Study of Subjective Sentence Identification Oriented to Chinese Microblog
Chinese Microblog include many opinions about hot topics.Mining opinion can realize early warning and public sentiment monitoring.Most of researches are usually based on English corpus.The existing researches generally confuse opinion mining and sentiment mining.Because of the specific stylistic features of Chinese Microblog,the traditional Chinese text opinion mining models cannot achieve ideal effects.In this paper,the features selections according to the analysis of the specific stylistic features of Chinese Microblog.Selecting declared verb,modal particles,degree adverb and fixed part of speech structures as the experiment features except the sentimental feature,which distinguish from sentiment mining.This paper used a CRFs(Conditional Random Fields) as the classification model.The results showed that recall ratio is only 32.1％,which is only used the sentimental feature.Added the other features,the recall ratio increased to 61.8％.This method was achieved an ideal effect with the opinion mining task of Chinese microblog which is held by China Computer Federation Technical Committee on China Information Technology.
Chinese microblogopinion miningCRFs modelopinion recognitionstylistic features
NETL
NSTL
万方数据
维普
