Research On Sentiment Analysis Method Of Chinese Micro-blog

With the development of mobile network terminals and sina micro-blog, micro-blog users grew rapidly, micro-blog data also showed explosive growth. Massive micro-blog data hiding a lot of emotional content, sentiment analysis of these data can obtain useful decision support information. Sentiment analysis of micro-blog data is advantageous to detection and monitor public opinion, forecasting and product sales, product improvements. Chinese micro-blog data is different from the ordinary text, so its sentiment analysis method. The existing sentiment analysis methods have advantages and disadvantages. To improve the sentiment analysis method of Chinese micro-blog, this paper carries on the following two parts research.Micro-blog sentiment expression elements are analyzed and assigned a weight, including sentiment words, emoticons, extent and negative adverbs, sentence pattern. Sentiment lexicon is a collection of sentiment words, is an important feature of sentiment expression. It is essential to build the sentiment lexicon, this paper propose a method to discover new sentiment words based on emoticons. Integrated degree and negative adverbs interaction to sentiment words, summarizes seven types of semantic rules which about modified way between sentiment words and adverbs, innovative add semantic rule which adverb is behind the modified sentiment word.Sentiment lexicon method and machine learning method are the two main sentiment analysis methods, we analyzed and compared the two main methods, proposed Lexicon-SVM which combines the advantages of these two methods. Lexicon-SVM method using SVM algorithm to micro-blog classification, it improve the feature item weight calculation method when constructing micro-blog text vector space model, feature item weight calculation combines the sentiment word polarity, semantic rules and sentence pattern. Lexicon-SVM use frequency words filtered, synonyms merged, the same level of sentiment words merged method to achieve dimensionality reduction.Lexicon-SVM method can effectively solve the problem of unknown sentiment words and neglecting context. Finally, to verify the performance of the proposed Lexicon-SVM method by micro-blog corpus, the results show Lexicon-SVM method shows higher accuracy.