一种面向主题覆盖度与权威度的评审专家推荐模型研究

目的/意义] 为投稿论文遴选出合适的审稿专家是论文发表过程中关键的一环。随着投稿论文和候选评审专家数量的持续增长，人工指定评审专家的方法在准确性和公平性上的弊端日益显露出来。因此，为进一步提高专家评审的客观性和准确性，笔者从专家知识与专家权威度两个维度对专家建模，并以此为依据为不同主题的投稿论文遴选推荐评审专家。方法/过程] 首先分析专家知识以及投稿论文的研究内容，并提取两者涉及的多个子研究主题；然后，计算专家知识对投稿论文子主题的覆盖度，并提出融合主题特征与时间特征的权威度算法TTAM来分析专家权威度；最后，提出融合主题覆盖度和专家权威度的专家推荐框架CAUFER，综合考虑覆盖度和权威度两个因素为投稿论文推荐合适的评审专家。结果/结论] 实验结果表明，与经典的基于向量空间模型、语言模型和作者主题模型3种专家推荐算法相比，笔者提出的算法能够较好地提高专家与投稿论文的匹配度，并可据此追踪专家权威度的变化，刻画专家在特定主题下的权威度，进一步提高专家推荐的准确性和科学性。

A Topical Coverage and Authority Unification Model for Expert Recommendation

Purpose/significance] The selection and recommendation of expert play an important role in the process of publications. Due to the remarkable increase in the number of submissions and candidate experts, selecting the proper experts manually for peer reviewer appears its weakness in terms of the accuracy and efficiency. Accordingly, an intelligent algorithm that automatically selects and recommends experts for submissions is of great importance.Method/process] In this research, the knowledge of each candidate expert and the research content are extracted, which represent several distinct sub-topics. Then, the topical coverage and the authority of reviewer candidates with respect to each submission, which are treated as indispensable evidences for expert recommendation, are deeply exploited. Finally, these two important factors are linearly combined in an integrated model for recommending proper experts, which is named as CAUFER (Coverage and Authority Unification Framework for Expert Recommendation).Result/conclusion] Different experiments were conducted and the results show that, compared with the Vector Space Model, Language Model and Latent Dirichlet Allocation, the proposed model will effectively grasp the dynamic authority change as well as the different level authority with respect different subtopics.