The Fine-Grained Topic Evolution Based on the Probability Model

摘要

Nowadays online social media like Weibo is gradually replacing the traditional Internet services, being a convenient and swift approach for information diffusion and access. Tracking the evolution of the topic and the trend of the situation according to the user-interested aspect across the social network has been an issue of considerable interest and has not been fully investigated. Prior research on topic evolution has often refinements to the full-analysis topic models which typically perform full analysis on the whole data to identify all emerging topics. However, such methods do not perform more detailed topic evolution analysis users always want on some specific aspects. In this study, we present a model based on targeted topic model that automatically discover and summarize the fine-grained evolutionary patterns of topics in a text stream. We evaluate our method on a Weibo stream containing over ninety thousand weibos. Experimental results show that the proposed method can discover the fine-grained evolutionary patterns of topic effectively.