An Automatic Paper-Reviewer Recommendation Algorithm Based on Depth and Breadth

Paper-reviewer recommendation is an effective method to match reviewers for papers in peer review. However, existing recommendation methods are either sensitive to the order of paper process or not excepted fair in total. How to efficiently and accurately recommend is still a tough task and needs to be further explored. Thus, in this paper, we propose a greedy-version automatic paper-reviewer recommendation algorithm regarding both aspects of depth and breadth, named as GMCTS. More specifically, from the depth aspect, we focus on the maximum weight matching between paper and reviewer for each recommendation. While from the breadth aspect, we consider the broadest of distinctive topics (expertise) covered by the recommended reviewers. Furthermore, to avoid the unfairness of recommendation, we consider three types of constraints both theoretically and experimentally, including paper demand constraint, reviewer workload constraint and multiple types of Conflict Of Interests (COIs). Finally, extensive experiments conducted on benchmark datasets demonstrate that GMCTS achieves an impressive performance. Specifically, GMCTS scores a 90.5% and 12.67% gain in topic coverage and recall respectively compared to some advanced recommendation methods.