[PDF] English-Arabic dictionary for translators

View - Download English-Arabic dictionary for translators

Previous PDF

Next PDF

English-ArabicDictionaryforTranslators

SabriElkatebandBillBlack

DepartmentofComputation

UMIST,POBox88,ManchesterM601QD

Abstract

Wepresentadesignofacomputerizedbilin-

PrincetonLexicalreference

database,intheform andMiller,1990).Ourgoalisnotonlytoaddthe enhancethevalueofthedesign.Ourdesignwill system).

1Introduction

Thenotionofwhatadictionaryishasunder-

goneadramaticchangewithdevelopmentsin computationallexicographyandincomputa- ofwordandsenserelationsmakespossiblead- hocquerieswithwhichtheuser(orthenatu- rallanguageprocessingsystem)canndsyn- formationaboutaword,anditspossibletrans- lationsinotherlanguages.Equally,declara- terminologicalknowledgebasescanenablethe searchforwordsrealizingaconcept,senseor lexeme,e.g.asproposedin(SierraandMc-

Naught,2000).Wedescribetheconceptualde-

signofaterminologybase,basedona'back- bone'derivedfromarelationalmodelofthe

WordNet(Denness,1996).Thedatamodelis

extendedbeyondanArabicreplicationofthe

2Approach

Wemainlyaimatdevelopinganexpandable,

browsableandsearchablecomputer-basedlex- calterminologyinArabic.Besidesthedesire thatthisdictionarycanmeettheneedsofvar- iousgroupsofusers,itismainlyintendedfor

Arabtranslatorswhoseektohavesatisfactory

informationaboutawordandanadequaterep- resentationofitsform,structureandsenses.

Oneoftheinterestingmodesoforganisation

ofthisconceptualdictionaryisthatindexing ofthesetsofwordsreplacesalphabeticalor- der.Thesetofwordsorsynonymsetsknown aconcept.Aword-conceptrelationsupports threequeryt ypeswithbothwordandconcept indexed:

Sensesofword

Wordsexpressingaconcept

Synonymsofaword.

3WordNetModel

WordNetisamonolingualEnglishLanguageon

Thislexiconisorganisedintermsofwordmean-

ingratherthanwordforms.WordNetorganises thelexiconbysemanticrelationsonthebasis ofsynonymy.Synonymyisasemanticrelation betweentwowordswithdierentformsandsim- ilarmeanings.Table1,extractedfromthedis- tributionoftheWordNetinPrologform,and editedtableformat,showshowthismaybe viewedintabularform.Wordnetrepresents sensesasthecollectionofwordshavingthat sense-asetofsynonyms,orasynset.The senseisnomorethanthatsetofwordsthat denoteit,butinadatabaseitisconvenientto representeachsuchsetwithauniqueidentier, showninthetableasSynset No.

SynstNoWord#WordCatS#

1000017421entityni

1000031351organismn1

1034475081plantn1

1050548181plantn2

1069624512

oran1

2012412921plantv1

Table1:Word-senserelationsderivedfrom

WordNet

4EuroWordNetModel

EuroWordNetisamultilingualdatabasewith

variouswordnetsforseveralEuropeanlan- guages(Dutch,Italian,Spanish,German,

French,CzechandEstonian).Thewordnets

adoptthesamestructureimplementedinAmer- icanwordnetforEnglish(PrincetonWordNet (Miller,1990)).Thereisauniquelanguage- ipantwordnet,andeachwordnetislinkedtoan tonwordnet.Thisindexmakesthelanguages interconnected,i.e.searchcangofromthe wordsinonelanguagetosimilarwordsinany otherlanguage.EuroWordNetapproachaims atbuildingthewordnetsmainlyfromexisting resources.Eachsiteintheprojectcanbuild internationalprojects.

5Word-senserelation

WordNetaimsatorganisingthelexiconby

semanticrelationsonthebasisofsynonymy.

Synonymyisasemanticrelationbetweentwo

wordswithdierentformsandsimilarmean- ings.Thatistosay,thelexiconisorganised intermsofwordmeaningratherthanword forms.ThismodeoforganisationmakesWord-

Asabasicprinciple,meaningsinWordNetare

representedbysynonymsetsorsynsets.A synsetisthesetofwordsthatdenotethesame howwordformscanbeusedtoexpressword meanings,andawordformispolysemousora synonymtoanotherwordform.F1expresses wordmeaningM1.F1andF2aresynonyms astheyrepresenttwoentriesinthesamerow.

F2ispolysemousbecauseithastwoentriesin

thesamecolumn.Thelexicalmatrixisbased pingbetweenformsandmeaningsi.e.itisrep- resentedthroughtheactualmappingbetween writtenwordsandsynsets.

Figure1:Lexicalmatrix

6Senserelations

Thethesauralrelationofhyponymyisread-

ilypicturedintherelationalmodel,asatran- sitiverelationfromsynsettosynset.Thus thehyponymyrelationbetweenthesynsetsen- tity,organismandorganism,plant/ oraisrep- instancesofothersenserelationsinthesame way,e.g.meronymyandantonymy.Hyponymy

Synset1Synset2

100001742105054818

Table2:Representinghyponymyinatable

realtionsaremainlyusedtosupportbrowsing relatedsenses,asinFigure2.

7AddingdataforArabicand/or

otherlanguages

Thereareseveralalternativewaysofaddinga

secondandsubsequentlanguagetoasenseenu- merativelexicon,some,butnotallofwhichare discussedin(Vossenetal.,1997).Tomakethe

Figure2:Treeviewershowingpartofhy-

ponymyrelations language(s).

Threepossibleextensionstothedatamodel

suggestthemselves: (a)Changethenameofthewordcolumnto

English,andtoaddnewcolumnsforArabic,

French,etc.

SynsetNoW#EngArabicCatS#

1000017421entitywuju:dni

1000031351organismka:inn1

1034475081plantmasna'n1

1050548181plantnaba:tn2

1069624512

oranaba:tn1

2012412921plantzara'av1

Table3:addingacolumntoWNScontaining

Arabic

(b)Addanewcolumninwhichacodeforthe languageofthetablerowisplaced. (c)ReproduceWN

Stableforeachlanguage.

Anadvantageofaddinganewtableisto

makeanewindependentconceptualdictionary forthesecondlanguage,whereasinsertinga newcolumnismoreeconomicalonspace.

TheArabicequivalentoftheWN

Stableis

createdtoincluderootandpatternofeachword asadditionalcolumnsaswellasanylanguage specicfeatures.Thisallowsthesystemtosup- portqueriesbasedonwords,rootsorpatterns, aswellasviasynonymy,hyponymyandthe

SynsetNOW#WordLangCatS#

1034475081plantEnglishn1

1034475081masna'Arabicn1

1069624512

oraEnglishn1

1069624512naba:tArabicn1

1050548181plantEnglishn2

1050548181naba:tArabicn2

2012412921plantEnglishv1

2012412921zara'aArabicv1

Table4:addingacolumntoWNScontaining

languageidentication

SynsetNoWordCatS#RootPattern

103447508masna'n1sn'maf'al

105054818naba:tn2nbtfa'a:l

106962451naba:tn1nbtfa'a:l

201241292zara'av1zr'fa'ala

Table5:ArabicWNStable

otherWordnetrelations,andbyEnglishtrans- lation.

8Queryingtranslations

Ineitheroftheabovedatabaseschemata,a

requiringajoinoftwotables,intheother asingletablequery.WehavejoinedWN S tablewithWN

SArabictoshowtheEnglish

word,Arabictranslationandthepartofspeech WN

Gtabletoaddtheglossesandexamples

columntothequery.Theuserortheintended userwhoissaidtobethetranslatororthelan- inonelanguagethatcanexplainthesenseof thewordforbothlanguages.

9Updatingtranslations

Inaninvirnmentofanopen-endedsystemfor

tionsmotivatedbythesecondorsubsequent language.IncaseswhereArabicwordshaveno canbeapplied:

1-AllocatenewSynsetnumber.2-Link

Synsetnumbertothenearesthypernymby

addingrowinWN

HYPtable.3-Addrow

SynsetNoWordArabiccatGloss

inWN

Stable.4-AddEnglishglossinthe

WN

GLOSStable.

10ArabicMorphology

Arabicishighlyin

ectionallanguageandcan expanditsvocabularyusingaframeworkthat islatentinthecreativeuseofrootsandpat- terns.Phonemesandlettersarethecompo- nentsoftheArabicword.Thesecomponents aremappedintoapredeterminedformknown ifsubjectedtocertainpatterns:

ArabicEnglishPOSPattern

katabawritevfa'ala kita:bbookn'a:l kita:bahwritingn'a:lah ka:tibwriternfa:'il ka:tibclerknfa:'il ka:tabacorrespondvfa:'ala maktabocenmaf'al maktabahlibrarynmaf'alah muka:tabahcorrespondencenmufa:'alah iktita:bsubscriptionnifti'a:l kita:biclericaladj'a:li ktb

Itisworthmentioningthattables7,8and

9areforillustrationpurposesonlyanddonot

formapartofthedatabase.

Consonantsremainunchangeableandarenot

subjectedtoanyconversionwhenderivinga newword,buttheyarederivedfromandbuilt upon.

GroupingthesetsofArabicwordsaccording

isusedinourdesigntoquerywordsfromagive relatedsensesTable8showsdierentnouns coinedaccordingtotheArabicpatterntaf'i:l whichreferstoaprocessoraprogressofsome activity:

ArabicwordEnglishword

tasi:sorigination tanzi:morganization ta'li:meducation tajmi:'assembly takri:rrening tashhi:mlubrication ktb

ThisfeatureofArabicisalsousedtoquery

Arabicwordsthatareformedaccordingtoa

tocoinnewArabictermsaccordingly.

NativeArabicspeakerscaneasilytellthepat-

ternofalmostanygivenword,butalsorecall thewordscoinedaccordingtothatpattern.In thedatawehavecollected,therearelistsof wordsthataresearchedaccordingtogivenpat- toaparticularverb.Therefore,infrontofev- erynouninalistofaparticularpatternthere isacorrespondingverbderivedfromthesame rootofthatnoun.Itisimportanttonotethat particularpattern.Forexample,thenounpat- pattern'tafa:'ala':

11Userinterfaceandtheediting

functionality areorarenotexpectedtohaveArabicenabled versionofWindowsalreadyinstalled,provide

JavatosupportthosewithnoAEWinstalled

ArabicnounMeaningArabicverb

taba:dulexchangetaba:dala taba:'udseparationtaba:'ada tata:bu'successiontata:ba'a taja:dhubattractiontaja:dhaba taqa:rubapproachtaqa:raba taka:thurmultiplicationtaka:thara tama:thulsimilaritytama:thala tana:furalienationtana:fara tana:fuscompetitiontana:fasa

Table9:ndingnounverbrelationthrougha

givenroot ersofArabicwhoarewellingtousesystems andkeyboardsoftheirownlanguages.Forthis purposeavirtualkeyboardiscreated,shownin

Figure3.

Theinterfaceusesinformationdisplaysthat

treateachelementasadistinctobjectrather thanatextportion.Allupdatesaremaderel- terfacehasaqueryfacility.Thisallowswords tobeenteredineitherEnglishorArabic(and additionallyArabicrootsandpatterns),anda numberofalternativequeriesinvoked.Since responsetoaqueryistodisplayawordsense matrix,showninFigure4.

Figure4:WordSenseMatrix

Thematrixallowscells,rowsorcolumnsto

beselected.Selectingacellorarowmakes aparticularsynsetcurrent.Thisinturnen- ablesthetree-viewofahierarchyofwordsto begeneratedandfocusedaroundtheselected sense.Atthesametime,theglossandexam- displayed.Whenasenseisselectedeitherfrom thewordsensematrixorfromthetreeviewer

Arabictranslationofthesenseaswellasroot

andpatternoftheArabicwordareretrieved.

12Conclusion

ThedesignandimplementaionoftheEnglish-

oftheWordNet.Thedesignenablesustostore lationsthanthoseintheoriginalwordnet.We lowtheconceptualdictionarytobeaugmented withmorphologicalanalysisandgeneration.

References

R.BeckwithandG.A.Miller.1990.Implement-

ingalexicalnetwork.InternationalJournal ofLexicography3,pages302{312.

S.M.Denness.1996.Adesignofastructure

foramultilingualconceptualdictionary.Msc dissertation,UMIST,Manchester,UK.

S.Elkatib.1991.Translatingscienticand

technicalinformationfromenglishintoara- bic.Master'sthesis,UniversityofSalford,

Manchester,UK.

C.Holes.1995.ModernArabic.Longman,

London,UK.

G.A.Miller.1990.Nounsinwordnet:Alexical

inheritancesystem.InternationalJournalof

Lexicography3,4.

G.SierraandJ.McNaught.2000.Designof

anonomasiologicalsearchsystem:Aconcept- orientedtoolforterminology.Terminology,

6(1):1{34.

P.Vossen,P.D?ez-Orzas,andW.Peters.1997.

Themultilingualdesignofeurowordnet.In

P.Vossen,N.Calzolari,G.Adriaens,A.San-

lippo,andY.Wilks,editors,Proceedingsof theACL/EACL-97workshopAutomaticIn- formationExtractionandBuildingofLexical

SemanticResourcesforNLPApplications,

Madrid,July12th,1997.

Figure3:ArabicVirtualKeyboard

Figure5:User'sinterface

quotesdbs_dbs22.pdfusesText_28

[PDF] Le guide de l 'allaitement maternel - Manger Bouger

[PDF] La Leche League France - Allaitement et maternage

[PDF] Hérédité monogénique

[PDF] Allemand CP : Se présenter - dsden 89

[PDF] Allemand CM2 - Académie en ligne

[PDF] Allemand-Français - pierregloria

[PDF] Allergie ? la benzocaïne - e-Allergo

[PDF] alu aeronautique - MDS métaux détails services

[PDF] outils vedettes allô prof pour le 1er cycle du secondaire - Allo prof

[PDF] outils vedettes allô prof - Allo prof

[PDF] En fin de Contrat d 'Avenir Si vous êtes sans emploi - SUD Education

[PDF] Guide des prestations 2017 - Caf

[PDF] Je travaille au Grand-Duché de Luxembourg - OGBL

[PDF] Demande d 'aide au logementpdf - MSA

[PDF] CONSULAT GÉNÉRAL DE FRANCE À TUNIS DEMANDE DE VISA