A topic modeling approach for research community mining

Mining community on the basis of hidden relationships present between the entities is important from academic recommendation point of view. Previous approaches mined research community by using network connectivity or by ignoring semantics-based intrinsic structure of the words and author's relationships present between the conferences. In this paper, we propose a novel Venue-Author-Topic (VAT) approach which can consider semantics-based intrinsic structure of words and authors correlations, simultaneously. We also show how topics and authors can be inferred for new conferences and authors correlations can be discovered by using proposed approach. Experimental results on the corpus downloaded from DBLP shows the effectiveness of proposed approach and the detailed interpretation of results reveals interesting information about the research community.