Sentiment Analysis Of Popular Events Based On Chinese Microblog Network

In recent years, microblog has become a very important platform for people’s online activities. When popular events happen, users can get information from microblog platform, and they can participate in the discussion and diffusion of the events via the functions provided by microblog, such as posting, retweeting, commenting and mentioning. Users can publish their opinions and express sentiment about events of different topics. Overall sentiment state and sentiment diffusion of the public can be discovered by analyzing the microblog data related to the popular events.In this paper, a sentiment classification method for Chinese microblog texts is proposed with the features extracted based on the short-text and novelty characteristics of Chinese microblog texts. One-step strategy and two-step strategy for classification process are proposed and compared. On the basis of classification method, this paper analyzes the sentiment dynamic properties of popular events on social, entertainment and sport topics with Sina microblog events dataset. The sentiment transition patterns in the sentiment dynamic diffusion are analyzed. And this paper constructs the model to predict the sentiment transition based on conditional random fields.The sentiment classification method proposed in this paper classifies the Chinese microblog texts into positive, neutral or negative sentiment polarities. The experiment result shows that the classification F1-measure is74.9%in one-step strategy and82.4%in two-step strategy. The sentiment classification method is effective and the result proves that the two-step strategy performs better than one-step strategy. The sentiment transition prediction model based on conditional random fields achieves60.2%F1-measure, which is3.7%better than the baseline SVM model performance.