欢迎来到天天文库
浏览记录
ID:39756231
大小:67.76 KB
页数:8页
时间:2019-07-10
《Information Retrieval based on Paraphrase》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、InformationRetrievalbasedonParaphrasePeterWallisDept.ofComputerScience,RMIT,Melbourne,Australia,(peter@cs.rmit.edu.au)TextRetrievalsystemsbasedonrankingusesim-2InformationRetrievalilarityasanapproximationtorelevance.Mostofthesesystemsignorewordmeaning.Weas-Examplesofinforma
2、tionretrieval(IR),ortextretrievalsumethatsomemeasureofparaphrasewouldsystemscanbefoundatmostlibrariesintheformofon-beabettersimilaritymeasure.Wedevelopalinecatalogsandcitationindexes.SuchsystemsusuallyconceptofparaphrasebasedonMeaning-Textmaketheirkey-wordbasednatureexplici
3、tbyaskingforTheoryandimplementanapproximationtothesomebooleancombinationoftermswhichmustappearinidealusingtheLongmanDictionaryofContem-atitleorsubject®eld.MorerecentdevelopmentsallowporaryEnglish(LDOCE).Theperformanceoftheusertoentersearchtermsintheformofanaturallan-thenews
4、ystemisassessedusingrecallandpre-guagequery.Forexampleausercouldmakethefollowingcisionaveragesontwostandardcollections.Werequest:discusstheresultsandconcludethatwecouldªI'minterestedincommunicationbetweendis-improveperformanceifonlytherestrictedvo-jointprocesses.ºcabularyof
5、thedictionaryentrieshadapartic-ularproperty.WethenproposeatechniqueforThesesystemsextractkeywordsfromthequeryandthenmodifyingtheentriesusingstatisticalmethods®nddocumentswhichcontainasigni®cantproportionofrecentlyusedininformationretrieval.theextractedwords.Forinstancethefo
6、llowingsetmightbeextractedfromtheabovequery:Key-words:LexicalSemantics,paraphrase,Informa-(ªinterestºªcommunicationºªbetweenºªdis-tionRetrieval,Meaning-TextTheory,semanticprimitives,jointºªprocessesº)LDOCEThedocumentsarealsotreatedassetsofterms,andrankedagainstthequeryusing
7、asimilaritymeasure.Thedocu-mentsareshowntotheuserbest®rst.Theuserlooks1Introductionthroughtherankedlistuntilthey®ndwhattheywant,ordecidetotryadifferentquery.Themosteffectivesim-Generalpurposequestion-answeringisanon-trivialtaskilaritymeasurestodatetreatdocumentsandqueriesas
8、beyondthereachofcurrentcomputerscience,buttextcollectionsofindependentfeatures.The
此文档下载收益归作者所有