Categorizing Authors in Co-Authorship Networks Through Textual Information Analysis: A Comparative Study of Two Scientific Communities

The analysis of scientific collaboration is commonly based on co-authorship networks, which provide valuable insights into the structural patterns of research communities. However, a major limitation of these approaches concerns the lack of information about external co-authors who appear in publications but do not belong to the core population under study. This work proposes a methodology that combines textual information and network analysis to classify all authors in a co-authorship network, including external collaborators of a given target population of scholars referring to a specific field. Using bibliographic data from Scopus, we analyse the scientific production of two Italian academic communities—Statistics and Management—over the period 2012–2022. The approach integrates textual metadata (author keywords and subject areas) with collaboration data by constructing a bipartite network linking authors and textual descriptors. Community detection algorithms are then applied to such two-mode network to identify thematic clusters that simultaneously group both domains: researchers and textual features. The proposed method allows us to classify authors according to the thematic areas emerging from the textual information associated with their publications. In doing so, it becomes possible to infer the disciplinary positioning of external co-authors and to better interpret the structure of collaboration networks. The results highlight differences in the thematic organization and collaborative patterns of the two scientific communities, statisticians and management scholars, and demonstrate the potential of integrating textual analysis with network methods for the study of scientific collaboration.