Game theory based emotional evolution analysis for chinese online reviews

Sentiment analysis has become one of the mainstream researches in social network analysis. Its impact can be seen in many practical applications, ranging from public opinion analysis to marketing of public praise and information prediction. However, most of the existing research has been performed in the sentiment classification for subjective text, the emotional evolution analysis for complex interactive text (e.g., online reviews) has not yet been thoroughly targeted by the research community. This paper is concerned on short-text Chinese online reviews collected from Tianya forum. First, an efficient affective computing framework is proposed to capture the underlying emotions of Chinese online reviews. It can accurately calculate the semantic orientation of the entire review, without requiring expensive manual labeling of seed words. As users' attitudes might influence with each other, predicting their future emotional behaviors that only relying on the emotional values of historical reviews is very one-sided. Therefore, we propose a game theory based emotional evolution prediction algorithm combining the affective computing, in which the mixed nash equilibrium strategies are calculated as the future emotional behavior of interactive users. Then, experimental results on the large-scaled review dataset are provided to demonstrate the effectiveness and accurateness of our approaches. Finally, by applying the research results on the pairwise happiness-popularity coordination evaluation, we have revealed some interesting phenomenon on the “World View” board in Tianya forum.