Post Recommendation in a Blogosphere

Abstract

In this paper, we present new post recommendation schemes that use link-based similarity in a blogosphere. Given the posts that a certain user has been interested in, the basic scheme first selects seed posts among them and then computes link-based similarity. With this similarity, k posts are selected and recommended, which are expected to have the most similar subjects to the seeds. To address the problem to recommend posts that are unrelated to the seeds, we propose an incremental recommendation method where selecting additional seed posts and calculating similarity are interleaved. We conducted experiments which confirm the effectiveness of our recommendation system.