Online Topical Clusters Detection for Top-k Trending Topics in Twitter

This paper tackles the problem of detecting temporal query oriented topical clusters for top-k trending topics from Twitter. There is an increasing demand to identify and cluster set of users who have similar topical interests as well as certain level of activeness on those topics. Most existing approaches focus on the contents generated by the social users and link structure of the underlying social network. However, the degree of users' topical activeness has not been thoroughly studied to identify its effect on the formation of topical clusters. This research investigates on how the users' behaviors and topical activeness vary with time and how these parameters can be employed in order to improve the quality of the detected topical clusters for top-k trending topics at different time intervals. The effectiveness of our proposed activity biased weight methodology is justified using a benchmark Twitter dataset.