A LDA model based text-mining method to recommend reviewer for proposal of research project selection
Yunhong Xu, Xianli Zuo · 2016
Reviewer recommendation for research projects proposals plays an indispensable role in funding agencies, because the opinions or feedback of reviewers will exert a direct impact on the result of the projects selection. Current methods mainly focus on grouping the proposals by declared disciplines or evaluating the reviewers with their individual profile, however, the two methods ignore the rich information with different types and formats of proposals and experts, such as subjective information (e.g., evaluation of colleague), objective information (e.g., publications' number). Besides, prior studies mostly applied to English documents, which has limitations when dealing with projects proposals in Chinese. In order to effectively solve the research gap that ignored the different information forms and Chinese contexts, this paper proposes firstly extract the topics words in proposal by LDA and expert's profile with text-mining; secondly automatically classify the information of proposal and profile, and integrate the information into several categories according to its different types. Each category represents the different dimensions of information of proposal and expert. Thirdly, we calculate the similarity of information in each category, and sort the similarity to select top 8 experts as candidate reviewers. Finally, we establish the evaluation model for the candidate reviewers to decide several reviewers to review proposal. A recommendation approach is proposed by integrating these categories of information. In future research, we will try to evaluate the proposed approach using real data.