Finding potential research collaborations from social networks derived from topic models

Author(s): Md Asaduzzaman Noor, John Sheppard, Jason Clark

Abstract

Community detection is a valuable tool for analyzing social networks given its potential for identifying groups with common characteristics and common interests. In this work, we focused on detecting scholarly communities based on researchers’ publication data to discover interdisciplinary collaboration recommendations (researchers working on different domains). Specifically, instead of using any physical or direct relationship between researchers, we utilized a topic model to obtain the topic-based similarity between researchers to construct the social network graph. Next, we employed an edge-weakening procedure to alter the initially constructed network to uncover a more refined community structure. Two community detection algorithms, Louvain and Spectral clustering, were utilized to find the community structures in the modified network. The results of our experiments revealed the ability to discover possible research communities for both algorithms that were comparable, which suggests that our method has the potential for identifying hidden interdisciplinary research collaboration recommendations using topical relationships as the basis for building and analyzing the social network graph.

Citation

Noor, M. A., Sheppard, J., & Clark, J. (2023). Finding potential research collaborations from social networks derived from topic models. In Proceedings of the 2023 10th International Conference on Behavioural and Social Computing (BESC) (pp. 1–7). IEEE. https://doi.org/10.1109/BESC59560.2023.10386981