An Efficient Bipartite Graph Based Tag Recommendation Algorithm

In this paper, we address the problem of diversified coverage based tag recommendation. Instead of just considering relevance of the recommended results like most of the existing methods, we consider the recommended results in terms of both diversity and relevance. To our best knowledge, we are the first introducing diversity in the problem of tag recommendation. By reducing the problem to the MIDS problem, it is proved to be an NP-hard problem. Specially, based on a constructed semantic similarity graph, an efficient tag recommendation algorithm, namely EDC, is developed. It first handles the cliques and the bipartites in the graph. Then, recursively searches the MIDSs in the remaining graph. The experimental results on the real datasets, i.e. MovieLens and Last.fm, show that EDC significantly improve the result diversity.