PUBLIC OPINION ANALYSIS BASED ON PROBABILISTIC TOPIC MODELING AND DEEP LEARNING

With the rapid development of Internet, especially the social media technologies, the public have gradually published their perception of social events online through social media. In Web2.0 era, with the concept of extensive participation of public in social-event-related information sharing, the effective content analysis and better results presentation for these media published online thus possesses significant importance for public opinion analysis and monitoring. In view of this, this paper proposes a novel method for public opinion analysis on social media website. First, the probabilistic topic model of Latent Dirichlet Allocation (LDA) is adopted to extract the public ideas about the distinct topics of certain event, and then the deep learning model named word2vec is used to calculate the emotional intensity for each text. Next, the underlying themes in the whole as well as the events of emotional intensity are investigated, and the variation trend of public’s emotion intensities is tracked based on time series analysis. Finally, the rationality and effectiveness of the method are verified with the analysis of a real case