wsdm-2011-final.pdf

wsdm-2011-final.pdf

ID:33927627

大小:248.36 KB

页数:8页

时间:2019-02-28

wsdm-2011-final.pdf_第1页
wsdm-2011-final.pdf_第2页
wsdm-2011-final.pdf_第3页
wsdm-2011-final.pdf_第4页
wsdm-2011-final.pdf_第5页
资源描述:

《wsdm-2011-final.pdf》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、ClusteringProductFeaturesforOpinionMining†‡††ZhongwuZhaiBingLiuHuaXuPeifaJia†StateKeyLabofIntelligentTech.&Sys.,TsinghuaNationalLabforInfo.Sci.andTech.,Dept.ofComp.Sci.&Tech.,TsinghuaUniversityzhaizhongwu@gmail.com‡Dept.ofComp.Sci.,UniversityofIllinoisatChicagoliub@cs.uic.eduABS

2、TRACTwordsorphrasesthatexpressthesamefeature,featureexpressions.Forexample,“picture”and“photo”arefeatureInsentimentanalysisofproductreviews,oneimportantproblemexpressionsreferringtothesamefeatureofcameras.Inthispaper,istoproduceasummaryofopinionsbasedonproductweassumethatallfeat

3、ureexpressionshavebeenidentifiedbyanfeatures/attributes(alsocalledaspects).However,forthesameexistingalgorithm.Therearemanysuchalgorithms[17-20,29,feature,peoplecanexpressitwithmanydifferentwordsor35,38].phrases.Toproduceausefulsummary,thesewordsandphrases,whicharedomainsynonyms

4、,needtobegroupedunderthesameGroupingfeatureexpressions,whicharedomainsynonyms,isfeaturegroup.Althoughseveralmethodshavebeenproposedtocriticalforeffectiveopinionsummary[26].Sincethereareextractproductfeaturesfromreviews,limitedworkhasbeendonetypicallyhundredsoffeatureexpressionst

5、hatcanbediscoveredonclusteringorgroupingofsynonymfeatures.Thispaperfocusesfromtextforanopinionminingapplication,it’sverytime-onthistask.Classicmethodsforsolvingthisproblemarebasedconsumingandtediousforhumanuserstogroupthemintofeatureonunsupervisedlearningusingsomeformsofdistribu

6、tionalcategories.Someautomatedassistanceisneeded.Unsupervisedsimilarity.However,wefoundthatthesemethodsdonotdowell.learningorclusteringisthenaturaltechniqueforsolvingtheWethenmodelitasasemi-supervisedlearningproblem.Lexicalproblem.Thesimilaritymeasuresusedinclusteringareusuallyc

7、haracteristicsoftheproblemareexploitedtoautomaticallybasedonsomeformofdistributionalsimilarity[6,10,22,24,32,identifysomelabeledexamples.Empiricalevaluationshowsthat34,37].Recentworkalsousedtopicmodeling[12,40].However,theproposedmethodoutperformsexistingstate-of-the-artweshowth

8、atthesemethodsdonotperformwell.Eventhelatestmet

当前文档最多预览五页,下载文档查看全文

此文档下载收益归作者所有

当前文档最多预览五页,下载文档查看全文
温馨提示:
1. 部分包含数学公式或PPT动画的文件,查看预览时可能会显示错乱或异常,文件下载后无此问题,请放心下载。
2. 本文档由用户上传,版权归属用户,天天文库负责整理代发布。如果您对本文档版权有争议请及时联系客服。
3. 下载前请仔细阅读文档内容,确认文档内容符合您的需求后进行下载,若出现内容与标题不符可向本站投诉处理。
4. 下载文档时可能由于网络波动等原因无法下载或下载错误,付费完成后未能成功下载的用户请联系客服处理。
相关文章
更多
相关标签