A Citation-Based Recommender System for Scholarly Paper Recommendation

Abstract

Several approaches have been proposed to help researchers in acquiring relevant and useful scholarly papers from the enormous amount of information (information overload) that is available over the internet. The significant challenge for those approaches is their assumption of the availability of the whole contents of each of the candidate recommending papers to be freely accessible, which is not always the case considering the copyright restrictions. Also, they immensely depend on priori user profiles, which required a significant number of registered users for the systems to work effectively, and a stumbling block for the creation of a new recommendation system. This paper proposes a citation-based recommender system based on the latent relations connecting research papers for the scholarly paper recommendation. The novelty of the proposed approach is that unlike the existing works, the latent associations that exist between a scholarly paper and its various citations are utilised. The proposed approach aimed to personalise scholarly recommendations regardless of the user expertise and research fields based on paper-citation relations. Experimental results have shown significant improvement over other baseline methods.