基于主题契合度的专家推荐模型研究

A Topic Relevance Aware Model for Reviewer Recommendation

评审专家遴选是会议、期刊等投稿论文评审的一项重要步骤。本研究首先依据主题在学科中出现频率,分析待评审论文及候选专家论文的主题重要性;其次,综合考虑主题重要性,构造整数优化模型以高效地为多篇待审文稿推荐评审专家,并加入专家专长与投稿论文的匹配度、评审专家权威性、评审专家工作量分配等因素以满足专家推荐的现实需求;最后,从平均覆盖率、人均审核率和重要性匹配度三个角度对本模型的有效性进行验证。结果表明,提出的优化模型可以较好地完成专家推荐任务。

Reviewer assignment is an important step in the phase of paper evaluation for conference organizers and journal editors.In this study,the importanceof a topic is initially estimated by its occurrence frequency within a specific research area,which helps to express submissions and reviewers’expertise.Next,inconsideration of topic importance in submissions,an integer optimization model is formulated to recommend a reviewer group.Also,different practical constraintsare reckoned in the optimization model,which includes the affinity between reviewers and submissions,reviewers’expertise,the burden assignment of each reviewer,etc.To evaluate the effectiveness,the proposed approach is benchmarked with two baseline algorithms in terms of coverage,average number of reviewers,relevancebetween reviewers and topics,etc.Comparative experiment results show that the proposed approach is capable to recommend reviewers effectively.