Sentiment Lexicon Network For Short Texts Sentiment Classification

With the vigorous development of web2.0 applications such as microblog, forums, blogs, a large number of user generated texts emerged on the web.Through the comprehensive analysis of short text, you can understand the opinions and attitudes of internet users for social events,people, products and evolutionary trend of public opinion.Therefore, the analysis process of these text information becomes particularly important and text sentiment analysis is one of the main core technology. In this paper, we analyzed the features of short text and for the sentiment analysis of short text, the following work is carried out :At the beginning, to build Sentiment Lexicon Network. Firstly, the non-negative matrix factorization algorithm is used to construct a sentiment lexicon network which can express the relationship between word-words and words-objects based on a large scale of corpus and synset. Experiments demonstrate the effectiveness of the Sentiment Lexicon Network in the sentiment analysis area.Then a sentiment classification method combined with sentiment lexicon network for Chinese shot texts is proposed. The method employs the sentiment lexicon network to extend the weight and the feature set of shot texts. The machine learning algorithms are combined with the sentiment lexicon network to fulfill sentiment classification task. Experiments show that the method could achieve higher precision and recall rates compared with other methods, and effectively solves the problems of the features sparse and the less information in short text sentiment classification task.At the last, the sentiment analysis is employed in the analysis of the evolution of internet public opinion. In this paper we treat weibo as the research object, in view of the short data features such as weibo,a evolution of network public opinion analysis method based on sentiment analysis is proposed. The method uses the sentiment value as the feature to time slicing weibo data then use the DTM model to evolution analysis the data. Experimen ts show that this method can divide microblogging data better and find public opinion changes time points efficiently.