机器翻译中基于语法、语义知识库的汉语词义消歧策略

♣100871WSDAStudyofChineseWordSenseDisambiguationinMTBasedonGrammatical&SemanticKnowledge-basesWang,HuiInstituteofComputationalLinguistics,PekingUniversity,Beijing100871,ChinaAbstractWordsensedisambiguation(WSD)playsanimportantroleinMachineTranslationandmanyotherareasofnaturallanguageprocessing.TheresearchonWSDhasgreattheoreticalandpracticalsignificance.ThemainworkinthispaperistostudywhatkindofknowledgeisusefulforWSDinsystem,andestablishamulti-levelWSDmodelbasedonsyntagmaticfeaturesandsemanticinformation,whichcanbeusedtodisambiguatewordsenseinMandarinChineseeffectively.ThemodelmakesfulluseoftheGrammaticalKnowledge-baseofContemporaryChineseasoneofitsmainmachine-readabledictionary(MRD),whichcanproviderichgrammaticalinformationfordisambiguationsuchasChineselexicon,parts-of-speech(POS)andsyntaxfunction.AnotherresourceofthemodelistheSemanticDictionaryofContemporaryChinese,whichprovidesathesaurusandsemanticcollocationinformationof68,000Chinesewords.TheresultsofthisstudyindicatethatthetwoMRDresourcesareeffectiveforwordsensedisambiguationinMTandarelikelytobeimportantforgeneralChineseNLP.Keywords:WordSenseDisambiguation,Chinese-EnglishMachineTranslation,GrammaticalKnowledge,SemanticDictionary♣973G1998030507-411.translationchoiceWordsensedisambiguation,WSD50[1,2]2080LeskLukOxfordAdvancedLearner’sDictionary[3,4]DaganGale[5,6]VoorheesResnikWordNet[7,8]Yarowsky1994Roget’sInternationalThesaurus[9]GaleChurch[10]2090[11]5LAM1997[12]45.5%1999[13]52.13%[14,15,16]senseChoueka1983[17]2[18][19]796%12[1]Thiswillbesubsidizedbythestate.[2]Livingallowanceswerequicklyhandedouttothepeopleinthestrickenarea.20513398993223.4%2001998122744219659227%1/5735,2007[20]♣♣3nanbncnenfngnh“”“”“”nena[1]accountantship[2]accountant[1][2]getready4[3][4]bribeinvite5.15973Sem_DictWordnet5220026.82think2miss3present1sprout5.2vegetablegreensdishcourse973Sem_Dict61big1|heavy2see2|*watch2|read635004319392694.8%401197.5%30.1%0030.05%321074.7%374586.8%329976.7%96422.5%3515.8%1382.2%70%575——1998167building5671188floor67231·7.81[1]Ide,Nancy;JeanVéronis.IntroductiontotheSpecialIssueonWordSenseDisambiguation:TheStateoftheArt,ComputationalLinguistics.Vol.24,No.1,1998.pp1-40[2]..:.2000.[3]Lesk,Michal.Automaticsensedisambiguation:Howtotellapinefromanicecreamcone.In:AssociationforComputingMachinery,eds.The1986SIGDOCConference.NewYark,ACM.1986.pp24-269[4]Luk,AlphaK.StatisticalSenseDisambiguationwithRelativelySmallCorporaUsingDictionaryDefinitions.In:ACLeds.The33rdAnnualMeetingofACL,Cambridge,Massachusetts.1995.pp181-188[5]Dagan,Ido,AlonItai,andShaulMarkovitch.TwoLanguagesAreMoreInformativeThanOne.In:The29thAnnualMeetingofAssociationforComputationalLinguistics,Berkeley,CA:ACL,1991.pp130-137[6]Gale,WilliamA,KennethW.Church,andDavidYarowsky.AMethodforDisambiguationWordSensesinaLargeCorpus.ComputerandtheHumanities.(26)1993.pp415-439[7]Towell,Geoffrey;EllenM.Voorhees.DisambiguatingHighlyAmbiguousWords,ComputationalLinguistics,Vol.24,No.1,1998.pp125-145[8]Resnik,Philip.SelectionandInformation:AClass-BasedApproachtoLexicalRelation.[Ph.D.Dissertation],USA:UniversityofPennsylvania.1993.pp23-54[9]Yarowsky,David.DecisionListsforLexicalAmbiguityResolution:ApplicationtoAccentRestorationinSpanishandFrench.In:ACLeds.The32ndAnnualMeetingofAssociationforComputationalLinguistics.LasCruces,NM:ACL.1994.pp88-95[10]Gale,WilliamA,KennethW.Church,andDavidYarowsky.Usingbilingualmaterialstodevelopwordsensedisambiguationmethods.In:TheInternationalConferenceonTheoreticalandMethodologicalIssuesinMachineTranslation,1992.pp101-112[11].[]..1993[12]LAMSZE-SING,KAM-FAIWONG,andVINCENTLUM.LSD-C–A.linguistic-basedword-sensedisambiguationalgorithmforChinese.ComputerProcessingofOrientalLanguages,Vol.10,No.4,1997,pp409-422.[]..1999[14]..[a].[c].:.1995[15].[].,1998[16]YangXiaofeng,LiTangqiu.AStudyofSemanticDisambiguationBasedonHowNet.ComputationalLinguisticsandChineseLanguageProcessing.Vol.7,No.1,2002,pp47-78[17]Choueka,Y.andS.Lusignan,AConnectionistSchemeforModelingWordSenseDisambiguation.CognitionandBrainTheory.6(1)1983,pp.89-120.[18],,,...1998.[19],,.[a].[c]1998..1998.pp361-367[20],.[a].[c]1998.:.1998.pp852-86310