欢迎来到天天文库
浏览记录
ID:37659428
大小:220.16 KB
页数:8页
时间:2019-05-27
《The Use of NLP Techniques in CLIR》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、TheUseofNLPTechniquesinCLIRBarbelRipplingerIAIMartin-Luther-Str.1466111Saarbrucken,Germanybabs@iai.uni-sb.deAbstractTheapplicationofnlptechniquestoimprovetheresultsofinformationretrievalisstillconsideredasacontroversialissue,whereasnlptechniquesincross-languageinformationre-t
2、rieval(clir)arealreadywellestablished,atleastforthetranslationpart.Inthispaper,theclircomponent-Mpro-IR-isrepresentedwhichhasbeendevelopedasthecoremod-uleofamultilingualinformationsysteminalegaldomain.Thiscomponentusesnotonlythelexicalbaseformforindexingbut,inadditionderivati
3、onalinformationandforGermaninformationaboutthedecompositionofcompounds.Thisinformationisprovidedbyaso-phisticatedmorpho-syntacticanalyser(mpro)andisexploitednotonlyforquerytranslationbutalsoforqueryexpansionaswellasthesearchandthedocumentranking.Mpro-IRisdevelopedforadomainwi
4、thafocusonrecall.Theobjectiveoftheclefevaluationwastoassessthislinguisticbasedretrievalapproachinanunrestricteddomain.Thefocusoftheinvestigationwasonhowderivationanddecompositioncancontributetoimprovetherecall.Duetotimerestrictionwecouldonlysubmitonecross-languagerunwithEngli
5、shasquerylanguageusingthephrasesearchcomponentofMpro-IRwhichisrestrictedtosearchonlyforonetranslation,andwherethequeriedtermshavetooccurinthesamesentence.Theserestrictionsresultinasmallnumberofretrieveddocuments,andthusinalowperformance.1IntroductionTheMpro-IRsystemisaclirsys
6、tembasedonquerytranslationandfocusesratheronabetterrecallthanonabalancedrecallandprecision.Toimprovetherecall,thesystemtriestotakeadvantageofasophisticatedlinguisticprocessingcomponentwhoseresultsareusedinthemono-lingualretrievalmodules.Basedontheoutputofamorpho-syntacticanal
7、ysiswhichprovidesthefullrangeofmorphologicalinformation,notonlyin ectionwhichwouldcorrespondtothepowerofastemmer(Porterstemmer)butalsoderviationalanddecompositionofcompoundnounsisexploited.Thisinformationisusedfortheindexing,queryexpansion,searchanddocumentrank-ing.Thetransla
8、tioncomponenttakesadvantageoftheprovidedpart-of-speechaswellasofthes
此文档下载收益归作者所有