The Study Of Emotion Classification Of Chinese MicroBlog Combining The Rules And Models

With the development of network and information technology, as a new network product, micro blog gradually attracts more and more users join in weibo, share news through weibo, and to express ideas and emotions. Domestic and foreign scholars began to study weibo for massive amounts of data, but among the issues, the more important is the sentiment analysis. Sentiment analysis is mainly on emotional polarity judgment, namely to determine a weibo message emotion of positive and negative, and neutral. Because Chinese weibo has a late start, so far, these studies mainly for weibo in English, with Chinese weibo research work is still in its infancy.Based on sina weibo platform, in view of the microblogging automatic tagging corpus emotion and classification method carries on the analysis and research, this paper proposes a weibo without manual intervention method of automatic tagging corpus, thus realize the microblog text automatic classification of emotion. Weibo emotion tendentiousness in this paper, divided into three classes of positive, negative and neutral, according to the characteristics of weibo language use of emotion words, emoticons, network new words and so on the many kinds of feeling carriers, to complete the automatic tagging corpus, then extract has shown the characteristics of the training corpus, weibo data classification for under test, realization of micro blog this emotion classification, finally to verify this method. Main work of this paper is:The first of all is that micro blog of the emoticon frequently appears in this phenomenon, respectively from the the position of expression symbols appear in the text, pragmatic functions, concurrence phenomena, several aspects, such as the simple function were studied. And the commonly used 84 default weibo emoticons emotional polarity are classified and construct an emoticons dictionary.The second that is in terms of feature extraction, adding the network new words, emoticons, emotional templates, emotional characteristics, such as considering the punctuation on the text of emotional expression often has a special function, based on micro blog this emotional features extraction, adding the "!" And "?" punctuation characters to complete the micro blog emotional polarity classification.Third thing is that for weibo "optional" characteristics of language, through the artificial collection of emotional words, network words, adversative, negativity and adverbs of degree words,to build an emotional lexicon, and then provide emotional judgment basis for automatic tagging of weibo corpora.Fourth is to proposeing a weibo corpora based on emoticons and emotional words automatic tagging method using the constructed based on the expression of symbol dictionary and emotional word dictionary. Base on the results of manual annotation, this article puts forward the method of automatic tagging corpus and the emotion word automatic tagging method, emoticons automatic tagging method, and verified the accuracy of the text proposed labelling method.Fifth, aimed at the characteristics of message of microblogs “varies in length”, this paper proposes a emotion classification method which combined rules with models of microblog.Finally, summarize the full text of the work, and points out the next research direction.