The Research Of Sentiment Analysis Techniques For Chinese Microblog

With the rapid development of Internet, the way of Internet information interaction becomes more and more diversified. In the Web2.0era’s mainstream social networking platforms, microblog has become one of the most loved social tools for the majority of Internet users. In the age of information with speed and efficiency as the standard, microblog not only provides people with the instant interactive communication platform that spans time and distance, but also provides people a dynamic display platform with self-expression, expression of emotions and showing the character. Microblog messages are renewed every moment, and most of them are texts with emotional. These texts make the microblog’s opinion analysis become possible. So far, the existing research is mainly aimed at English microblog, the research of Chinese microblog is still in its beginning stages. The main research contents of this paper include the following: (1) Studied linguistic characteristics of the topic based Chinese microblog. According to the relevant characteristics, this paper analyzed the feature for viewpoint sentence recognition and sentence’s polarity determination, and at the same time analyzed the problem of extraction opinion elements from the microblog corpus.(2) Constructed seven categories of emotional dictionaries, including opinion dictionary, network language dictionary, the emoticon dictionary of tencent microblog, dictionary of subjectivity word, the modal particle dictionary, dictionary of adverbs and the dictionary of negative word. In this paper, we offered an new method to automatic expand the opinion dictionary.(3) Used methods combined the rules and machine learning to practice the three emotional analysis tasks of microblog corpus: viewpoint sentence recognition, sentence’s polarity determination and emotional elements extraction. For viewpoint sentence recognition this paper firstly used rules to filter out parts of the none-viewpoint sentence, and then used the SVM method to classify viewpoint sentences by analyzing features such as emoticon, emotion word etc.. Based on the result set of classifying viewpoint sentences, this paper used the SVM to classify the emotional polarity, and used the positioning rule of emotional words and CRF to extract the evaluation objects and judge their polarity.