基于中文微博的情感分析分析

《基于中文微博的情感分析分析》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、AbstractMicroblogisbecomingamostpopularinternetapplication.Accordingtothestatistics,morethan100milliontweetspublichedineveryday.Thesetweetsnotonlyconveythedescriptionoffacts,butalsocontaintheemotionalstatesofmassivemicroblogusers.Andtheseemotionalinformationsmaybehelpforusertodecidewhe

2、therbuyaproduct,provideveryimportantreferencevalueforcompaniestomakemarketstrategy,andevenmakemassivedataavailableforgovernmenttomonitoringpublicopinion.Inlightofthis,weproposedasentimentanalysismethodbasedonacombinationofsyntacticdependenciesandtextclassificationtechniquesforChinesetwe

3、ets.Themethodadoptsthesyntacticdependenciestoperformsentimentanalysis,atthesametime,computesaconfidenceforeverytweet.Choosentweetswhichconfidenceaboveacertainthresholdastrainingsamples,trainatwo-stepsentimentclassifierbyusingthecontentfeaturesandmediafeaturesoftweets.Finally,classifyth

4、esentimentorientationoftweetsagain.Inaddiation,wealsoproposedamethodthatservescommonemoticonsasthesentimentclasslabelsoftweetsandimplementsanincrementallearningmethodtotackletheproblemofreal-timesentimentanalysis.Experimentalresultsshowthattheproposedmethoddramaticallyimprovesthepreci

5、sionandtherecallby6%and3%repectivelycomparedtothemethodthatonlybasedonsyntacticdependencies.Andtheperformanceofourtwofeaturesetsarealsobetterthanunigramfeatures,theprecisionandtherecallbothare88%intermofsubjectiveclassifier,andtheyare72」％and71.5%forsentimentclassifie匚Apartfromthis,theme

6、diafeaturesaregoodfortracklingtheproblemofreal-timesentimentanalysis.Keywords:ChineseMicroblog,SentimentAnalysis,Syntacticdependencies,TextClassification目录摘要IABSTRACTII1绪论1・1课题研究背景(1)1.2课题的研究目的和意义(2)1.3国内外研究现状(3)1・4论文的研究内容(7)1.5论文结构安排(8)2微博相关研究工作综述2.1微博社会网络的特性分析(9)2.2微博文本内容的语义分析(12)2.3

7、微博中的应用研究分析(13)2.4本章小结(14)3基于句法依存关系的微博情感分析3.1句法依存关系概述(15)3.2微博预处理(17)3.3Bag-of-Targets模型(20)3.4评价词及评价对象抽取规则(20)3.5情感倾向性判别(23)3.6实验评估(29)3.7本章小结(33)4基于文本分类的微博情感分析4.1文本分类技术概述(36)4.2训练样本集的选择(37)4.3样本预处理(42)4.4主客观分类特征(43)4.5情感倾向性分类特征(45)4.6实验分析(46)4.7本章小结(53)5总结与展望5.1论文工作总结(54)5.2进一