资源描述：
《holder and topic based analysis of emotions on blog texts a case study for bengali》由会员上传分享，免费在线阅读，更多相关内容在工程资料-天天文库。
1、HolderandTopicBasedAnalysisofEmotionsonBlogTexts:ACaseStudyforBengaliDipankarDasandSivajiBandyopadhyayAbstractThepaperpresentsanextendedapproachofanalyzingemotionsoftheblogusersondifferenttopics.Therulebasedtechniquestoidentifyemotionhold-ersandtopicswithr
2、especttotheircorrespondingemotionalexpressionshelpstode-velopthebaselinesystem.Ontheotherhand,theSupportVectorMachine(SVM)basedsupervisedframeworkidentifiestheholders,topicsandemotionalexpres-sionsfromtheblogsentencesbyoutperformingthebaselinesystem.Theexis
3、tenceofmanytomanyrelationsbetweentheholdersandthetopicswithrespecttoEk-man’ssixdifferentemotionclasseshasbeenexaminedusingtwowayevaluationtechniques,oneiswithrespecttoholderandotherisfromtheperspectiveoftopic.Theresultsofthesystemwerefoundsatisfactoryincom
4、parisonwiththeagreementofthesubjectiveannotation.Theerroranalysisshowsthatthetopicofablogatdoc-umentlevelisnotalwaysconveyedatthesentencelevel.Moreover,thedifficultyinidentifyingtopicfromablogdocumentisduetotheproblemofidentifyingsomefeatureslikebigrams,Nam
5、edEntitiesandsentiment.Thus,weemployedasemanticclusteringapproachalongwiththesefeaturestoidentifythesimilaritybetweendoc-umentleveltopicandsententialtopicaswellastoimprovetheresultsofidentifyingthedocumentleveltopic.KeywordsEmotion·Blog·Holder·Topic·SVM·NE
6、·Discourserelation1IntroductionEmotionanalysisisarecentsubdisciplineatthecrossroadsofinformationre-trieval[1]andcomputationallinguistics[2].AwiderangeofNaturalLanguageProcessing(NLP)taskssuchasclassificationofnewspaperarticles[3],blogs[4],questionanswerings
7、ystems[5],moderninformationretrievalsystems[6]areusingD.Das(B)·S.BandyopadhyayDepartmentofComputerScienceandEngineering,JadavpurUniversity,Kolkata,Indiae-mail:dipankar.dipnil2005@gmail.comS.Bandyopadhyaye-mail:sivaji_cse_ju@yahoo.comT.Özyeretal.(eds.),Mini
8、ngSocialNetworksandSecurityInformatics,127LectureNotesinSocialNetworks,DOI10.1007/978-94-007-6359-3_7,©SpringerScience+BusinessMediaDordrecht2013128D.DasandS.Bandyopadhyayemotionalinformation.Theresearchactiv
