Building And Application Of Microblog Smiley Emotion Lexicon For Sentiment Orientation Analysis

As a new medium, microblog is a platform that spreads shares and accesses to information based on user’s relation. The microblog has rapidly entered the life of people plays an important role in initiating and spreading of internet public opinion and becomes a one of the most important data sources of user’s opinion mining and internet public opinion analysis. Moreover, most microblog users’ comments are short, and contain a variety of data, such as emoticons and videos. Thus the information contained in other forms of data besides text can not be ignored.As one kind of new network language, smileys is widely used in microblog, to a certain extent representing the user’s mood and thinking, and will also affect the microblog results of sentiment analysis. In this thesis, the idea of constructing a smileys emotion lexicon based on the data statistical results of microblog is proposed. Through analyzing the sentiment of text co-occurring with smileys in a large amount of microblog, the emotion of smileys is determined, and then a smileys sentiment lexicon is constructed for providing a support for other users using smileys to analyze sentiment. Moreover, the smileys sentiment lexicon affects to the microblog text, and generates a new sentiment lexicon. The new sentiment lexicon combines with exiting sentiment lexicon in sentiment analysis, for obtaining more accurate sentiment analysis results. And for the microblogs with both smileys and text, this thesis presents a fusion of text sentiment lexicon and smiley sentiment lexicon to improve the correctness of such data for sentiment analysis.Experimental results show the effectiveness of this method, and analyzed how to set relevant thresholds for affecting results, so as to obtain more effective smileys sentiment lexicon and new sentiment lexicon.