Chooser User Manual
Contents
1. CLIPPING2. 2 3
3. literal 1 literal 1 Definition usage category member bg v 1 category domain bg n 5 hypernym bg n cekrop 4 cekrop I M 1 Ha
4. CLIPPING 24 jor
5. Ei Fig 10 Chooser s search dialog References Koeva S S Leseva B Rizov E Tarpomanova T Dimitrova H Kukova M Todorova Design and Development of the Bulgarian Sense Annotated Corpus In Proceedings of 22 Department of Computational Linguistics Institute for Bulgarian Language BAS 2012 the Third International Corpus Linguistics Conference CILC 7 9 April 2011 Valencia Spain Koeva S B Rizov S Leseva Chooser a Multitask Annotation Tool In Proceedings of the 6th Language Resources and Evaluation Conference Marrakech Morocco 28 30 May 2008 ISBN 2 9517408 4 0 23
6. Berge Collins 1995 literal 2 literal 9 literal 9 literal 6 literal 6 Definition PY usage bq derivative n 1 CLIPPING 4 SS SSS KAKTA NANUANTARAT HAKAM ARTANA He rnvualtun 4 1
7. Berge Collins 1995 literal 2 literal 9 literal 9 literal 6 literal 6 or T Definition APY A usage derivative n 1 pei 4 1 CLTPPTNG 4 Fig 8 After the insertion is confirmed freshly added LU ce appears in the text Append Ap
8. Ha main tree synset 2 9 9 Synset bg 2 2 form gt Pos 30 02767308 BCS 4 lemma xl feng x 05 literal 2 4 Cancel 18 24
9. euo ue 4 6 A 4 1 literal 1 literal 1 literal 1 Fig 6 Chooser s Edit dialog The example on Fig 6 shows the Edit dialog with the lemma of the component electronic feminine singular corrected from electronic masculine singular used as citation form to form the correct MWE lemma 16 Department of Computational Linguistics Institute for Bulgarian Language BAS 2012 electronic mail e The lemma of a MWE is also corrected in the Edit dialog To do that 1 after grouping the elements of the MWE into a single LU on how to do that see the relevant section above select it using the Left Right Arrow keys or a mouse pointer click 2 Select Edit from the Word menu 3 If needed correct the lemmas of the elements in the lemma field so that they
10. Ei Ha e nvMara ckaHnan 1 ad 1m El bg n 4 1 bg 1 bg 4 1 Ebbg 6 p cekTop 4 o 20 ous wo Ei z Fig 3 The Info view displaying the Hydra Tree view mode A detailed description of Hydra is available in the Hydra user manual
11. 24 BM ckaHnan na EC tree synset 2 1 E Synset bg 2 1 POS n lifeng 30 07966927 n BCS 4 literal 2 4 literal 1
12. POS x ILI eng 30 02644234y BCS 3 literal 1 literal 1 CLIPPING form literal 1 lemma literal 1 literal 1 c ce no ronan Berge Collins 1995 70007
13. 4 5 6 literal 1 UneHa Definition usage category member bg v 1 category domain bg n 5 hypernym bg n 4 Fig 5 Chooser s Word menu 3 2 5 1 Edit operations Edit affords modifications of the wordforms and lemmas of the LUs in the corpus In such a way misspellings typos and wrong lemmatisation are corrected directly in Chooser s interface in parallel with the annotation and updated immediately To edit a LU 1 press E
14. main tree synset 2 9 9 El Synset bg 2 2 Pos v x iu eng 30 02767308w BCS 4 literal 2 CLIPPING Kenway 1996 70 oT 18 24
15. Li main t t ree synse 2 1 Edit F4 a E Synset bg n 2 1 insert POS n ILI eng 30 07966927 n BCS 4 cb3fat Delete Del ond literal 2 npocnepute BPen literal 1 CLIPPING ce Ha 24 nu e BM
16. derivative n 1 CLTPPTNG Vj SN a 1 ce Fig 9 Delete dialog Useful tips In the corpus two or more individual tokens may be attached to each other without a separator as a result of wrong tokenisation for instance a number to a following word a special symbol 96 etc to a previous word etc Besides one word compounds may be typed as separate words and as a result constitute more than one token and vice versa multiword expressions may be typed as a single token 20 Department of Computational Linguistics Institute for Bulgarian Language BAS 2012 To split two or more tokens use the Insert and Edit functions 1 insert an additional token type the relevant wordform and lemma 2 edit the original token s wordform and lemma To merge two or more tokens use the Delete and Edit functions 1 delete any of the tokens 2 edit the other s lemma and wordform 3 2 5 3 Search Chooser supports simple and regular expression employing Python s regular expression syntax search functions over i wordforms ii lemmas Case sens
17. W literal 1 Definition usage category member bg v 1 category domain bg n 5 hypernym bg n 4 Ha TA M Fig 2 The Text view showing the currently selected LU coloured in red 3 1 3 Info view The Info view displays portions of the information available in the Wordnet database for a currently selected item in the List view The programme embeds the three display 8 Department of Computational Linguistics Insti
18. 1 Sentence Ends Phantom POS n ILI eng 30 07966927 n BCS 4 Ha literal 2 Current Word Current Sense CLIPPING Ce 24 BM
19. 2 Ha 3 Fig 7 The language unit ce of course is just added in dialog but not yet confirmed 18 Department of Computational Linguistics Institute for Bulgarian Language BAS 2012 anew ERE File Pass Word Stats End e ca
20. Edit operations e The grouping of non contiguous MWE components and word order eat variants of MWEs is performed in the same way e To ungroup a MWE left click on the individual words while holding the Ctrl key Ellipted components of MWE need to be restored in MWE lemma 11 Department of Computational Linguistics Institute for Bulgarian Language BAS 2012 Consider the following example red and white wines Select white wines as a MWE It receives the lemma white wine and is mapped to the corresponding synset white wine 1 In order to be able to map red to red wine 1 in the Wordnet database select red and expand the lemma to red wine the lemmas are represented in the brackets red red wine and white wines white wine 3 2 3 Annotation When a LU is selected in the text by means of the Left Right Arrow keys or mouse pointer click a list of the available annotation options is displayed in the List view Fig 2 The annotation of a LU is performed by 1 selecting the appropriate sense from the List view by i browsing the list with the Up Down Arrow keys ii pressing the number or letter key corresponding to the number letter preceding the relevant definition in the List view iii mouse pointer click on the selected item in the list 2 pressing Enter The respective definition in the List view is highlighted the LU in the Text view is coloured in the relevant colour e The users are ad
21. and BulSemCor a corpus of Bulgarian annotated with wordnet senses http dcl bas bg en corpora en html This manual covers the semantic annotation module The senses used in the semantic annotation are those encoded in the Bulgarian wordnet Therefore Chooser is coupled with the system for wordnet development Hydra Department of Computational Linguistics Institute for Bulgarian Language BAS 2012 and accesses the senses available in the wordnet database through Hydra s API Chooser s interface embeds a fully fledged visualisation of the wordnet synsets The changes made to the wordnet database are dynamically updated and displayed Any corrections or additions such as newly created synsets and synonyms are made accessible to Chooser immediately after they are performed In such a way the semantic annotation takes place simultaneously with the wordnet development The basic annotation functionalities implemented in Chooser are i fast and easy to perform annotation ii run time access to detailed information for the annotation candidates through the associated wordnet senses with all the info pertaining to the respective synsets synonyms explanatory definition semantic relations notes on usage grammar pragmatics etc iii identification of MWEs with contiguous and non contiguous constituents iv different strategies for corpus traversing v operations over the language units in the corpus edit insert
22. and delete functions vi a flexible search strategy allowing both simple and regular expression search according to wordform or lemma All the changes made to the corpus by means of the edit insert delete functions are immediately displayed and made accessible in the programme s interface to view and use in the annotation process The functionalities involving the manipulation of the corpus files minimise considerably the need for prior normalisation The system provides multiple user concurrent access Department of Computational Linguistics Institute for Bulgarian Language BAS 2012 Chooser is written in Python and tested under Linux and Windows 2 Getting started with Chooser 2 1 Installation For the initial setup of the programme confer the installation manual http dcl bas bg Tools Chooser Chooser InstallationManual pdf 2 2 File format Pre processing The files need to be tokenised and lemmatised Lemmatisation is essential because a language unit in the corpus is mapped to synsets in wordnet only if the lemma of the LU matches the lemma of a literal in one or more synsets File format The used format is a flat xml The root element is lt text gt The attribute current stores the last position where the file is viewed lt xml version 1 0 gt lt text current 5196 gt lt text gt The text is encoded as a list of xml elements labeled word The relevant information is stored in sepa
23. in the Pass 10 Department of Computational Linguistics Institute for Bulgarian Language BAS 2012 menu or traversing all the instances of the particular sense of the LU by checking the Current sense box Grouping words in multiword expressions To group two or more words in a multiword expression 1 select the individual words that form the MWE by left clicking on each of them one after the other while holding the Ctrl key In order for a MWE to be assigned an appropriate lemma and consequently to be identified in the Wordnet database the individual words must be grouped in the order in which they appear in the lemma of the corresponding literal in wordnet The individual words as part of a MWE may not be in their citation form The relevant wordform must be typed in the MWE s lemma field Consider the compound noun electronic mail The adjective electronic lemmatised in the same way agrees in gender with the feminine noun mail and is hence in the feminine singular In order for the LU in the corpus to be mapped to the corresponding synset 1 1 e mail 1 email 1 electronic mail 1 e mail 3 email 3 it has to be assigned the correct lemma To this end user needs to edit the MWE s lemma as shown in Fig 6 For details see the section on
24. 2 3 4 5 6 Fig 4 The Pass menu 3 2 5 Corpus searching and editing Chooser supports common search operations and two types of operations that involve making changes to the corpus content edit and add remove The options are selected 14 Department of Computational Linguistics Institute for Bulgarian Language BAS 2012 from the Word menu Fig 5 GHOSE porone noA File Pass Word Stats End paj Eind Find Again
25. Department of Computational Linguistics Institute for Bulgarian Language BAS 2012 Chooser User Manual Table of Contents TD PTE OCU CUOI P 1 2 Getting started with IGRI ET PR 3 FNIT 3 2 2 File WEA PEN EEE EEA 3 2 3 Starting Bree RR 4 Swag e ee aa ane e e E A ER AS S REESE EENEN EER NEER 4 3 1 User interface TEXT VIEW PT Sae 5 3 1 2 1 C P 7 VW TO M 8 3 2 Corpus annotation and me a pde kp DOR det 10 3 2 1 Loading and saving M Rosa 10 3 2 2 S lec ng LU T 10 aae a AREE eas 12 3 2 4 Traversing the COrpuS sseseesesessesersssessessresressessresressesseesresseesressessreseese 13 3 2 5 Corpus searching and or seo e bn PO SR IRR DU S Padus 14 3 2 51 Edit Operation S prits pe Qua 15 3 2 5 2 Adding deleting WONG ede 17 3 2 5 3 Search References 1 Introduction Chooser is an OS independent multi functional system for linguistic annotation adaptable to annotation schemata for different language levels It has been used in the creation of BulPosCor A POS annotated corpus of Bulgarian
26. NG Kenway 1996 70 main tree synset 2 9 9 Synset bg 2 2 S POS v x iu e
27. age BAS 2012 3 1 User interface On launching Chooser the programme s window appears on the screen It has a tripartite display area consisting of a Text view the main pane a List view bottom pane and an Info view right hand pane Different types of information are displayed in each field when a file is loaded Fig 1 Chooser on startup 3 1 1 Text view The Text view is the main pane where corpus files are loaded and displayed as well as Department of Computational Linguistics Institute for Bulgarian Language BAS 2012 where the corpus is traversed and the language units are selected for annotation Many of the operations performed on the language units take place in or are initiated from the Text view i identifying language units LUs A language unit is a word or a group of words that is treated as a single entity i e assigned a single sense Usually a token corresponds to a single word LU However sometimes the tokens need to be normalised For instance numbers may be attached to the following word e g a numbered list in the corpus For the sake of proper annotation they need to be split identified and subsequently annotated as separate LUs Chooser supports a run time edit functionality which allows such operations to be performed in parallel with the annotation ii grouping words in multiword expressions MWE Chooser provides a function for MWE identification and groupi
28. are in the appropriate form 4 Press ok The definition of the relevant synset appears in the List view e Ifthe constituents are ungrouped they will restore their original lemmas Useful tips To add a punctuation mark in the corpus text append the relevant symbol to the wordform in the form field and not in the lemma field of the word it should be attached to Capital letters marking the beginning of a sentence are also inserted in the form field and not in the lemma field For instance the lemma of a common noun such as bus is bus However if the word bus begins a sentence it should be capitalised in the form field If the lemma contains capital s they should be accounted for in the lemma field e g the lemma of a proper noun such as New York should be New York 3 2 5 2 Adding deleting words 17 Department of Computational Linguistics Institute for Bulgarian Language BAS 2012 Insert To insert a word 1 select Insert in the Word menu An Edit dialog appears on the screen 2 Type the wordform you want to add into the text and the respective lemma 3 Press ok to accept or cancel to reject the operation e The new word is inserted before the current LU BET new File Pass Word Stats End Ha
29. asing frequency of selection calculated on the basis of the previous choices made by the annotators The List view and the Info view are synchronised so that when an item in the list is selected the corresponding synset is displayed in the Info view To browse and select items in the List view and view the respective synsets in the Info view use any of the following methods i choose the relevant number or letter in front of a given definition in the List view using the number or letter keys ii browse the List view with the Up Down Arrow keys iii click on a particular definition in the List view The Text view in Fig 2 shows the currently selected LU coloured in red the Department of Computational Linguistics Institute for Bulgarian Language BAS 2012 List view displays the definitions of the available wordnet senses the one highlighted in red corresponds to the wordnet sense that has been assigned to the LU 2 1 1 business business sector Finally Hydra s main view for the synset 2 1 1 is displayed in the Info view chooser D9015 0776 blz EE File Pass Word Stats End
30. dit in the Word menu 15 Department of Computational Linguistics Institute for Bulgarian Language BAS 2012 A popup dialog with two fields form and lemma appears on the screen Fig 6 2 To edit the wordform retype correct the form in the form field 3 To edit the lemma retype correct the form in the lemma field 4 Press ok to accept the change or cancel to reject it DUBIO TD YTCHBIERE File Pass Word Stats End EA ral main HR synset 1 1 Synset bg v 1 1
31. http dcl bas bg Tools Hydra Hydra UserManual pdf Department of Computational Linguistics Institute for Bulgarian Language BAS 2012 3 2 Corpus annotation and editing 3 2 1 Loading and saving files To load a corpus file 1 use Open from the File menu A browsable dialog appears on the screen 2 browse the file you wish to load and select it Once a corpus file is opened it is displayed in the Text view The system saves the file automatically To explicitly perform the save operation use the Save button in the File menu To change the name and or location of the file use Save as 3 2 2 Selecting LUs To select traverse the words in the corpus file use either of the following actions 1 use the Left Right Arrow keys Unless another option is selected from the Pass menu the Arrow keys perform linear pass selecting one LU at a time to the left or to the right respectively 2 left click on a particular word in the Text view This command allows the user to select an arbitrary word in the text For the pass strategies confer the relevant section below Traversing the corpus When a language unit is selected the user is able to view the corresponding annotation candidates annotate and edit delete the LU insert other words with respect to it define search and traverse operations on it The possible pass options for a particular LU are traversing all the instances of the LU by checking Current word
32. itive search and search direction forward backward are selected from the menu To enable the search function 1 select the option Search from the Word menu A Search dialog pops up 2 type a word in the word field 3 to search for this particular form in the corpus check the form box 4 to search for all the wordforms of a lemma check the lemma box 5 check any of the other possible criteria search direction and case sensitivity 6 press ok to initiate the search or cancel to disable it 21 Department of Computational Linguistics Institute for Bulgarian Language BAS 2012 7 to resume the search after a match is found use the F3 Ctrl key The default search direction is forward Search backwards may be selected from the search dialog F3 Chooser home zarka chooser new portion4 19 01 07 chsr txt n File Pass Word Stats v End Ha Ha Ha A e CLIPPI
33. ng accounting for word order variations of the MWE components and for MWEs with non adjacent components For details see the section Grouping words in multiword expressions iii edit operations Chooser provides a functionality for editing the wordforms and lemmas of the LUs In such a way typos and wrong lemmatisation do not impede the annotation process For a description in more detail see Edit operations iv add remove operations This feature enables the insertion and deletion of tokens in the corpus files This is particularly useful in cases of wrong spelling of multiword expression as single words and vice versa For details see the Adding deleting section Department of Computational Linguistics Institute for Bulgarian Language BAS 2012 The status of the LUs with respect to annotation is denoted by means of different colour codes Several types of units are recognised a untraversed units units in the corpus that have not been traversed in the current session so no information on their annotation status is displayed b annotated units language units that are assigned a sense c non annotated units units that have not been assigned a sense d compound words multiword expressions 3 1 2 List view The List view is a standard list control that displays the definitions of the wordnet synsets suggested as annotation candidates for a LU in the corpus The synsets are listed according to decre
34. ng 30 02767308w BCS 4 ca Gt literal 2 literal 2 al 9 word form search backwards C lemma case sensitive Bl 9 6 6 Cancel ition Berge Collins 1995 H usage G derivative n 1 Ej TPPTNG 4 OU amp U Nh 7 Ha Ha Ha e craHe
35. pend is used to add a word to the end of a file To append a word follow the steps described for the Insert operation Delete To delete a LU 1 select Delete from the Word menu or press Delete on the keyboard A popup dialog asking to confirm or cancel the operation appears on the screen 19 Department of Computational Linguistics Institute for Bulgarian Language BAS 2012 2 Press either of the two options to confirm or cancel Chooser home zarka chooser new portion4 19 01 07 chsr txt eL X File Pass Word Stats End e D main ires synset Ha 2 9 9 Ha t y dy F y Tisryenus Synset bg v 2 2
36. rate attributes wordform w lemma 1 sense s annotator u time stamp t sentence end A special attribute is reserved for a parent ID that links the individual tokens of a compound p An annotated unit contains the following basic information Department of Computational Linguistics Institute for Bulgarian Language BAS 2012 word 1 p 1529023764 s 1100001720 t 1298483182 w 3aToBa gt Minimum restrictions are imposed on the extension of the specified file format so that it may permit addition of flat and or hierarchical annotation schemata without affecting the current one thus enabling other levels of annotation 2 3 Starting Chooser To launch Chooser using a command line run the following command python chooser py The examples below show how to run the programme in a Linux environment provided that it is located in the local directory home boby chooser 1 from the local directory chooser where the executable file chooser py is stored boby tornado chooser python chooser py 2 using the full path to the executable file boby tornado python home boby chooser chooser py 3 using a relative path boby tornado python chooser chooser py 3 Overview The following sections give a description of Chooser s user interface and functionalities Department of Computational Linguistics Institute for Bulgarian Langu
37. rses those LUs mapped to literals that have been removed from the Wordnet database after the LU has been annotated v sentence endings Sentence End This pass option has been defined for the purposes of manual validation of sentence splitting The above options can be combined with the Current word pass thus enabling the 13 Department of Computational Linguistics Institute for Bulgarian Language BAS 2012 traversing of the instances of a current LU that meet the selected option A user may also traverse the instances of the particular sense assigned to the current word Traversing is performed using the cursor movement Left Right Arrow keys of the keyboard Pass Word Stats End CT eT a xopa All Cp Not chosen ca 2 1 Ambiguous a Synset bg n 2
38. tute for Bulgarian Language BAS 2012 views of the system for wordnet development Hydra right pane Main view Tree view and Synset view The synchronisation between the List view and the Info view enables the annotators to make their choices based on a detailed inspection of all the available information associated with a synset HEW BOTHOTTH File Pass Word Stats End 4 DU nain ume B mm Relation Path Unit Path M
39. vised to consult the additional information in the Info eat view other synonyms in the synset usage examples the relations pertaining to the synset before performing the annotation The sense distinctions in wordnet may be very fine grained which sometimes necessitates close inspection of similar senses 12 Department of Computational Linguistics Institute for Bulgarian Language BAS 2012 e There is no specific operation to override the annotation operation Instead the user needs to select another sense e Changes made to the Wordnet database in Hydra are updated immediately in Chooser To view the changes concerning a currently selected LU refresh the List view by first disselecting and then re selecting the particular LU This is performed by jumping to a previous or next word using the Left Right Arrow keys or by clicking on any other word and then clicking again on the LU under consideration 3 2 4 Traversing the corpus Chooser supports several strategies for traversing the corpus Fig 4 i The default option is a linear pass of all the LUs option All To choose another strategy check the respective option in the Pass menu ii unannotated LUs option Not Chosen traverses the LUs that have not been assigned a sense yet iii ambiguous LUs option Ambiguous passes the LUs to which more than one wordnet sense corresponds iv LUs removed from the Wordnet database Phantom trave
40. z POS v iti eng 30 02767308 BCs 4 literal 2 CLIPPING Kenway 1996 7 oT Ca B ca Ca literal 2 literal 9 B 18 24 literal 9 literal 6 c o literal 6 Definition APY
Download Pdf Manuals
Related Search