欢迎来到天天文库
浏览记录
ID:34802161
大小:12.24 MB
页数:62页
时间:2019-03-11
《面向舆情监控的虚假网络信息过滤平台的研究与设计》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、::10128UDC::20151100093ConvolutionNeuralNetworkCNNword2vecSkip-GramCNNIAbstractWiththedeepeningdegreeofsocialinformation,thenetworkpublicopinionandpeople'sdailylivesarebecominginseparable.Inthefaceofalargenumberofonlinecommunitiesinourcountry,theorganizationswithulteriormotiveshirethenavytofocuson
2、somesociallysensitiveissuesandhotspotstoanonymouslypublishmanyguidingideastoguidethetopicsinthewrongdirection.Thisisnotconducivetosocialstabilityanddevelopment.Therefore,theuseofcomputertechnologytomonitortheInternetpublicopinionhasbecomeahotareaandhasacertainsocialvalue.Thisarticleexploresthefilt
3、eringofdeceptivenetworkinformationinkeytechnologiesofnetworkpublicopinionmonitoringsystem.Asamajorpartofdeceptivenetworkinformation,deceptiveopinionsalsohavecertainharmtosociety.Soitisurgenttoidentifydeceptiveopinions.Thedeceptiveopiniondetectionmainlyextractsopinionscontentcharacteristics,andusem
4、achinelearningtoachievethegoalofdetectingdeceptiveopinion.Mostofthetraditionalmachinelearningisashallowstructure,sothecomplextextcannotberepresented.Atthesametime,deceptiveopinionsdonotonlyreflectintheopiniontextattribute,butalsoreflectinthecommenter'sbehavior.So,fromasinglepointofviewoftextualcon
5、tentattributes,onlyopiniontextattributescannotfullyconsiderthecharacteristicsofdeceptiveopinions,whichcanallleadtolossoffeatures.ThepaperusestheConvolutionNeuralNetwork(CNN)whichistheoneofdepthlearningframeworktoidentifythecontentofdeceptiveopinion.Thepapermarkstheonlinepublichotelopinionstogetala
6、belleddataset.Then,thepaperpreprocessesthedata,suchasChinesewordsegmentationandstopwords,andusesthemodelSkip-Gramoftoolword2vectoobtainthewordvector.IntheaspectofCNNmodelconstruction,thepaperproposesthemodelwhichmixestextcontentattributesandbehavioralattributes.Meanwhile,thetraditionalConvolutionN
7、euralNetworkisbeingimprovedfromthepointofthewordorder,tomakeconvolutionneuralnetworkmoresuitablefortextclassification.Theexperimentalverificationshowsthatthemodelproposedinthispaperhasachievedgoodresultsinthedete
此文档下载收益归作者所有