Detecting Opinion Leader Dynamically in Chinese News Comments

摘要

Nowadays, more and more users publish their opinions by means of Web 2.0. Analyzing users' opinion, discovering the relationship between opinions, and detecting opinion leader are important for Web public opinion analysis. Opinion leader is regarded as the most influential comment and user during the information dissemination process. However, most existing researches pay less attention on their internal relation, implicit relationship between users' opinions and the changes of opinion leader over time. In this paper, we focus on modeling comment network with explicit and implicit links for detecting the most influential comment dynamically, and modeling user network and clustering users for detecting the most influential user. We propose an approach with sentiment analysis, explicit and implicit link mining for modeling comment network in Chinese news comments. We also propose an algorithm for detecting most influential comment from the comment network dynamically. Moreover, we model user network based on the comments network, and detect the most influential user from the user network. Experiments using Chinese Sina news comment dataset show that our approach can detect opinion leaders and the changes of them over time dynamically.