Mining an enriched social graph to model cross-thread community interactions and interests.

ABSTRACTIn this paper, we present a text mining approach to generate an enriched social graph to model cross-thread community interactions and interests of Web forum users. In addition to modeling reply-to relationships between users, the proposed approach models message-similarity relationship to keep track of all similar posts resulting out of deviated discussions in different threads. The generated social graph resembles a network of clusters, where the clusters are the group of similar posts and the binding links are the reply-to relationships between them. The graph can be presented at the granule of users who authored the posts to generate a social network, and at the same time it keeps information for all other users with similar interests.