The article describes the problem of finding and selecting experts for reviewing grant applications, proposals and scientific papers. The main shortcomings of the methods that are currently used to solve this problem were analyzed. These shortcomings can be eliminated by analyzing large collections of sci-tech documents, the authors of which are potential experts on various topics. The article describes a method that forms a ranked list of experts for a given document using a search for documents that are similar in topic. To evaluate the proposed method, we used a collection of grant applications from a science foundation. The proposed method is compared with the method based on topic modeling. Experimental studies show that in terms of such metrics as recall, MAP and NDCG, the proposed method is slightly better. In conclusion, the current limitations of the proposed method are discussed.
PDF at the CEUR Workshop Proceedings journal: ceur-ws.org/Vol-2523/paper25.pdf
DAMDID/RCDL 2019 conference in the archives of the Kazan Federal University with a link to collection of proceedings (PDF): https://dspace.kpfu.ru/xmlui/handle/net/151948
In the collection of proceedings from the DAMDID/RCDL 2019 conference (PDF, EPUB) at SpringerLink, pages 163—180: https://link.springer.com/book/10.1007/978-3-030-51913-1
Semantic Scholar: https://api.semanticscholar.org/CorpusID:209437163
Zubarev D. V. Devyatkin D. A., Sochenkov I. V., Tikhomirov I. A., Grigoriev O. G. Expert Assignment Method Based on Similar Document Retrieval // Data Analytics and Management in Data Intensive Domains: ХХI International Conference DAМDID/RCDL'2019 (October 15–18, 2019, Kazan, Russia): Conference Proceedings. Edited bу Alexander Elizarov, Boris Novikov, Sergey Stupnikov. – Kazan: Kazan Federal University, 2019. – pp. 339–351.