Academic Paper Recommendation Based on Community Detection in Citation-Collaboration Networks

Academic search engine plays an important role for science research activities. One of the most important issues of academic search is paper recommendation, which intends to recommend the most valuable literature in a domain area to the users. In this paper, we show that exploring the relationship of collaboration between authors and the citation between publications can reveal implicit relevance between papers. By studying the community structure of the citation-collaboration network, we propose two paper recommendation algorithms called Adaptive and Random Walk, which comprehensively consider several metrics such as textural similarity, author similarity, closeness, and influence for paper recommendation. We implement an academic paper recommendation system based on the dataset from Microsoft Academic Graph. Performance evaluation based on the assessments of 20 volunteers show that the proposed paper recommendation methods outperform the conventional search engine algorithm such as PageRank. The efficiency of the proposed algorithms are verified by evaluation.