Literature recommendation by researchers' publication analysis

Abstract

Scholarly paper recommendation has been an important research topic in the field of information filtering because scholars find thousands of publications that match their search queries but are largely irrelevant to their latent information needs. Many existing methods are based on the users' online behaviours to construct user interest model. However, an author's published works constitute a clean signal of his latent interests. In previous research, Sugiyama et al. recognized the important user profile and proposed the scholarly paper recommendation via user's recent research interests. However, in reality, a user's interests can be diverse and a researcher's published papers often involve his multiple research directions. On the other hand, it is often more desirable to recommend papers based not only on content relevance but also on the reputation of venues. To deal with the above mentioned limitations and problems, in this paper, a novel scholarly paper recommendation method is proposed. The main distinctive features of the proposed model include: (1) user's information needs are generated in term of scholar's multiple research directions; (2) user's interests are represented by his research content and recommendation standard; (3) research content model and recommendation standard of each scholar is generated from his published papers, citation papers and reference papers; (4) in order to distinguish the scholar's different attention on his research contents, we present the penalty factor concept. Extensive experiments are designed to evaluate the effectiveness of the proposed model by using one real data set. The results show that the proposed model significantly outperforms existing methods.