Cross-Domain Citation Recommendation Based on Co-Citation Selection

Recommending information across domains has recently gained much attention among research and academic communities. Traditionally, a cross-domain recommender system has emerged to assist users in finding relevant information from the target domain given the initial information from the source domain. However, in the area of citation recommendation, mapping terms across different domains could be problematic due to the term mismatch. In this paper, we propose a cross-domain citation recommendation framework to suggest relevant research publications given a patent as the source domain. Two main approaches are implemented and compared in this study. The first is a baseline approach which is based on simple keyword mapping technique. The second approach, Co-Citation Selection (CCS), is based on the collaborative filtering in which neighboring papers is selected and weighted into publication citation prediction. To compare between two approaches, we adopt the Cosine, Jaccard, and KL-Divergence as the similarity measurement. The evaluation results are reported in terms of mean precision, recall, F-measure, and reciprocal rank. The best improvement of 22.6% in mean reciprocal rank was achieved with the Jaccard similarity.