Categorizing authors in co-authorship networks through textual information analysis: a comparative study of two scientific communities

Abstract Scientific collaboration is a key driver of research productivity and innovation, enabling researchers to share knowledge, leverage expertise, and improve research outcomes. While co-authorship is commonly used to represent collaborative behaviour, most studies focus on specific single scientific fields or national networks. A significant limitation of these studies is the lack of comprehensive methods to classify all authors within co-authorship networks, particularly in relation to their field or macro-disciplinary affiliations. This study addresses these limitations by analysing co-authorship networks within two distinct Italian scientific target groups—Statistics and Management. Using a dataset of publications from 2012 to 2022, we develop a novel methodology that combines bibliographic and textual data, such as keywords and subject areas, to classify all authors in these communities. By constructing a two-mode network and applying a clustering algorithm, we categorize authors into distinct thematic areas, offering a more nuanced analysis of collaboration. Additionally, we compare the collaborative structures of the two target groups, revealing important similarities and differences in their research dynamics. This approach provides new insights into the evolution of scientific collaboration and offers a deeper understanding of how researchers from different fields engage in knowledge production.