Related Topic Network

Topic Detection and Tracking provides a flat and unorganized view of a document collection and cannot adequately reflect the content of the complete collection as some of the information is lost in the process. Topic models account for more information and lead to a more organized view of the document collection. In this paper, we propose a more efficient model named Related Topic Network with a new term weighting method. Empirical evaluation using two real-world datasets consisting of 953 and 5,550 news documents demonstrates the utility of the proposed model and shows that the new term weighting method leads to performance improvement.