A recommender system of reviewers and experts in reviewing problems

In this study, we propose the architecture of a content-based recommender system aimed at the selection of reviewers (experts) to evaluate research proposals or articles. We introduce a comprehensive algorithmic framework supported by various techniques of information retrieval. We propose a well-rounded methodology that explores concepts of data, information, knowledge, and relations between them to support a formation of a suitable recommendation. In particular, the developed system helps collecting data characterizing potential reviewers, retrieving information from relational and unstructured data, and formulating a set of recommendations. The designed system architecture is modular from the functional perspective and hierarchical from the technical point of view. Each essential part of the system is treated as a separate module, whereas each layer supports a certain functionality of the system. The modularity of the architecture facilitates its maintainability. The process of information retrieval includes classification of publications, author disambiguation, keywords extraction, and full-text indexing, whereas recommendations are based on the combination of a cosine similarity between keywords and a full-text index. The proposed system has been verified through a case study run at the National Center for Research and Development, Warsaw, Poland.