Topic correlation and individual influence analysis in online forums

Over the last few years, online forums have gained massive popularity and have become one of the most influential web social media in our times. The forum document corpus can be seemed to be composed of various topics evolved over time, and every topics is reflected on a volume of keywords and social actors. In this paper, we attempt to study the interesting problem: for the evolving topics, were there any correlation between them? We propose a method for discovering the dependency relationship between the topics of documents in adjacent time stamps based on the knowledge of content semantic similarity and social interactions of authors and repliers. We introduce mutual information measure to estimate the correlation between the topics. Applied to the realistic forum data, we show how topics are related and which postings can be recommended to another as similar topics. We also show how the authors impact the topics and propose a new way for evaluating author impact.