Deep Learning for Sentiment Analysis A Survey

《Deep Learning for Sentiment Analysis A Survey》由會員上傳分享，免費(fèi)在線閱讀，更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫。

1、DeepLearningforSentimentAnalysis:ASurveyLeiZhang,LinkedInCorporation,lzhang32@gmail.comShuaiWang,UniversityofIllinoisatChicago,shuaiwanghk@gmail.comBingLiu,UniversityofIllinoisatChicago,liub@uic.eduAbstractDeeplearninghasemergedasapowerfulmachinelearningtechniquetha

2、tlearnsmultiplelayersofrepresentationsorfeaturesofthedataandproducesstate-of-the-artpredictionresults.Alongwiththesuccessofdeeplearninginmanyotherapplicationdomains,deeplearningisalsopopularlyusedinsentimentanalysisinrecentyears.Thispaperfirstgivesanoverviewofdeeple

3、arningandthenprovidesacomprehensivesurveyofitscurrentapplicationsinsentimentanalysis.INTRODUCTIONSentimentanalysisoropinionminingisthecomputationalstudyofpeople’sopinions,sentiments,emotions,appraisals,andattitudestowardsentitiessuchasproducts,services,organizations

4、,1individuals,issues,events,topics,andtheirattributes.TheinceptionandrapidgrowthofthefieldcoincidewiththoseofthesocialmediaontheWeb,forexample,reviews,forumdiscussions,blogs,micro-blogs,Twitter,andsocialnetworks,becauseforthefirsttimeinhumanhistory,wehaveahugevolume

5、ofopinionateddatarecordedindigitalforms.Sinceearly2000,sentimentanalysishasgrowntobeoneofthemostactiveresearchareasinnaturallanguageprocessing(NLP).Itisalsowidelystudiedindatamining,Webmining,textmining,andinformationretrieval.Infact,ithasspreadfromcomputerscienceto

6、managementsciencesandsocialsciencessuchasmarketing,finance,politicalscience,communications,healthscience,andevenhistory,duetoitsimportancetobusinessandsocietyasawhole.Thisproliferationisduetothefactthatopinionsarecentraltoalmostallhumanactivitiesandarekeyinfluencers

7、ofourbehaviours.Ourbeliefsandperceptionsofreality,andthechoiceswemake,are,toaconsiderabledegree,conditioneduponhowothersseeandevaluatetheworld.Forthisreason,wheneverweneedtomakeadecisionweoftenseekouttheopinionsofothers.Thisisnotonlytrueforindividualsbutalsotrueforo

8、rganizations.Nowadays,ifonewantstobuyaconsumerproduct,oneisnolongerlimitedtoaskingone’sfriendsandfamilyforopinionsbecausetherearemanyuserr