Abstract:
The paper tackles the problem of selecting candidates with expert knowledge in a particular field. We propose methods for assessing similarity based on citation graphs and topical similarity of documents retrieval methods to select experts. The paper also provides a methodology for assessing the accuracy of the proposed methods and the results of experiments that were carried out on a dataset of grant applications from the Russian Foundation for Basic Research. The experimental results show that the classical methods of citation graph comparison and deep learning provide similar results. In addition, similar document retrieval methods have higher accuracy in selecting experts than citation-based methods. The proposed methods can be used not only for selecting experts for the evaluation of grant applications of scientific foundations but also for the assignment of reviewers for the analysis of any objects with text and citations.
Keywords:expert selection, graph analysis, SimRank, DeepWalk, topically similar document retrieval.