面向中文微博的情感分类研究

《面向中文微博的情感分类研究》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、万方数据面向中文微博的情感分类研究感分类方法具有良好的执行效率和较高的扩展性，改进后的SVM算法能够在不影响分类效果的情况下提高训练效率。关键词：情感分类；情感词典；特征选择；SVM算法；MAPREDUCEII万方数据江苏大学硕士学位论文AbstractMicroblogisaproductofrapiddevelopmentoftheInternetandhi曲一speedflowofinformation，whichattractsalargenumberofuserswithadvantagessu

2、chasbriefcontent，convenienceincommunication，andrapidspeed．Massinformationgeneratedbylarge-scalemicroblogusersismainlypostedcommentsrelatedtosocialhottopics，onlinegoods，publicfigures，andSOon，andsuchcommentshavestrongsentiments．Byanalyzingsentimentsbehindth

3、esedata,hugecommercialvaluesandsocialvaluesCanbefound．Sentimentclassificationismainlyevaluatingtendencyofasubjectivetext,includingpositivesentimentsandnegativesentiments．Aconventionalcontextualsentimentclassificationtechnologyisalreadymature，butfeaturesex

4、istinginthemicroblog，suchasaninformallanguage，ashorttext，andhugeamountsofdataundoubtedlybringalotofdifficultiestotheresearchofthesentimentclassification．Abasisofthesentimentclassificationisidentifyingsentimentwords；however,newwordsandanagramsappearinmicro

5、blogtextscausethatacommonsentimentlexiconisdifficulttobeeffectivelyusedinsentimentanalysisofthemicroblog．Forthisproblem，thispaperproposestoconstructamicroblog—sentimentlexiconasasupplementtothebasicsentimentallexicon．Candidatesentimentwordsarefirstextract

6、edfromamicroblogcorpusdatabase，awordsimilarityisthencalculatedtoconstructagraphmodel，thelabelpropagationalgorithmisusedtoannotateasentimentalpolarityofthecandidatesentimentwords，andamicroblog-basedsentimentallexiconisfinallyconstructed．Anexperimentindicat

7、esthatthemicroblogsentimentlexiconlaysagoodfoundationforimprovingaccuracyofthesentimentclassification．Becauseamicrobloglanguageisshort，precisionofsingle-featuresentimentclassificationisnothigh．Inthispaper,commontextualfeaturesinthecorpusdatabaseayeextract

8、edbyusingaCHIfeatureselectionalgorithm，thesentimentalIII万方数据面向中文微博的情感分类研究lexiconiscombinedtoextractsentimentallexiconfeatures，suchasthenumberofsentimentwords，sentimentscores，andthenumberofemoticonstosupplementthemic