Discovering Topics from Reviews and Emotion Analysis Based on Total Probability Model

摘要

With the rapid development of Internet economy, reviews have become an important way to obtain information about the user experience and to drive consumers to shopping. A topic discovery method on reviews is proposed in paper. First, discovering candidate topics quickly by LDA total probability model, and refining topics in the time windows based on the features of review. Then, multi-feature fusion method was used to compute emotion tendency of topics combined with Chinese language Characteristics. The proposed approach also solves the problem of subjective information missing, and enhances the effectiveness of text mining. The experimental results confirmed the usefulness of proposed method.