Topic Detection Based on User Intention

Topic detection always plays an important role in social network analysis. In this paper, we focus on a very simple question that how to choose the terms that can represent a topic better before topic detection. To tackle this problem, we propose an effective model named Topic Model based on Entropy and LDA (TMELDA). The model is built on the user intention, which means different users have different knowledge for topic detection. What´s more, the choice of terms in TMELDA is not only based on semantic relevance but also on the consideration of evenness extent of term distribution. An extensive empirical study using real Sina Weibo data clearly demonstrates that our method has a better performance in topic detection.