shown in Figure 8 2 23 English Function Type dog Str Y man men Str gt N mkA cold Str gt A Function Type mkN vino Str gt N mkA caro s SEX gt A Italian Figure 8 Inflection paradigms needed for the Foods grammar 2 As mentioned earlier mkN makes a noun out of an input string and mkA makes an adjective from its input string In the above paradigms we also see application of function types with variables For example in dog Str gt N the word dog is just a representative of the input type which should be a noun here and could be any other noun As another example in man men Str gt N the words man and men are used just to say that the function takes two input strings which are nouns and the second one is the irregular plural form of the first one This form was explained in more formal terms in Section 1 1 under the resource module for Italian called ResIta on page 20 Now we are ready to write the concrete syntax modules We begin with English The concrete syntax opens SyntaxEng and ParadigmsEng in order to get access to the required resource libraries 2 24 concrete FoodsEng of Foods open SyntaxEng ParadigmsEng in linca
19. with the most common words in the same positions as in the user s answer is selected as the reference For the final evaluation I have applied a strategy very similar to the Mastermind s game 35 Words in green stand for correct words in their correct place in the sentence and yellow words mean these words are correct and are part of the right answer but they are misplaced while red words stand for words which do not exist in the right answer and are either misspelled or totally wrong lexically or grammatically The Quiz also displays blanks for missing words 2 2 Solution Details In this section solution details including the design implementation and testing phases are discussed in a general sense for more technical details refer to Appendix C Section 6 3 Developer s Guide 2 2 1 User Interface The user interface which is in form of a web page has been designed to be extremely compact and narrow in order to fit in a mobile phone screen Figure l4 shows its initial state The layout and styling issues has been mostly handled in a CSS external file as explained in Section Methods and Technology in Subsection 2 1 5 Also most HTML elements like the Menubar and other buttons are added dynamically from within 34 the JavaScript code applying DOM instead of being statically defined in the HTML code This makes the application more flexible and gives the programmer a better control of the application and also makes the main
20. 1 3 1 Why learning a foreign language matters at all As we mentioned earlier people in different parts of the world speak dif ferent languages thus in order to be able to communicate we need to find some way to understand each other The simplest yet most effective way is to get to know the other languages Of course it is not possible for everyone to learn all those foreign languages out there but it is very com mon that some people will need to learn one or more foreign languages in their life time On the other hand languages as an advanced complex means of commu nication are known to be an exclusive ability which has made the intellec tual human being distinctive from other creatures on earth Therefore the ability of applying languages for communication is known as one of the mind s most important properties used to describe humans intelli gence Thus learning languages is not only looked upon as a requirement but also as a brain exercise for intelligence training and as most intelli gence related topics its fun factor for many should not be neglected as well 1 3 2 How can computers help us in the language learning process in general and GF s specific role Learning a language is not possible without training and exercising even for the most talented people On the other hand as professionals in language education usually state creating exercises and exam questions is one of the most difficult dull and time consuming tasks
21. Google Chrome www2 nok se laromedel sprakverktyg nokplayer nokplayer html source riva01verb01 amp name Verb 20presens amp chapter E Skriv r tt verb Sofia komm fr n Grekland komm Tips Flera bokst ver saknas Several letters are missing Figure 4 Rivstart samples several missing letters Advantages The exercises are designed specifically to support a method and are therefore very coherent with the rest of teaching material and thus benefit a lot from their support at the same time for example learners can always refer to the book for examples and explanations Feedback is very precise namely up to the letter s level and there fore very useful for the learners This also provides the possibility to train independently Drawbacks Having static exercises the same problems with Moodle and En gOnline about limited diversity and need for more manual work hold here as well It is only targeting one language namely Swedish 1 3 Background and Inspiration This subsection is about the reasons and motivations behind this work First we discuss the essence of language learning then we continue with describing the desired role of computers in the language learning process This leads us to introduce GF and its capabilities which is our major source of inspiration and actually the Quiz is built upon its strength We will discuss the GF system itself in detail in Section 2 1 1
22. LNAZ ABOLAHRAR September 2011 Examiner AARNE RANTA Prof Department of Computer Science and Engineering Chalmers University of Technology SE 412 96 G teborg Sweden Telephone 46 0 31 772 1000 Cover Multilingual Grammar Based Language Training Computational Methods and Tools Department of Computer Science and Engineering G teborg Sweden September 2011 Acknowledgment I would like to express my special thanks to Aarne Ranta my supervisor for his invaluable help and support and also Thomas Hallgren for his generous assistance with the JavaScript part as well as Krasimir Angelov and Ramona Enache for all their help I also appreciate Olga Caprotti s helpful comments and advice on the report which highly improved it Last but not least I want to thank all friends and family who have helped and supported me by all means to accomplish this work Abstract Living in the communication era we face an ever increasing demand for communication Considering the human languages role as the major means of communication has inspired the author to do this thesis work which we believe is an effort towards solving the challenge of human communication The main goal of this thesis is introducing a teach ing learning aid which we have called the GF Translation Quiz or in short the Quiz to be used in training human languages in general and specifically targeting their grammatical aspects The Quiz is designed as a web applicat
23. a simplified step by step approach 1 The Hint button triggers the show_hint function in the JavaScript code which in return calls the make all answers function if it is the first time Hint is called and this question has not been checked before Otherwise it calls the continue checking in Step 3 2 What happens here in the make all answers function we have already explained in Step 2 of Check Answer which we avoid re peating 3 continue checking will continue running the checking process in three branches 3 1 check_answer_exam2 is called if the quiz is in the Exam Mode which follows similar steps as the Check Answer but behaves differently in terms of showing immediate reaction to user s actions e g it reveals the final score only at the end of the exam 3 2 check answer2 is called which we have discussed in the previous phase Please see Step 3 3 in Check Answer 3 3 show_hint2 is called in case the Hint button was clicked The first action taken in show_hint2 is preprocessing the raw user answer which we have explained in the previous functionality Check Answer 45 3 3 1 Please see Step 3 3 1 in Check Answer 3 3 2 Please see Step 3 3 1 in Check Answer 3 3 3 show_hint2 then compares the processed user an swer to all possible correct answers computed in Step 2 by make_all_answers function and displays the result as explanation below Words in green stan
24. aa Bb oe ee 1 1 1 Multilinguality 6a xeu e REEL EDO da 1 1 3 Stand Alone Exercise Tool 1 2 State of the ar i sc ka eee kor neo EES RRS d 1 2 1 Moodle A IA A 123 R vstart 1 3 Background and Inspiration 2 2 2 ko 9 esas 1 83 4 Why learning a foreign language matters at all 1 3 2 How can computers help us in the language learn 0 o oa 0 N N N LM O 1 8 8 Translation Quiz command in GF shell 10 13 2 1 Methods and Technology 13 AR A E 13 TON 20 ae ua M Wero era 28 PPP A ee 30 JO COSS pee 2 ee ae da eb RUP e 31 J L5 JEVADOHBi e x oc som XXE g dren gts d 32 2 1 7 BLEU an Automatic MT Evaluation Method 33 2 2 Solution Details a2 2 esa eee doe Bee we ee aw 34 2 2 1 User Interfacel 2 sso o ok OX Rn 34 Sp fe Biche QR Od ae up le 4 g r aen 38 2 2 3 Functionalities yume x ege hao ae de o 40 a AA TE 46 2 3 1 How to make your own quiz 46 2 3 2 How the Translation Quiz handles morphology ques E A A A ee 48 49 3 1 Limitations and Drawbacks 49 3 2 Advantages a uae uox gor ku do Re rss ss 49 50 51 52 6 1 Appendix A User manual 52 6 1 1 Customization 2 2 RE 52 Ones a A EDU 53 6 1 3 Technical Concerns 54 54 A 54 Ea CR a Rays 55 0 2 8 Technical Concerns 2 55 6 3 Appendix C Developer s Guide
25. ages related GF grammar files PGF Portable Grammar Format files are the portable format for grammars written in GF The following command is required for this purpose 2 gt gf make name lt name of pgf file gt lt concrete grammars gf gt In the above command 1 to n grammars may be applied at the same time The only condition is that all these concrete grammars should share the same abstract syntax grammar The resulted PGF file is a binary file 11 which contains a pre compiled and optimized multilingual grammar as defined in the applied concrete grammars 2 Note that you should be in the directory that contains the grammar files For example here we have the demo grammar for Swedish and English located in gf lib src demo and we have named the PGF file as DemoSweEng gt gf make name DemoSweEng DemoSwe gf DemoEng gf And then they can start GF by calling the name of the PGF file to import the grammars like this in our example gt gf DemoSweEng pgf As one can see it is not a trivial task to work with GF directly from a system s shell specially for those who are not much familiar with com puting let alone GF The greatest achievement by the web based Quiz application is bringing GF capabilities to users with no prior acquain tance with computers or linguistic background Accordingly one of our goals has been to design the application in a way that even school chil dren and retired people find it easy to
26. ammars Although these methods do not cover the complete GF grammar notation until this moment but they are very much easier than the previously mentioned ways and still meet the needs of non GF grammarian users Here we discuss them briefly 46 GF online editor for simple multilingual grammars For the means of GF grammar writing a recent feature has been introduced which makes writing grammars much easier specially for non specialist users It is called GF online editor for simple multilingual grammars This online editor written by Thomas Hallgren is available at http www grammaticalframework org demos gfse Traditionally GF grammars are created in a text editor and tested in the GF shell Text editors know very little if anything about the syntax of GF grammars and thus provide little guidance for novice GF users Also the grammar author has to download and install the GF software on his her own computer On the contrary in the online editor all that is needed is a reasonably modern web browser 37 The editor also guides the grammar author by showing a skeleton gram mar file and hinting how the parts should be filled in When a new part is added to the grammar it is immediately checked for errors Last but not least in spite of its name the editor runs entirely in the web browser so once you have opened the web page you can continue editing grammars even while you are offline 37 Thus users can write t
27. anguage Eng Parse in question language IN You PI Run LinearizeAll to answer language Swe ni springer Eng LinearizeAll to answer language S AON You_Sg Run Swe du l per du springer ni l per Figure 17 How all possible correct answers are generated 44 Analyzing the user s answer and Providing Instructive Feed back Hint Besides the basic feedback provided in the Check Answer see previ ous functionality there exists a more sophisticated feature in the Quiz called the Hint This feature is created to fulfill one of the main goals of this project namely providing users with instructive and helpful feed back The Hint button is available in all Quiz Modes but the Exam Mode With automatic update upon input change and no usage restric tion it is designed to be extremely handy in the Easy Study Mode Ac tually with the help of this feature it is very hard for the users to go wrong The evaluation process combines a simplified version of the MT evaluation method called BLEU together with a strategy very similar to the Mastermind s game 35 For technical details of the applied methods in the Quiz Hint see Section 2 1 7 Although there is some interoperabil ity between Check Answer and Hint the Hint mechanism is designed to work completely independent of the Check Answer As we did for the Check Answer phase here we explain the Hint procedure in
28. anguage to use in the linearization the linearization result text the abstract syntax tree to server linearizeAll linearize language to use in the linearization a list of all possible linearizations text a list of abstract server parse the string to be parsed syntax trees as plain strings Table 1 Calls to the PGF server defined in the JavaScript API in pgf_online js The following calls were used indirectly in the Quiz for they exist in the minibar application I7 We have slightly modified minibar to meet our Quiz requirements and integrated it into the Quiz application as a component options grammars url server get_ grammarlist server switch grammar server get languages For a list of all supported calls by the GF Web Service API please refer to the GF Wiki page available at 13 http code google com p grammatical framework wiki GF WebServiceAPI 29 Also in another guide by Thomas Hallgren you will find these calls il lustrated by examples in addition to a detailed reference on how to make these calls from JavaScript using the API defined in pgf_online js L7 I8 This guide is available at 18 More information about how to run a web server and launch the Quiz application with your own grammar is available in Appendix B 6 2 2 1 4 HTML and DOM Hyper Text Markup Language abbreviated as HTML is the predominant markup language for web pages where a markup la
29. ation Explanation INO the correct answer s is are de d r ostarna r italienska Current Quiz Mode Answered Questions Your Score Current Quiz Mode Answered Questions Your Score Medium Study Mode o L3 o Medium Study Mode Figure 16 Hint vs Check Answer 2 2 2 Configurations The application has three different ways to be fine tuned Through User Customizations designed in the interface for end users Configuration Variables and Mode Settings found within the code which are mostly aimed at instructors and teaching supervisors and Server Configura tions which are related to the GF web services and are mostly aimed at developers and partly to instructors and teaching supervisors User Customizations They are handled from the user interface and are designed to be as easy as possible From the Menubar users can customize the quiz to their needs i e they can choose their desired Grammar From source and To target languages and also a Quiz Mode Easy Study Mode In this mode word magnets are available to help users they can type and or click on them freely to construct their answer Delete last and Clear buttons can be used to delete the last word or the whole answer in case required These 38 word suggestions come from the selected grammar and therefore it keeps users from misspelling words and even making grammat ically wrong sentences Also users have unlimi
30. brary called ParadigmsEng for example 2 lin Wine mkN wine This will make a noun out of the string wine The Resource Grammar API is divided into language specific and language independent parts 2 22 The syntax API is language independent i e has the same types and functions for all languages It is called SyntaxL for each lan guage L The morphology API is language specific i e has partly different types and functions for different languages It is called ParadigmsL for each language L In Figure 7 we see a table of all phrasal categories and rules and also the structural words we require for the Foods grammar example which is a small fragment of the whole Resource API 2 Category Explanation clause sentence with all tenses she looks at this very warm red house the red house very this one place adjective warm common noun house Function Type NP gt AP Sener CL Example mkNP Det gt CN gt NP this old man mkCN N gt CN mkCN AP gt CN gt CN very big blue house NA SL Lt old very very old mkAP A gt AP mkAP AGA gt AP gt AP Function Type In English this Det Det this that_Det Det that these Det Det this those_Det Det that very AdA AdA very Figure 7 Resource Grammar API presentations for phrasal categories and rules as
31. ch English Finnish French German Italian Nepali Norwegian bokm l Persian Polish Punjabi Roma nian Russian Spanish Swedish and Urdu while more languages are under construction 12 For an up to date list of available languages please refer to the GF official website at http www grammaticalframework org The Resource Grammar Library has manifold language processing appli cations such as translation multilingual generation software localization natural language interfaces and spoken dialogue systems In this thesis work we have applied it as the resource for getting grammatically precise translations between the From source and To target languages in the Quiz application Here we will rewrite our example the Foods grammar from the previous section by applying the GF Resource Grammar Library Just as a re minder the whole examples and explanations in Sections and 2 1 2 are quoted or summarized from the Grammatical Framework Program ming with Multilingual Grammars by A Ranta 2 The abstract syntax module remains unchanged and is the same as in Figure 6 Before going to the Foods grammar example we need to give some preparatory explanations here In the Resource Grammar Library the goal is to achieve grammatically correct combinations of words regardless of their meaning This makes it possible to cover a much larger subset of languages compared to semantic grammars Based on this approach a Res
32. ction 2 1 1 1 1 3 Stand Alone Exercise Tool Although the Quiz application can be applied as a very helpful aid inside a language learning methodology it is initially intended to be a stand alone exercise tool for self training Therefore providing learners with helpful guidance and instructive feedback is a high priority This pur pose is covered by a combination of features and functionalities in the application namely Check Answer Hint and History Specially the Hint which analyzes the input answers and gives accurate feedback on wrong answers This makes the application easy to use by the learners with out the aid of a teacher or other help resources A simplified version of BLEU 4 which is an automatic machine translation evaluation method combined with the Mastermind s game strategy has been applied to im plement this feature See more details about the Quiz functionalities in Section 2 2 3 1 2 State of the art In this section we discuss briefly some existing computerized language training and exercise generating applications as listed below 1 2 1 Moodle What is Moodle Moodle abbreviation for Modular Object Oriented Dynamic Learn ing Environment is a Course Management System CMS also known as a Learning Management System LMS or a Virtual Learning Environment VLE It is a free and open source e learning web application that educators can use to create effective on line learning sites Its dis
33. ction this time by means of the Resource Grammar Library 2 1 2 GF Resource Grammar Library Writing multilingual domain specific application grammars also called semantic grammars in GF as we saw an example of in the previous section becomes even simpler by applying the GF Resource Grammar Library Such grammars cover similar semantic fragments in several nat ural languages and therefore are extremely redundant Thus applying the Resource Grammar Library also saves us a considerable amount of time and effort 9 The great benefit we get from GF Resource Grammar Library is that it makes writing multilingual domain specific grammars also called appli cation grammars in GF The GF Resource Grammar Library is a set of natural language gram mars implemented in GF These grammars share a common abstract 20 syntax that is in the format of a tree structure Each language is then obtained via compositional mappings from abstract syntax trees to a concrete syntax specifically written for that language The grammar de fines for each language a complete set of morphological paradigms and a syntax fragment The GF Resource Grammar Library plays the role of the standard software library much similar to the Standard Template Library of C or the Java API 12 The library is available as open source software under the GNU LGPL License Currently it covers 25 languages including but not limited to Bulgarian Catalan Danish Dut
34. d for correct words in their correct place in the sentence and yellow words mean these words are correct words and are part of the right answer but they are misplaced while red words stand for words which do not exist in the right answer and are either misspelled or to tally wrong lexically or grammatically Hint also displays a blank for each missing word Possibility to have the whole Quiz Show Quiz History The application keeps a record of the quiz questions user s answers and the explanations as a quiz is running At the end of the quiz which is either invoked manually by the user clicking on the End Quiz button or by the application itself e g when the pass condition occurs or an exam ends a Show Quiz History button will be displayed Clicking on this button will reveal all quiz questions together with user s answers and explanations from the application in a separate window where the user has the possibility to review save a copy and also print the History 2 3 Extra Features 2 3 1 How to make your own quiz This part is mostly aimed at teachers and education supervisors who wish to make their own desired quizzes This simply requires writing their own grammar and applying it in the current web interface of the Quiz Apart from the ordinary way of writing a GF grammar and by help of GF Resource Grammar Library that we saw in Sections and 2 1 2 there are two additional possibilities for writing a GF gr
35. e with Google Chrome version 7 0 517 44 28 as default browser It has also been tested and verified with Mozilla Firefox version 3 6 12 27 However it does not work with TE version 6 0 29 26 due to JavaScript restrictions Another important issue is that occasionally some disorder in displaying the word magnets has been observed in Chrome browser which seems to be a rendering bug in Chrome However this problem has been by passed from within the Minibar application 36 in its released version on Oct 26th of 2010 and after that and thus the same holds for the Quiz application 97 References 1 10 11 12 13 Spoken languages in the world http www infoplease com askeds many spoken languages html Accessed 2 November 2010 Ranta A Grammatical Framework Programming with Multilin gual Grammars CSLI Publications Stanford 2011 Google translator http translate google com Accessed 1 Novem ber 2010 Papineni K Roukos S Ward T and Zhu W J Bleu a Method for Automatic Evaluation of Machine Translation IBM Research Report RC22176 W0109 022 September 17 2001 Moodle http moodle org Accessed 2 November 2010 EngOnline https learning portal chalmers se Accessed 2 November 2010 CEFR Common European Framework of Reference for Lan guages http www coe int t dg4 linguistic cadre en asp Ac cessed 14 December 2010 Rivstart http w
36. eckAns Determines whether the user can use the Check Answer button Currently the default value is true for all modes but the Exam Mode max hint times Determines how many times the user can use the Hint button for each question Currently the default value is unlimited times for Easy Study Mode three times for Medium Study Mode only once for Hard Study Mode and evidently zero times for Exam Mode 6 2 3 Technical Concerns Which Browser to use The Quiz application has been tested on different browsers like Google Chrome 28 Mozilla Firefox 27 and IE Windows Internet Explorer 26 Due to technical prob lems with IE and previous graphical issues with Chrome the author recommends Firefox over all tested browsers as the most consistent browser for the Quiz application However presently no problem has been detected with Chrome 95 How to make your own quiz A detailed discussion about this can be found in Section 2 3 1 How to install and run a PGF server locally For a detailed up to date tutorial please refer to the Grammatical Framework Wiki page available at Belor chat you aed to install GF itself for this step refer to the ee page How to apply a new grammar For a detailed up to date tuto rial please refer to the Grammatical Framework Wiki page available 6 3 Appendix C Developer s Guide Finally in this section some information is provided for users at a higher le
37. ection 1 2 3 50 5 Conclusion In brief the aim of this thesis work is introducing an automated exercise generator to be used in training grammars in general as well as the lexical and syntactic aspects of different human languages This tool is designed and implemented as a web application which we have called the GF Translation Juiz or in short the Quiz that makes it also available on mobile phone platforms such as iPhone and Android We have a very good reason to call it thus GF Grammatical Framework and its mostly unique and conceptual approach toward translation and multilinguality is the major source of inspiration to this work and forms the basis that this thesis is built upon its strength Therefore most characteristics of the Quiz application i e grammatical precision support for multilinguality and coverage of natural languages in small fragments are provided by GF From another point of view the Quiz application plays the role of an interface so that ordinary users with no prior computer knowledge can also benefit from GF s capabilities for the purpose of language learning The GF Translation Quiz provides an extendable frame work from two different dimensions Creating new exercises by defining new abstract syntax grammars Adding new languages by writing the corresponding concrete syn tax grammar However although the current frame work is quite accomplished and ro bust but there are a lot of possib
38. education supervisors who would like to use this application as a means of training and or evaluating their students language skills Finally in Appendix C some information is provided for users at a higher level e g programmers developers who might want to apply the Quiz s code within their own applications 1 1 Objectives This section focuses on the main objectives of this project each objective describes one advantageous aspect of the project 1 1 1 Multilinguality One important goal in this project has been introducing a multilingual application and by multilingual we mean that the system is available in many languages that the user can choose from and it does not nec essarily mean that all languages are applied at the same time 2 The generated exercises are in the preliminary form of sentences in a certain language which need to be translated to another one In other words the application is intended to be general in terms of languages and give the users many language choices For this purpose GF s Resource Grammar Library have been applied as the constructive base of the project Fur thermore the possibility to extend the system by adding new languages is highly intended which is a major achievement in GF 1 1 2 Grammatical Precision and Language Coverage The question to be answered here is that how much grammatical pre cision we actually need in the Quiz application and how big chunk of each language should it cover at t
39. g item n This det Sg questo questa That det Sg quel quella These det Pl questi queste Those det Pl quei quelle Mod quality kind ls table n gt kind s n quality s kind g n g kind g gt Wine noun vino vini Masc Cheese noun formaggio formaggi Masc Fish noun pesce pesci Masc Pizza noun pizza pizze Fem Very a s table g n gt molto qual s g n Fresh adjective fresco fresca freschi fresche Warm regAdj caldo Italian regAdj italiano Expensive regAdj caro Delicious regAdj delizioso Boring regAdj noioso gt 18 Here is the resource module called ResIta that is used by the FoodsIta concrete syntax 2 resource ResIta open Prelude in param Number Sg P1 Gender Masc Fem oper NounPhrase Type ls Str g Gender n Number Noun Type s Number gt Str g Gender Adjective Type s Gender gt Number gt Str det Number gt Str gt Str gt Noun gt NounPhrase n masculine feminine commonNoun gt s table Masc gt masculine Fem gt feminine commonNoun g commonNoun s n g commonNoun g n n l3 noun Str gt Str gt Gender gt Noun vino vini g gt 4 s table Sg gt vino Pl gt vini 8 8 i adjective nero nera neri nere Str gt Adjective nero nera neri nere
40. h the Con trolled Languages concept defined in GF makes it even easier to write application grammars which cover similar fragments in several natural languages at the same time The so called Controlled Languages are fragments of natural languages designed to be clear and unambiguous so that they become mechanically processable Thus they get the bene fits of formal languages and since they are actually fragments of natural languages they perfectly fit the demands of natural language training applications such as our Translation Quiz Thus GF saves us a consider able amount of time and effort required to add new languages to and or extend the existing grammars 9 A detailed example on how grammars actually work in GF is discussed in Sections and 1 3 3 Translation_Quiz command in GF shell This thesis project was initiated by an attempt to improve the GF Mul tilingual Grammar Tour which is a multilingual web document designed for teaching the grammar of supported languages in GF by A Ranta There are a lot of examples and exercises in this document which rely on the GF shell for execution Among these exercises there is a shell com mand called translation_quiz that was a starting point for the Quiz application We borrowed the idea and developed it into our current web based Quiz application with its many features which will be discussed in detail in Section 2 2 Here I briefly describe how this command works in the GF shell
41. he initial state 36 D GF Translation Quz gt C fi Ofile c E a Ye en RY Signin Farsi C3 other bookma Foods pof JN Easy Study Made Restart Quiz End Quiz 213 these boring pizzas are very warm LA eee ee ee ee ee ee eee ee eee ee a o mo RR ee a eros fenes o e varma d r Vat i m ST Study Mode 2 Figure 15 The Easy Study Mode 37 L GF Translation Quiz C fi file C Elnaz Chalmers Master 20Tl yz A 23 Chalmers Studen Sign In Farsi Other bookmarks Show Information 3 GF Translation Quiz C ft O file C Elnaz Chalmers Master 20T1 Yz A 23 Chalmers Studen sign In Farsi Sd Other bookmarks Show Information Grammar Foods poi gt From Eng To Quiz Mode gt N Restart Quiz End Quiz Grammar Foods pof__ From Eng To Quiz Mode M d study Mode Restart Quiz End Quiz Quiz Question Quiz Question 1 those cheeses are Italian ji those cheeses are Italian Your Answer Your Answer gt de d r osterna ar italianska de d r osterna r italianska lt Previous Question Next Question gt Hint Check Answer Previous Question J Next Question gt Hint Check Answer f Hint de d r osterna r italianska l Explan
42. he same time According to an old wisdom in natural language processing coverage and precision cannot be maximized at the same time thus there is a trade off between coverage and precision 2 Therefore if we put aside a hard coded application which cannot go wrong but has not much to offer as well there remains a choice between the so called data driven and grammar driven transla tion systems that the Quiz application could rely on Where data driven methods e g Google translator 3 mostly based on statistical methods provide a much larger coverage of the language compared to grammar driven ones but of course with no warranty for grammatical accuracy 2 However in our case making the decision is not difficult at all for 2 the purpose of the application which is teaching languages enforces grammatical precision as an inevitable priority Evidently the error tol erance of the system should be absolute zero Who could imagine a reliable learning tool that generates wrong output which is totally self contradicting Moreover it is more effective to expose learners to a new language in a gradual manner rather than throwing them into the entire language all at once The same holds for the Quiz application therefore we do not need large language coverage Conversely we would rather have languages in small fragments That is what we can get from the Controlled Languages concept available in GF See more about this in Subse
43. heir grammar with the online editor in an easy interactive way and test them in the Minibar L7 or Quiz frameworks afterwards The gfm format Another approach to grammar writing which is even easier is writing them in a very simple and straight forward format called the gfm format In this format a multilingual grammar is a simple text file with a gfm suffix which contains language elements including words phrases and whole sentences in all destined languages in a multi column style so that they could be easily mapped to one another Here is an example which relates some English Italian and Swedish nouns and noun phrases langs Eng Ita Swe man uomo man woman donna kvinna boy ragazzo pojke kille girl ragazza flicka tjej a good man un buon uomo en god man a good woman una buona donna en god kvinna these girls queste ragazze dessa flickor de h r flickorna these boys questi ragazzi dessa pojkar de h r pojkarna 47 The text is then saved in a file named Foods gfm The compiler recog nizes the suffix gfm and creates these four files Foods gf FoodsEng gf FoodsIta gf and FoodsSwe gf The first line gives the list of applied language names Multi word expres sions and free variations are allowed Words of different languages are separated by a while free variants within a language are separated by a As a grammar grows larger as in the above example the benefits of
44. if the two previous conditions fail The first action taken in check_answer2 is preprocessing the raw user answer For this purpose we have applied regular expressions 3 3 1 First all unwanted characters including digits punc tuation marks and in brief all none letter characters are replaced by a space character apostrophe is exceptionally allowed for its semantic role in the many grammars These characters are defined by the following regular expression in terms of ranges of Unicodes Nu0021 Nu0026 Nu0028 Nu0040 u005b u0060 Nu007b Nu007e g 3 3 2 T hen all extra spaces defined in this regular expression Ns Ns g are removed and finally all white space characters stated by the regular expression s g are replaced with a single space which is required in the an alyzing phase taking place in the next phase namely the Hint feature of the Quiz This approach makes the appli cation more robust and flexible towards the user s subtle mistypings and even towards a careful high standard in put including punctuations that might not be supported in the grammars 42 3 3 3 check_answer2 then compares the processed user answer to all possible correct answers computed in Step 2 by make all answers function and finally displays the result of this phase in the explanation part of the Quiz interface 43 generate_question Generated random abstract tree generate_question Linearize to question l
45. ion that makes it also available on mobile phone platforms such as iPhone and Android It may be applied either as an aid in a language learning methodology or as a stand alone exercise tool for self training The Quiz is intended to help teachers and education supervi sors by reducing the burden of creating exercises and exam questions and at the same time help learners by providing them with instructive feedback About the background many efforts have been done in this mutual area between computer science and linguistics among which we have focused on GF Grammatical Framework and its unique conceptual approach to wards translation and multilinguality Thus this thesis work relies heavily on GF from many aspects i e most characteristics of the Quiz applica tion i e grammatical precision support for multilinguality and coverage of natural languages in small fragments are provided by GF From an other point of view the Quiz application works as an interface that makes GF capabilities available to ordinary non specialist users List of Abbreviations GF Grammatical Framework PGF Portable Grammar Format MT Machine Translation BLEU BiLingual Evaluation Understudy JS JavaScript CSS Cascading Style Sheets HTML Hyper Text Markup Language DOM Document Object Model CEFR Common European Framework of Reference Contents 1 Introduction Ll Obpjectives s ecis esre gipon oa
46. ion for more information about how to write your own grammar for the Quiz 2 1 3 GF Web Service API Some of the functionalities available in the GF shell are also available via the GF Web Services API 13 which is a small application that exposes the PGF API in form of a web service The application uses FastCGI 14 as communication protocol to talk with the web server and the data protocol in use is JSON 15 16 We can list the most important tasks in the Quiz application as follows Generating the question Evaluating the user answer Analyzing the user answer and providing instructive feedback 28 In all these phases we need to call the PGF server which is handled by the GF Web Services API 13 Moreover as the Quiz application is written in JavaScript we have used an additional interface for JavaScript provided by Thomas Hallgren which is available in pgf_online js 17 18 Here is a table containing calls to the PGF server that we used directly in the Quiz in the format they appear in the JavaScript API defined in pgf_online js 17 18 We have only mentioned the input and output elements that were used in the Quiz and not all existing ones More details about how the calls were applied in the Quiz can be found in Section Functionalities Call as in JavaScript API Input Output the generated get d ori as ae abstract syntax tree the abstract syntax tree to server linearize linearize l
47. is introducing a teaching learning aid to be used in training the lexical morphological and syntactic aspects of human languages by discussing computational methods as well as implementing a tool for this purpose The tool is designed as a web application that makes it also available on different mobile phone platforms such as iPhone and Android From this point on we call this web application the GF Translation Quiz or in short the Quiz You can find more details about the application itself in Chapter B Solution Description This chapter explains the objectives of the project state of the art and also the background and inspiration behind this thesis work which the author believes is somewhere in the path towards solving the challenge of human communication Later on in Chapter 2 we explain our solu tion that is the resulted application by describing first the applied methods and technologies and then the application itself including im plementation and technical details In the 3rd Chapter the outcomes of the project are evaluated and both advantages and drawbacks are discussed In the following Chapter some future work possibilities are suggested Afterwards we have concluded the whole thesis work in the final Chapter which is followed by three Appendices A B and C Ap pendix A explains Quiz features and functionalities and how they work with an intention for end users while Appendix B is mostly aimed at teachers or
48. itional problems that exist mostly in structured programming paradigm especially as the code was grow ing larger It was very likely to end up in a redundant spaghetti code However enforcing a modular design manually by the pro grammer can help a lot in this case This solution in return requires a thoughtful design to make it also easy to keep track of the flow control with all those modules 2 1 7 BLEU an Automatic MT Evaluation Method In order to be able to give instructive feedback we need to analyze and evaluate the answer given by the user For this purpose a simplified ver sion of MT Machine Translation evaluation methods has been applied Although it might seem a totally different problem our purpose has a lot in common with MT evaluation process In both cases the input is a given translation which needs to be evaluated by being compared against one or more reliable reference translations Therefore we found it very useful to use the same techniques The main technique applied here has a lot in common with the famous method known as BLEU BiLingual Evaluation Understudy 4 Here we describe briefly how BLEU works then will continue with the method applied in the Quiz More general end user information can be found in Section D 2 3 under Analyzing the user s answer and Providing Instructive Feedback Hint First of all let s see how Blue answers this fundamental question How can we judge a translation The closer a
49. larity Measures Based on Multiple Lin guistic Layers Presented in MOLTO workshop GF meets SMT Gothenburg November 5 2010 Available from http www molto project eu sites default files MT evaluation seminar pdf Accessed 2 December 2010 Mastermind Available from http en wikipedia org wiki Mastermind board game Accessed 2 November 2010 Minibar Documentation Available from http www grammaticalframework org 41296 minibar about html Accessed 2 November 2010 GF online editor for simple multilingual grammars http www grammaticalframework org demos gfse about html Accessed 20 February 2011 60
50. le future work which can enhance the application further Some are as follows Creating a user account system for keeping a long term history for each user which allows users to keep track of their training in the long run Designing specific grammars in order to fulfill the Quiz demands Controlling and directing the random generation of phrases and sentences and so on For more future work suggestions please see Section 51 6 Appendices 6 1 Appendix A User manual This chapter explains all the Translation Quiz s features and functional ities and how they work with an intention for end users See also Figures and 15 for sample states of the user interface 6 1 1 Customization From the Menubar users can customize the quiz to their needs i e they can choose their desired Grammar From source and To target lan guages and also a Quiz Mode Easy Study Mode In this mode word magnets are available to help users they can type and or click on them freely to construct their answer Delete last and Clear buttons can be used to delete the last word or the whole answer in case required These word suggestions come from the selected grammar and therefore it keeps users from misspelling words and even making grammat ically wrong sentences Also users have unlimited Hint which is automatically updated as the answer is being modified It is also possible to go to one previous question a
51. machine translation is to a professional human translation the better it is This is the central idea behind the BLEU method To judge the quality of a machine trans lation one measures its closeness by a numerical metric to one or more 33 reference human translations Therefore BLEU requires two ingredi ents first a numerical translation closeness metric and second one or more good quality reference translations 4 In our application we already have the second part i e at least one grammatically accurate one hundred percent correct translation namely linearizations of all possible abstract trees to the target language by the PGF server which we will use as our reliable reference For the first part in the BLEU method The main idea is to use a weighted average of variable length phrase matches against the reference translations 4 34 We should note that our requirements is different from judging and deciding which translation is a better one our goal is to give accurate feedback We also have free variations explicitly existing in our reference translations so we do not need to care about that either Because of these differences the metric applied in the Quiz is also different i e it is simplified so that it only considers words as phrase elements Moreover it keeps track of the position of words in the phrase It works like this the user answer is compared with all possible correct answers and the one
52. ndom generation of phrases and sen tences Adding an initial test for determining user s knowledge level can be added This can help users to know where to start i e which level Defining and making the Quiz configurable for different levels of lan guage knowledge by means of designing specific grammars and defin ing equivalent levels adapting to standards like the Common European Framework of Reference for Languages Learning Teaching and Assess ment abbreviated as CEFR 7 Introducing an offline version of the Quiz which does not require a constant internet connection and is much faster in the case of low speed connections or remote users Introducing a user account system with capability of keeping the records of learners training so that they can keep track of their overall progress and find their weak and strong points This information can be useful for teachers or supervisors the same way Thus the training can be more organized and fruitful The above history system can also provide some useful feedback for the Quiz itself for example the questions that the user got wrong could appear again soon or the ones he got right may not appear again and so on Enhance the Hint so that it can catch user errors in letters scale in addition to words scale For example it can detect wrong or misplaced letters in a word Maybe something similar to what is done in Rivstart web page For more details see Rivstart in S
53. ne submission chance to increase their score further submissions don t have any impact on their score Therefore it is recommended to use the Hint first to make sure they have got the right answer of course if Hint is available see below Moreover following the learning from your mis takes approach it is more effective for the learners to first find where they were wrong rather than seeing the right answer all at once Here we explain the underlying procedure in a simplified step by step approach 1 The Check Answer button triggers the check_answer_quiz func tion in the JavaScript code which in return calls the make all answers function if it is the first time this question is being checked and Hint has not been called before Otherwise it calls the continue checking in Step 3 2 Here we explain what happens in make all answers function There are cases where a single word in the source From language translates to two or more different words with distinct meanings and applications in the target To language For example the word you in English becomes du singular you or ni plural you in Swedish Therefore there are more than one correct answers in these cases Moreover GF grammars may include free variations which are alternative concrete syntax objects that map to the same abstract syntax tree 2 This means that a single word in a cer tain language may have more than one translation e
54. ng In both cases and specially in the later one providing the user with helpful feedback is an important functionality of the application This is han dled by the Hint feature of the Quiz which analyzes the user input and gives instructive feedback on wrong answers The application is generally planned to be easy to use for a wide range of users without the help of a teacher or other help resources required In this chapter we will discuss the methods and technology applied the Quiz application details as well as known application s limitations 2 1 Methods and Technology 2 1 1 GF GF Grammatical Framework has more than one description based on different angles from which we look at it It can be described as a special purpose functional language for defining grammars 10 and also as a grammar formalism which is based on constructive type theory to ex press the semantics of natural languages for multilingual grammars and their applications TO In both senses GF uses a Logical Framework LF 11 for a description of an abstract syntax module accompanied by any number of concrete syntax modules The abstract syntax is a tree like representation that captures the semantically relevant structure of the language fragment while the concrete syntaxes map the tree struc ture with linear text representations of target languages Although GF 13 grammars themselves are purely declarative they can be used both for lineari
55. ng of the translation_quiz js file a JavaScript component of the Quiz The minimum percentage of right answers score counter required to pass the Quiz with current default value of 0 75 which equals to 75 var pass percentage 0 75 The minimum number of questions required to pass the Quiz with current default value of 10 var min no questions 10 The number of questions in the Exam Mode which is a fixed number with current default value of 20 var exam_quesNo 20 54 The maximum times the user may answer the question to increase his her score with current default value of 1 However with the current Check Answer mechanism which reveals the right answer it does not make much sense to change this variable var max_answer_times 1 6 2 2 Modes Settings Here are the modes settings with their default values available under the same title Modes Settings in the set_mode function in the translation_quiz js file a JavaScript component of the Quiz have_minibar Determines whether the Quiz has the word mag nets available from the Minibar application I7 Currently the de fault value is true only for the Easy Study Mode and false for all other modes have_ prevQuestion Determines whether the user can use the Previous Question button Currently the default value is true for Easy Study Mode and Medium Study Mode and false for Hard Study Mode as well as Exam Mode have_ch
56. nguage is a set of markup tags designed to describe web pages The web page content con sists of HTML elements defined by tags surrounded by angle brackets like lt html gt HTML tags normally come in pairs like lt b gt and lt b gt The first tag in a pair is the start tag the second tag is the end tag they are also called opening tags and closing tags The role of a web browser is to use the tags to interpret the content of the page to be able to read HTML documents and display them as web pages 19 The base of the Quiz application is a web page written in HTML where HTML elements form its building blocks It serves as a means to create structured documents by denoting structural semantics for text such as headings paragraphs lists links forms quotes and other items All the Quiz functionalities are added in JavaScript to accompany this HTML base 19 The HTML base and the JavaScript component have evolved in paral lel and at the same time in an interactive way through the application development history One example of this interaction is transferring to a DOM Document Object Model 20 in order to interact dynamically with the HTML elements through JavaScript rather than static elements in the HTML file Here we explain DOM and how this change was done in the Quiz The Document Object Model DOM is a cross platform and language independent convention for representing and interacting with objects in HTML XHTML
57. ook The exercises are of various types like multiple choice questions phrase matching blank filling and etc Again the exercises we are most interested in are the blank filling ones which have a very precise feedback system which catches the user s errors up to letter s scale Below are some sample screen shots of these exercises and the system s feedback For more information about Rivstart and its on line exercises see 8 Natur och Kultur Google Chrome WWW2 nok se laromedel sprakverktyg nokplayer nokplayer html2source riva0iverb01 amp name Verb 20presens amp chapter E Skriv r tt verb Sofia arbeta p Karolinska Institutet arbeta Tips En bokstav saknas One letter is missing Figure 1 Rivstart samples missing letter Natur och Kultur Google Chrome vw2 nok se laromede sprakverktyg nokplayer nokplayer htmi source riva01verbO18name Verb 20presenstchapter C23 Skriv r tt verb Sofia rbetar p Karolinska Institutet rbetar Tips En bokstav r fel One letter is wrong Figure 2 Rivstart samples wrong letter Natur och Kultur Google Chrome o www2 nok se laromedel sprakverktyg nokplayer nokplayer html source riva01verb0 1 amp name Verb 20presens amp chapter E Skriv r tt verb Sofia tallar grekiska engelska och svenska tallar Tips En bokstav f r mycket One letter is extra Figure 3 Rivstart samples extra letter Natur och Kultur
58. ource Grammar has two kinds of categories and two corresponding type of rules 2 lexical lexical categories to classify words lexical rules to define words and their properties 21 phrasal also known as combinational or syntactic phrasal categories to classify phrases of arbitrary size phrasal rules to combine phrases into larger phrases In an abstract syntax the lexical rules are fun functions that take no arguments while phrasal rules are functions that do take arguments 2 There is also another classification inside the lexical categories and that is open vs closed categories In general closed categories contain struc tural words also known as function words Here are examples of the closed category words 2 Det stands for determiner e g this that AdA stands for ad adjective e g very And below are some examples of the open category words 2 N stands for noun e g wine pizza A stands for adjective e g delicious warm The major benefit we get from defining such classification is reusability for the words in closed categories can be listed once and for ever in a library For the Foods grammar case we can use the following structural words from the SyntaxEng module existing in the library 2 this Det that Det these Det those Det Det very AdA AdA For the linearizations of the open words used in the Foods grammar we can apply the morphological paradigm li
59. quivalents For example the verb run in English may have springer and l per as its equivalent translations in Swedish concrete syntax Taking care of these possibilities the function make_all_answers pro duces an array containing all possible correct answers by following these steps also illustrated in Figure 17 2 1 The random abstract tree generated by generate_question function in the question generation phase is linearized to the source language 41 2 2 The result text is then parsed to the source language This may result in additional trees to the one we started with thus we can cover the first case that we explained in the above example 2 3 The resulted abstract tree s are then linearized to the target language separately for this step the linearizeAll command is used instead of ordinary linearize in order to get all translation alterna tives including free variation 2 4 Then it will continue with continue checking in Step 3 3 continue checking will continue running the checking process in three branches 3 1 check_answer_exam2 is called if the quiz is in the Exam Mode which behaves differently in terms of showing immediate reaction to user s actions We will discuss it later 3 2 show_hint2 is called in case the Hint button was clicked which we will discuss later in the following functionality called Hint 3 3 check_answer2 is called
60. r det Number gt Str gt s Number gt Str gt s Str n Number n det noun gt s det noun s n n noun Str gt Str gt s Number gt Str man men gt s table Sg gt man Pl gt men regNoun Str gt s Number gt Str car gt noun car car s adj Str gt ds Str cold gt s cold copula Number gt Str table Sg gt is Pl gt are gt 17 Next to discuss is the Italian concrete module of the Foods grammar that is shown below The Italian version is different from the English one in a few aspects First of all you might ask where did all param types and oper definitions go To explain this we should note that these parts are completely independent of any abstract syntax and actually they may be applied in many concrete syntaxes for different abstract syntaxes and some times even for different languages Therefore to increase reusability GF provides a module type called resource which can contain oper and param judgments Then it is enough that the the concrete module opens the resource module to be able to use its definitions 2 concrete FoodsIta of Foods open ResIta in 1 lincat Comment s Str Kind s Number gt Str g Gender Item s Str g Gender n Number Quality s Gender gt Number gt Str lin Pred item quality ls item s copula item n quality s item
61. rwards We will also mention the methods and technologies behind their implementation very briefly More details about the methods and technologies themselves can be found in Section Generating a new question Next Question Preprocessing and Evaluating the user s answer Check Answer Analyzing the user s answer and Providing Instructive Feed back Hint Possibility to see the whole Quiz Show Quiz History Generating a new question Next Question Triggered by the Next Question button the generate_question function from the JavaScript is called which in turn sends a random tree generation request to the PGF server by this command server get_random finally the server 1linearize will linearize the generated random ab stract tree to the selected source language and display the result to the user 40 Preprocessing and Evaluating the user s answer Check An swer Users can check whether their answer is right by either clicking on this button or pressing the Enter Return key of their keyboard What hap pens from a user s point of view is that he she will receive a basic feed back in this phase which is either approval Yes that was the correct answer or denial No the correct answer s is are which reveals the correct answer s Users have to be careful because checking the an swer also submits their answer and for each question they have only o
62. stances as arguments and producing modules as values Here is the functor implementation of the Foods grammar The module header uses the keyword incomplete to indicate that Foodsl is a functor 2 25 incomplete concrete FoodsI of Foods open Syntax LexFoods in lincat Comment Cl Item NP Kind CN Quality AP lin Pred item quality mkCl item quality This kind mkNP this_Det kind That kind mkNP that_Det kind These kind mkNP these_Det kind Those kind mkNP those_Det kind Mod quality kind mkCN quality kind Very quality mkAP very_AdA quality Wine mkCN wine_N Pizza mkCN pizza_N Cheese mkCN cheese_N Fish mkCN fish_N Fresh mkAP fresh A Warm mkAP warm_A Italian mkAP italian_A Expensive mkAP expensive_A Delicious mkAP delicious_A Boring mkAP boring_A To obtain a complete concrete syntax this function Foodsl functor needs to take two instances of the interfaces Syntax and LexFoods as its argu ments This action is called a functor instantiation 2 Accordingly the English concrete syntax can be written as shown in Figure 9 concrete FoodsEng of Foods FoodsI with Syntax SyntaxEng LexFoods LexFoodsEng Figure 9 The English concrete syntax by a functor 2 It remains to show how interfaces and their instances actually look like Figure 10 shows the LexFoods interface and an English instantiation of it is shown in Figure 11 2
63. t Comment Cl Item NP Kind CN Quality AP lin Pred item quality mkCl item quality This kind mkNP this_Det kind That kind mkNP that_Det kind These kind mkNP this_Det plNum kind Those kind mkNP that_Det plNum kind Mod quality kind mkCN quality kind Wine mkCN mkN wine Pizza mkCN mkN pizza Cheese mkCN mkN cheese Fish mkCN mkN fish fish Very quality mkAP very_AdA quality Fresh mkAP mkA fresh Warm mkAP mkA warm Italian mkAP mkA Italian Expensive mkAP mkA expensive Delicious mkAP mkA delicious Boring mkAP mkA boring Different languages tend to use syntactic structures in similar ways to express the same meanings We also know that all languages in GF Resource Grammar Library implement the same syntactic structures Therefore in most cases we only need to rewrite the lexical parts of a concrete syntax for a new language Thus we again encounter copy and paste programing which is not wise to do specially in a functional programming paradigm In order to solve this problem we introduce functors A functor is a function that operates on modules In GF a functor is a module that opens one or more interfaces An inter face is a module similar to a resource but it only contains the types of opers and not their definitions The definitions are then given in instances of this interfaces Thus a functor is a module level function taking in
64. t a time and of course users can check their answer s correctness Medium Study Mode In this mode users can use the Hint button for a maximum of 3 times for each question and they can go to one previous question at a time as well as checking their answer s correctness Hard Study Mode In this mode users can use the Hint button only once for each question and they may not use the Previous Question button but they can still check their answer s correct ness Exam Mode In this mode users cannot use the Hint Previous Question or Check Answer buttons and they will not see their score until the end of the exam This piece of information is also available in the application through a link just above the Quiz Menubar 52 6 1 2 Functionalities As most of the Quiz s functionalities are quit self explanatory in this part only the none evident functionalities are explained to make them clear and prevent probable misinterpretations Next Question As it is expected to it will bring up the next question and it will prompt you before doing so if you have not checked your answer to the current question so that you don t miss a question by mistake The only matter you should know about it is that if your answer to the previous question has been correct Check Answer moves you automatically to the next question so that you won t need an extra click on the Next Ques
65. ted Hint which is automatically updated as the answer is being modified It is also possible to go to one previous question at a time and of course users can check their answer s correctness Medium Study Mode In this mode users can use the Hint button for a maximum of 3 times for each question and they can go to one previous question at a time as well as checking their answer s correctness Hard Study Mode In this mode users can use the Hint button only once for each question and they may not use the Previous Question button but they can still check their answer s correct ness Exam Mode In this mode users cannot use the Hint Previous Question or Check Answer buttons and they will not see their score until the end of the exam This piece of information is also available in the application through a link just above the Quiz Menubar Configuration Variables and Mode Settings There are some global variables defined in the code that control the application s general settings We have divided these variables in two groups Configuration Variables which affect the whole quiz and Mode Settings that define the Quiz Modes differences and thus makes it possible to change the Modes behavior as desired Generally these configurations and settings are aimed at instructors and teaching supervisors and they are only modifiable from within the code They are explained in more detail in
66. tenance process much easier For this purpose functions in support js from the Minibar application 36 have been applied To demonstrate the strength and flexibility provided by this method we give a small example The Check Answer button the user sees in the Easy Study Mode of the Quiz Figure looks the same as it does in other Quiz Modes however it is a totally different button i e it performs completely differently in this mode At application start up it is defined as a submit button so that the Enter key will be its shortcut Unfortunately this useful shortcut will become a programmer s nightmare in the Easy Study Mode where the Enter key has another role ie to take the typed text by the user and add it to the answer being constructed This causes a serious conflict with the role of submitting the user s answer for checking In order to avoid this disaster and keep the benefit of checking the answer by the Enter key the Check Answer button is removed and replaced with a none submit one while changing to the Easy Study Mode and vice versa See more details about DOM in Section 2 1 4 35 L GF Translation Quiz WES le C Elnaz Chalmers M vz 32 c fi of e E Y RI Foods pgf w From Easy Study Mode RR asy Study Mode Medium Study Mode Hard Study Mode Quiz queexamMode k isplayed here 4 Explanations are displayed here at NE Uu Figure 14 T
67. the applica tion It requires a high speed internet connection otherwise the latency in receiving responses to each server call will be quite annoying from a user s perspective specially in the case of remote users The auto fill feature which is set as default nearly in all browsers is not so desirable in the Quiz because it keeps everything the user has typed and suggests them while they want to write a new answer This is both distracting from a learner s point of view and also might be considered as cheating from a teacher s point of view in case of a serious exam It lacks a long term history mechanism for example a user account system so that the learners can keep track of their training in the long run 3 2 Advantages Easy to use which covers a large range of users including school chil dren Being based on GF makes it possible to consider and take care of com plex language concepts like free variations where translation alterna tives are possible and the reverse situation of ambiguity where a single word has more than one translations with distinct meanings in another language Extendable by adding grammars and languages 49 4 Future Work Here is a list of possible future work some of which come directly as solutions to the problems stated in the Limitations and Drawbacks in Section Designing specific grammars in order to fulfill the Quiz demands Controlling and directing the ra
68. the modular grammar writing are sensed better However this method works very well for small grammars with simple and short phrases 2 3 2 How the Translation Quiz handles morphology ques tions Aarne Ranta has suggested a simple and splendid idea that makes it pos sible to apply the Quiz for training morphology inside a specific language as well This is possible by defining a grammar in a way that the abstract syntax module contains the morphology forms while one concrete syntax maps these forms to their descriptions and another concrete grammar maps them to their actual forms Thus a translation from descriptions to real forms simply makes a morphology quiz One example grammar of this kind is available on the grammatical framework server it is called MorphoQuizVerbsEng pgf and is written by A Ranta for means of train ing English irregular verb forms It gives the form s description s and expects the real form s as an answer The rest is quite the same as with the ordinary grammars This feature adds another dimension to the Quiz and again more specific grammars can be written to enhance this usage of the Quiz application 48 3 Evaluation This section is organized in two parts First we discuss Limitations and Drawbacks and then Advantages They contain the author s observa tions and concerns and also some users points of view 3 1 Limitations and Drawbacks Here is a list of limitations and some undesirable aspects of
69. tion Check Answer You can check whether your answer is right by either clicking on this button or pressing the Enter Return key of your keyboard Exception In the Easy Study Mode you cannot use the Enter Return key for checking your answer because it has another functionality which is adding typed words to your answer Please note that by checking your answer you also submit it and for each question you have only one submission chance to increase your score further submissions don t have any impact on your score Use the Hint if you are not sure and of course if it is available Finally please note that Check Answer moves automatically to the next question if your answer is right Hint You will find the Hint button very handy in all Quiz Modes In fact it works very much similar to the famous Master mind game 35 Words in green stand for correct words in their correct place in the sentence and yellow words mean these words are correct words and are part of the right answer but they are misplaced while red words stand for words which do not exist in the right answer and are either misspelled or totally wrong lexically or grammatically Show Quiz History The application keeps a record of the ques tions user answers and the explanations as a quiz is running At the end of the quiz which is either invoked manually by the user clicking on the End Quiz button or by the application itself e g when
70. tribution is via standard core packages which the users need to install on their own web servers which can be nearly any platform currently available on Mac Windows and Linux where PHP and a database are already installed 5 Advantages Moodle is a general framework for education and is highly cus tomizable so it can meet diverse demands of a vast range of users It is open source and therefore available to all It is more than a mere application it is a growing community of users developers and literally anyone who is interested which adds great potential for support and further development Drawbacks In order to create a quiz in Moodle the user should first create the questions however he she can also save them in a question bank for later use Therefore in Moodle the burden of designing and creating quiz questions remains unresolved The question banks are reusable of course but still there stand two major difficulties Firstly to get a well designed quiz the user still needs to check the drawn question manually to make sure that they meet his her requirements Secondly as this method is rather hard coded the diversity of the questions is limited to the number of existing questions in the bank which is a static number The interesting point is that Moodle s major advantage which is being general is at the same time a drawback because it cannot cover specific needs of all subjects at the same time
71. use Now the only knowledge required to use the Quiz application is how to open a browser and of course a link to the Quiz s web page Note The GF shell command translation quiz that we explained in this section has only been a source of inspiration to our current web based Quiz application which is a part of this thesis work and has not been used in the application development by any means 12 2 Solution Description The implemented tool is a teaching learning aid aimed at training the lexical morphological and syntactic aspects of human languages The tool is designed as a web application that we have named the GF Trans lation Quiz or in short the Quiz This design decision makes the Quiz also available on mobile phone platforms Basically the Quiz application as it is evident from its name generates quiz questions by randomly constructing phrases from the input GF grammar The generated ques tions are in the preliminary form of phrases in a certain language which need to be translated to another one The Quiz is intended to support grammatical precision and multilinguality In order to meet these re quirements it relies on GF and its Resource Grammar Library which also makes it very easy and systematic to extend by adding new languages to and or extending the existing grammars The Quiz application may be applied either as an aid in a language learn ing methodology or as a stand alone exercise tool for self traini
72. vel than end users e g programmers developers who might want to apply the Quiz s code within their own applications 6 3 1 Code Organization and Modules The code is organized as follows translation quiz html which contains the main HTML ele ments of the Quiz in its general state translation quiz js the main JavaScript code which contains the global variables and different functions of the Quiz can be found here quiz pre start js which prepares and sets the variables prior to page load quiz support js it handles the lower level functions applied in translation quiz js minibar quiz js the JavaScript code that is in charge of the word magnets in the Easy Study Mode of the Quiz It is based on the original minibar js from the Minibar application 17 which is slightly modified in order to serve the Quiz s requirements 56 support js which is the original file from the Minibar application L7 It handles the lower level functions used in minibar quiz js and translation quiz js minibar quiz css and brushed metal png imported from the Minibar application 17 with slight changes in the CSS file handle the styling and graphics in the Quiz application pgf online js which is the original file from the Minibar appli cation I7 It creates the server object with desired parameters 6 3 2 Technical Concerns The Quiz application has been designed and implemented to be compat ibl
73. ww nok se rivstart Accessed 21 November 2010 Ranta A and Angelov K Implementing Controlled Languages in GF CNL 2009 CEUR Workshop Proceedings vol 448 2009 Ranta A Grammatical Framework A Type Theoretical Grammar Formalism The Journal of Functional Programming 14 2 145 189 2004 Bart Jacobs Categorical Logic and Type Theory Elsevier 2001 ISBN 9780444508539 Ranta A The GF Resource Grammar Library Linguistic Issues in Language Technology CSLI Publications Volume 2 Issue 2 2009 Grammatical Framework Wiki page GF Web Service API http code google com p grammatical framework wiki GFWeb ServiceAPI Accessed 2 November 2010 FastCGI http www fastcgi com devkit doc fcgi spec html Ac cessed 19 February 2011 58 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 JSON JavaScript Object Notation http www json org Ac cessed 14 February 2011 GF Wiki page Launching the Web applications http code google com p grammatical framework wiki Launch WebDemos Accessed 2 November 2010 Minibar Application http www grammaticalframework org de mos minibar minibar html Accessed 2 November 2010 GF web services API examples http www grammaticalframework org demos minibar gf web api examples html Accessed 5 November 2010 HTML Hyper Text Markup Language http www w3 org TR html401 Accessed
74. zing abstract syntax trees and for parsing strings Additionally both formal and natural languages can be described in GF The key no tion of this description is a grammatical object which in fact is a record that contains all information on inflection and inherent grammatical fea tures such as number and gender in natural languages or precedence in the case of formal languages 10 2 Here we demonstrate how these GF concepts are applied by quoting an example grammar from the book Grammatical Framework Program ming with Multilingual Grammars by A Ranta 2 This is a rather simple grammar named the Foods grammar which describes short com ments on food Here are some examples This cheese is very expensive Those fish are fresh That wine is very very delicious As stated before GF grammars consist of two module types abstract syntax and concrete syntax Abstract syntax module contains the syntactic functions which are applied for building abstract syntax trees in that grammar while concrete syntax modules explain how the abstract syntax trees are linearized to actual phrases strings First let s take a look at the abstract syntax module of the Foods grammar 2 abstract Foods 4 flags startcat Comment cat Comment Item Kind Quality fun Pred Item gt Quality gt Comment This That These Those Kind gt Item Mod Quality gt Kind gt Kind Wine Cheese Fish Pizza Kind

Improving landfill monitoring programs

Contents

Download Pdf Manuals

Related Search

Related Contents