GuiTAR stands for "General-purpose Tool for Anaphora Resolution".It was developed in Java at Essex University.(see http://dces.essex.ac.uk/research/nle/GuiTAR/)
Be the first to post a text review of GuiTAR. Rate and review a project by clicking thumbs up or thumbs down in the right column.
------------------------------------------------------------------------------ - Modifications (on 23/Feb/2006): By Mijail A. Kabadjov - Logging of feature values - Google key extracted from code and placed as parameter in TagSet ini file. ------------------------------------------------------------------------------ - Modifications (on 18/Feb/2006): By Mijail A. Kabadjov - The usage of DN classifier has been parametricised, so now there is only one version of class DefiniteDescription and it runs, depending on input parameter given by user, with SVM classifier, MaxEnt classifier, or No DN classifier at all (i.e., original VP algorithm ). - Still to do: Integrate the above change to the GUI, which still assumes only SVM classifier at the moment. ------------------------------------------------------------------------------ - Modifications (up to 10/8/2005): By Mijail A. Kabadjov 1) In class Agreement the method newInstance() was retouched. Basically some agreement features were updated and some were added (e.g. for "its"). 2) In class NominalGroup the "switch" statement was modified, so as to trigger an independent resolution for possessive pronouns (see next bit). 3) A new class PossessivePronoun was created by cloning class PersonalPronoun and modifying it by adding a constraint that a possessive pronoun cannot corefer with the possessed thing (similar to the one in DefiniteDescription). 4) In GTAR_main the code around the call to the DiscourseModelImplementer constructor was modified. Basically a new variable qCounter was created and it is updated after every file has been processed. It keeps track of the queries thrown at Google for the processing of each file. It was put in place of "i" (i.e. the first parameter of the call dmi.setSVM(...)), which only accounted for the index of the file being processed and not the number of queries as it should have. 5) In DefinteDescription there were three references to POS tag values, which were hard coded (e.g. "JJS"). I rewrote them by refering to classes TagSet and Local. - Known Bugs: These are not really bugs, but rather comments. 1) For some reason, the '.bat' files in 'data' folder did not work on my machine. I presume it might be because of the different operating system - my machine runs Windows 2000. 2) After compiling the code and producing the GuiTAR.jar by hand, I had another problem. When I ran the system like this: java -jar GuiTAR.jar ... I got the "ClassNotFound" exception for class "svm" in file libsvm.jar. I then regenerated GuiTAR.jar to include both, libsvm.jar and googleapi.jar, but still had the same problem. I guess in order for the system to find the other .jar files probably this must be specified somehow in the MANIFEST file. I got round this by setting the classpath accordingly and running GTAR_main directly: java -classpath ... GTAR_main .... 3) Regarding the new Evaluation module: 3.1. I didn't see how I can run only the evaluation module, without having to reprocess the files. That is, after I have processed the files, what I used to do is run several times the evaluation module over the "processed.*" files with different parameters to produce separate perfromance statistics for every anaphor type that the system resolves. With the new version I ran it several times, but every time it regenerated the 'processed.*' files. In other words, I see you can skip the preprocessing step or the evaluation step, but it seems you can't skip the resolution step, which of course in a certain context wouldn't make sense, but in the context described above it might be useful. 3.2. The AnaphoraResolutionEvaluation.csv file has changed. This was the file that I used the most to do the error analysis, and in the new version there is information that is not there anymore. For instance the list of NPs only includes the anaphors evaluated and not all the NPs as before. Is this the only difference? 4) After processing the files with the option '-log', for some reason the file AnaphoraResolution.log was always empty. - Modified Files: GTAR_main.java NominalGroup.java Agreement.java PossessivePronoun.java DefiniteDescription.java ------------------------------------------------------------------------------ - Modifications (up to 08/08/2005): By Philippe J. Goux Incorporated Mijails changes to the new version of GuiTAR, including into GTAR_main. Made modifications to Mijails new code to improve performance. - Modified Files: GTAR_Evaluation.java GTAR_main.java GTAR_Runner.java Agreement.java Cf.java DefiniteDescription.java DiscourseEntity.java DiscourseModel.java DiscourseModelImplementer.java DiscourseModelImplementer2.java AboutDialog.java BrowsePanel.java CheckBoxPanel.java FileBrowser.java FileObject.java Gui.java MainFrame.java OptionPanel.java ResultPanel.java Utils.java IOXMLUtils.java LexicalWord.java NominalGroup.java PersonalPronoun.java PreProSyntacticHeuristics.java ProperName.java History.java IniFile.java Language.java Local.java Log.java TagSet.java ------------------------------------------------------------------------------ - Modifications (up to 08/08/2005): By Mijail A. Kabadjov Integrating the LIBSVM classifier DiscourseModel.java: Added two versions of the method getClassification(), one for MaxEnt, the other for LIBSVM Lines 9, 20-36 DiscourseModelImplementer.java: Added code related to MaxEnt and LIBSVM classifiers, a Global query counter (that's for Google) and a sleeping bit in case the number of query surpasses the daily limit. Lines 9-21, 43-49, 100, 104-129, 171-193, 838-845, 848 (Line 17) Cf.java: Declared a public query counter GTAR_Runner.java: (Lines 15, 17-19, 23) DefiniteDescription.java: Replaced the previous file. The new file contains code to compute the input features for every definite description used by the discourse-new classifier. After resolving a DD as before the classifier decides whether to keep it or rule it out. For this version Google features are read from the XML file and not queried on the fly. Haven't updated the spooling, e.g. Local.out.print... NominalGroup.java: Added the resolution of possessive pronouns (i.e., the switch case) (Line 209) - Modified Files: DiscourseModel.java DiscourseModelImplementer.java Cf.java GTAR_Runner.java DefiniteDescription.java NominalGroup.java ------------------------------------------------------------------------------ - Modifications (up to 04/08/2005): By Mijail A. Kabadjov Replaced the files AnaphoraResolutionEvaluation.java and GTAR_Evaluation.java, and recompiled the whole thing leaving out GTAR_main.java. Then I ran the system as before using the .bat files I had. Only updated the .bat files to evaluate in addition for proper names (i.e. pn). ------------------------------------------------------------------------------ - Modifications (up to 01/08/2005): By Philippe J. Goux Created the history class containing all the history values for the GUI. Created the language class to store all the different sentences in the GUI and output of GuiTAR. Corrected bug using Arrays.binarySearch(). Created our own methods for string search, case sensitive or not. Modified value in LexicalWord from String to String[] array to enable more flexibility when adapting different tag sets. Modified the IniFile so that an empty initialise() method is called. This can be every useful when overridden by caller of object (see language class and local class). GuiTAR_main parameters changed. Also added version parameter and tag set parameter, enabling the user to decide which tag set to use. - Warnings: In order to run GuiTAR on Olivias data, on must first modify the tagset.ini file or create a new one. - Modified Files: GTAR_main.java Agreement.java DefiniteDescription.java DiscourseEntity.java AboutDialog.java BrowsePanel.java CheckBoxPanel.java FileBrowser.java FileObject.java Gui.java MainFrame.java OptionPanel.java ResultPanel.java Utils.java IOXMLUtils.java LexicalWord.java PersonalPronoun.java PreProSyntacticHeuristics.java ProperName.java History.java IniFile.java Language.java Local.java Log.java ------------------------------------------------------------------------------ - Modifications (up to 22/07/2005): By Philippe J. Goux Added many different icons to GUI. Changed GUI so that it is now able to be in multiple languages. The languages are stored in the language file and can be easely edited. The way the language file is dealt with makes it easy to add any new language to the file, and thus to the GuiTAR GUI without having to recompile the whole tool. Added flag icons for most of the contries world in the image folder for futur use with the language file. changed Log class to be able to store errors in a file so that they can be analysed afterwards. Changed Local class to add an error printstream. Added a history system so that the GUI of GuiTAR recalls the last actions of the user. Corrected the fact that the GuiTAR GUI could be used only once. Now, the tree works fine and the GUI is reliable. Corrected bugs here and there. Modified rules to better results for proper name resolution. - Known Bugs: Due to all the added fancy functions to the GUI, it is slightly longuer to load and the GuiTAR.jar file if bigger. - Warnings: Still a lot of work should be done on the name property capure as most of the errors in proper name resolution come from there. The best way to greatly increase the reliability of proper name resolution would be to add Named Entity Recognition during pre-processing to determine what the names are (i.e. persons, companies, objects, etc.). No history exists for log types and evaluation types. They should be coming next. - Modified Files: GTAR_main.java DiscourseEntity.java AboutDialog.java BrowsePanel.java CheckBoxPanel.java FileBrowser.java FileObject.java MainFrame.java OptionPanel.java ResultPanel.java Utils.java LexicalWord.java ProperName.java IniFile.java Local.java Log.java ------------------------------------------------------------------------------ - Modifications (up to 15/07/2005): By Philippe J. Goux A first glimpse to a simple Graphical User Interface. Bugs corrected here and there. - Known Bugs: Once the process has been launched once using the GUI, the tree created will not be refreshed for the second process launched. - Warnings: The GUI is far from flawless. Bugs still subsist and many other changes may be done. This GUI is only temporary and has been done in only one days work. - Modified Files: GTAR_main.java DiscourseEntity.java DiscourseModelImplementer.java AnaphoraResolutionEvaluation.java AboutGUI.java BrowsePanel.java CheckBoxPanel.java FileBrowser.java FileObject.java MainGUI.java OptionPanel.java ResultPanel.java Utils.java IOXMLUtils.java PreProSyntacticHeuristics.java ProperName.java IniFile.java Local.java Log.java ------------------------------------------------------------------------------ - Modifications (up to 13/07/2005): By Philippe J. Goux Modified the GetName() method in the DiscourseEntity class to better the quality of the name found. Corrected bug in the creation of CSV file by the evaluation process. Corrected bug for displaying multiple types of traces. - Known Bugs: Bugs still subsist in the evaluation process as proper name resolution is still ignored - Warnings: Although proper name resolution is performed, it is ignored by the rest of the GuiTAR system, and so the result cannot be shown. - Modified Files: GTAR_main.java DiscourseEntity.java DiscourseModelImplementer.java AnaphoraResolutionEvaluation.java IOXMLUtils.java NominalGroup.java PreProSyntacticHeuristics.java ProperName.java IniFile.java Local.java Log.java ------------------------------------------------------------------------------ - Modifications (up to 12/07/2005): By Philippe J. Goux Created new Log system to enable a selection of the traces to display. This facilitates greatly the debugging process. Corrected some errors in the way the name sets are created. Modified the evaluation process to enable the creation of a proper CSV file. Modified some rules in proper name resolution. Created a new Local class to store any variables common to many classes (such as the log or the IniFile). Modified the IniFile class so that many INI files can be called for different purposes. Such as one for tag sets and tag groups, one for history content or other (once a GUI is created). Completely revisited the batch file to enable an automatic up to date of any new class created. - Known Bugs: Some errors can still be found in the name set creation and this is the major reason why proper name resolution is not working. - Modified Files: GTAR_main.java Agreement.java DefiniteDescription.java DiscourseEntity.java DiscourseModelImplementer.java AnaphoraResolutionEvaluation.java GenerateErrorsInHTML.java LexicalWord.java NeAlignment.java NominalGroup.java PersonalPronoun.java PreProSyntacticHeuristics.java ProperName.java IniFile.java Local.java Utterance.java ------------------------------------------------------------------------------ - Modification (up to 06/07/2005): By Philippe J. Goux Loading only once the DOM Document from an XML for entire GuiTAR procedure. The way each file is treated by GTAR_main. Use to be all files do step 1, then all files do step 2... Now, the entire process is executed on file 1, then on file 2... Added a parameter to enable to log information if desired. A couple of rules added in ProperNoun.java. Not much... The IniFile has been modified to enable group of variables to be made. This clarifies the file to facilitate the modifications to be made by the user. Also, a comment tag has been added. Any line starting with ":" will be ignored when the IniFile is read. [...] - Known Bugs In DiscourseEntity.java, method getName() does not find all proper names. Some Cf tagged pn have no name property. Proper names resolved in ProperNoun.java are not shown in the HTML output. Still no significant progress on proper name resolution. - Files Modified: GTAR_Evaluation.java GTAR_main.java GTAR_Runner.java DiscourseEntity.java DiscourseModel.java DiscourseModelImplementer.java AnaphoraResolutionEvaluation.java GenerateErrorsInHTML.java IOXMLUtils.java LexicalWord.java NeAlignment.java NominalGroup.java PersonalPronoun.java PreProSyntacticHeuristics.java ProperNoun.java Segment.java IniFile.java ------------------------------------------------------------------------------ - Modification (up to 03/07/2005): By Philippe J. Goux The structure of the file to download has been changed to clarify which files are created for test purposes and which are used to compile GuiTAR. Batch files have been added to create documentation, to run test, and even to create a GuiTAR.jar file which contains classes, source and documentation. GuiTAR has now a GUI opening with quick usage explanations. Some classes have had comments modified to enable a smooth run of javadoc on GuiTAR. Code has been updated to use generics. Thus, GuiTAR now needs to be compiled with javac version 1.5 or higher. Only a few classes haven't been modified due to the structure of the code. GuiTAR now uses a INI file to store POS tags. If none exist, a default IniFile.ini is created were GuiTAR is lauched. More rules have been implemented for proper name resolution. More bugs have been corrected. [...] - Known issues: Cf with a pn tag exist with no PN pos tags in them, and thus no name Property objects. The proper name resolution return a posible antecedant and it is ignored by GuiTAR. - Modified files: GTAR_Evaluation.java GTAR_main.java Cf.java DiscourseEntity.java DiscourseModel.java DiscourseModelImplementer.java AnaphoraResolutionEvaluation.java IOXMLUtils.java LexicalWord.java NeAlignment.java NominalGroup.java PersonalPronoun.java PreProSyntacticHeuristics.java ProperNoun.java Segment.java IniFile.java - Warning: the ProperNoun.java class does not work correctly yet. Only excact matches may have a chance to be resolved. More indepth ------------------------------------------------------------------------------ - Modifications (up to 17/06/2005): By Philippe J. Goux Added class GTAR_main.java to enable a single process to run all of GuiTAR's functions. - Modified files: GenerateErrorsInHTML.java GTAR_Evaluation.java GTAR_Runner.java PreProSyntacticHeuristics.java - Warning: the ProperNoun.java class is hollow, the code within it does not work. The modification made for GTAR_main.java are basic and do not change the fact that information goes back and forth to files instead of staying in memory. Therefore, the performance of the overall algorithm is only slightly improved. ------------------------------------------------------------------------------ - Modifications (up to 16/06/2005): By Philippe J. Goux Added VPCTest and VPCTrain corpuses. Added two batch files to run tests and evaluation. adding the name set to DiscourseEntity.java. adding constant VALID_PROPER_NOUN_TAGS to LexicalWord.java. adding search loop to ProperNoun.java. adding NAME constant to Property.java. - Modified files: DiscourseEntity.java: adding the name set. LexicalWord.java: adding constant VALID_PROPER_NOUN_TAGS. ProperNoun.java: adding search loop. Property.java: adding NAME constant. - Warning: the ProperNoun.java class is hollow, the code within it does not work. ------------------------------------------------------------------------------ - Modifications (up to 13/06/2005): By Philippe J. Goux Source code from Mijail set up to work using JBuilder 9 or above. Added class ProperNoun.java. Modified class NominalGroup.java by adding a NominalGroup.PN to switch-case of method newInstance(). - Warning: the ProperNoun.java class is hollow, the code within it does not work. ------------------------------------------------------------------------------ - Modifications (up to 29/03/2005): By Philippe J. Goux Delivered the Local.java file which should contain all the variables which depend on the local environment (OS, directory, etc.). Delivered the IniFile.java file which should contain all the functions and variables to enable GuiTAR to set and get constant variables stored within an external text file. - Requirements: GuiTAR is not present in this delivery. Next update should contain GuiTAR updated using the proposed text file system to store constants.
Be the first person to add a text review.
Copyright © 2009 Geeknet, Inc. All rights reserved. Terms of Use
Thanks for your rating!
Would you also like to write a review?