基于中文微博的情感分类技术研究

《基于中文微博的情感分类技术研究》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、万方数据东北大学硕士学位论文AbstractWiththerapiddevelopmentofnetworktechnology,Weibo，asanewsocialplatform，hasgraduallypenetratedintoeveryaspectofpeople’Slives．Weiboisfullofusers’viewsandopinionstowardsproducts，entertainments，socialeventsandSOon，whichcontainsawealthofemotionalinformation．Analyzingusers’e

2、motionalattitudeduringaperiodoftimeortowardsaparticulartopic，classifyingtheemotionscontaininWeiboeffectivelyhasgreatcommercialvalueandsocialvalue．Itnotonlyallowsbusinessmentoobtainusers’viewsinstantly,butalsoallowsgovernmentdepartmentstokeepabreastofsocialdynamics，listentothevoiceofthepeop

3、le，whichhasagoodmonitoringroleofpublicopinion．Inthisthesis，wewilldodeepresearchonChineseWeibosentimentclassification，includingsentimentpolarityclassificationandsentimentfine-grainedclassification．(1)Sentimentpolarityclassificationisstudied．WeclassifythesentimentofChineseWeibointopositivean

4、dnegative，andimprovesometechnologythatinvolves．Firstly,informationgain，thetraditionalmethodoffeatureselectiondoesnotconsiderthefeatures’appearanceinintra-classandinner-class，tosolvetheproblem，weintroducetwofactors，concentrationratioanddistributedratio．Secondly,weconsiderWeibo’Scharacterist

5、icswhenwecalculatethefeatureweight，wecombinethefeature’Semotionalinformationand10cationinformationwithtlletraditionalTF．IDFcalculationmethod．(2)Sentimentfine—gainedclassificationisstudied．WeclassifythesentimentofChineseWeibointosevencategories，includinghappiness，like，surprise，anger,sadness

6、，fearanddisgust．Weanalyzeandimprovethetraditionalmethods．Firstly,weexpendstheexistingmulti·classemotionaldictionarytocompensateforitslackofwordscoverage．WemakeuseofWeibotrainingcorpustogeneratecandidatesentimentfeatures，proposeasentimentfeatureselectionTF—IDFmethodbasedonvariance．Thencalcu

7、latethefeature’scategoryandemotionalstrengthandadditintoIII万方数据东北大学硕士学位论文Abstractthemulti．classemotionaldictionary．Secondly,wecalculatetheWeibofine-grainedsentimentscorebasedontheexpandedemotionaldictionary．IntheprocessofWeibosentimentfine—grainedclassificatio