欢迎来到天天文库
浏览记录
ID:40365023
大小:902.82 KB
页数:12页
时间:2019-08-01
《Unsupervised Sentence Enhancement for Automatic Summarization 》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、UnsupervisedSentenceEnhancementforAutomaticSummarizationJackieChiKitCheungGeraldPennUniversityofTorontoUniversityofToronto10King’sCollegeRd.,Room330210King’sCollegeRd.,Room3302Toronto,ON,CanadaM5S3G4Toronto,ON,CanadaM5S3G4jcheung@cs.toronto.edugpenn@cs.toronto.eduAbstractInput:BilMarFoodsCo.,ame
2、atprocessorownedbySaraLee,announcedarecallofWepresentsentenceenhancementasacertainlotsofhotdogsandpackagedmeat.noveltechniquefortext-to-textgenera-Input:TheoutbreakledtotherecallonTues-tioninabstractivesummarization.Com-dayof15millionpoundsofhotdogsandcoldparedtoextractionorpreviousapproachescut
3、sproducedattheBilMarFoodsplant.tosentencefusion,sentenceenhancementOutput:TheoutbreakledtotherecallonTues-increasestherangeofpossiblesummarydayoflotsofhotdogsandpackagedmeatssentencesbyallowingthecombinationofproducedattheBilMarFoodsplant.dependencysubtreesfromanysentenceFigure1:Anexampleoffusin
4、gtwoinputsen-fromthesourcetext.Ourexperimentsin-tencesintoanoutputsentence.Thesectionsofthedicatethatourapproachyieldssummaryinputsentencesthatareretainedintheoutputaresentencesthatarecompetitivewithasen-showninbold.tencefusionbaselineintermsofcon-tentquality,butbetterintermsofgram-maticality,an
5、dthatthebenefitofsen-tenceenhancementreliescruciallyonaneventcoreferenceresolutionalgorithmus-manticanalysisbeyond,say,adependencyparseringdistributionalsemantics.Wealsoandlexicalsemanticresources.considerhowtext-to-textgenerationap-Theoveralltrajectorypursuedinthefieldcanproachestosummarizationca
6、nbeex-becharacterizedasamoveawayfromlocalcon-tendedbeyondthesourcetextbyexam-textsrelyingheavilyontheoriginalsourcetextto-ininghowhumansummarywritersincor-wardsmoreglobalcontextsinvolvingreformula-poratesource-text-externalelementsintotionofthetext.Whereassentenceextractionandtheirsummarysentenc
7、es.sentencecompression(KnightandMarcu,2000,1Introductionforexample)involvetakingonesentenceandper-hapsremovingpartsofit,traditionalsentencefu-Sentencefusionisthetechniqueofmergingsev-sioninvolvesreformulatingasmallnumberofre
此文档下载收益归作者所有