Using Coauthor Networks to Extract Topics in Information Systems

Detecting and labeling various research groups in the Information Systems (IS) field is crucial to understand the community. Building author collaboration networks and analyzing highly ranked historical publication records are straightforward to approach this goal. In this paper, we collect top IS journal papers to build multiple implicit coauthor networks. We study structural properties of networks, especially eigenvector centrality because it indicates the influence of a node. We propose a hierarchical community detection algorithm to identify different research groups. Topic modeling is applied to extract topics for each community. Our results show that the coauthor network of Decision Science Systems (DSS) journal has the highest level of collaboration compared to coauthor networks of other journals. We also found that topics on the individual community level tend to be more specific compared to those on the overall level. In the future, we plan to study dynamics of networks and their implications.