欢迎来到天天文库
浏览记录
ID:34598926
大小:11.15 MB
页数:62页
时间:2019-03-08
《基于微博的情感-分析关键技术-研究》由会员上传分享,免费在线阅读,更多相关内容在教育资源-天天文库。
1、万方数据II万方数据AbstractResearchonSentimentAnalysisandOpinionMiningofWeiboCandidate:JuYu,Supervisor:Prof.QiaoWANGSchoolofInformationScienceandEngineering,SoutheastUniversity,ChinaAftersteppingintotheeraofmobileinternet,peoplearebecomingmorewillingtosharetheiropinionsandlifeexperienceswithothersthroughint
2、ernet.Meanwhile,itresultsinanexplosiveaccumulationofsubjectiveinformationonsocialnetworkslikeWeibo.Andinthiscontext,thesentimentanalysisofWeibokeepsbothgreatsocialandcommercialvalues;asforgovernors,itCanfacilitatetheirunderstandingtowardspublicopinions;whileformerchants,ithelpsthemtofollowupconsume
3、rs’attitudes.HereIaimtodevelopasetofmethodsforsentimentanalysisandopinionminingofWeibo.Myworkmainlycomprisesfourparts,i.e.unknownwordsrecognition,Weibocorpuspre-processing,sentimentlexiconsextensionandalgorithmsforsentimentanalysis.Throughthework,wehavesignificantlyimprovedtherecognitionrateofthewo
4、rdsegmentationsystem.Andasentim、entlexiconforWeibowasbuiltbyus.WehavealsoconstructedintelligentminingmethodforthesentimentpolarityanalysisofWeibo.Finally,weachievedtobuildupaproto—systemforthesentimentanalysisandopinionminingofhottopicsonWeibo.Andthroughitwecangetthesentimentpolarityoftopics,aswell
5、asacquiretherealopinionsofWeibousers.Inthispaper,wefirstgiveabriefintroductiontothecurrentstudyofsentimentanalysisandopinionmining,anditssocialandcommercialvalue.AfterthedissectionofthelinguisticcharacteristicofcontentonWeibo,weevaluatedtheeffectivenessofdatapre—processingmethodsandalgorithmsforsen
6、timentanalysisofWeibo.Ondatapre—processing,weprovideanalgorithmforunknownwordsrecognitionwhichbasedonco-occurrencefrequency,anditcaneffectivelyrecognizeunknownwordsfromWeiboandimprovetheperformanceofwordsegmentationsystem.Thesentimentlexiconsexpansionisstartedbypre—processingofWeibocorpusbasedonthe
7、wordsegmentationsystem.Andthenistheclarificationofroughdatatomakeitwellstructured.Eventuallyweexpandtheprimarysentiment1exiconsbasedonwell.processeddata.Andinthisprocess,wehavecomparedtheadvantagesanddisadv
此文档下载收益归作者所有