User interests modeling in online forums

This paper studies the problem of user modeling in online forums from a personality viewpoint. A novel hierarchical user profiling mechanism is proposed, which utilizes the user-generated content, the reply relations among users and the topics of the discussions. The hierarchical model represent the users' interests across different topics. The obtained user profiles are applied to three forum-related tasks: new discussions recommendation, external news articles recommendation and user retrieval. The experimental results show that, comparing with the traditional methods, the hierarchical user profiling approach achieves a better performance in all three tasks.