Co-mention network of R packages: Scientific impact and clustering structure

Abstract

Abstract Despite its rising position as a first-class research object, scientific software remains a marginal object in studies of scholarly communication. This study aims to fill the gap by examining the co-mention network of R packages across all Public Library of Science (PLoS) journals. To that end, we developed a software entity extraction method and identified 14,310 instances of R packages across the 13,684 PLoS journal papers mentioning or citing R. A paper-level co-mention network of these packages was visualized and analyzed using three major centrality measures: degree centrality, betweenness centrality, and PageRank. We analyzed the distributive patterns of R packages in all PLoS papers, identified the top packages mentioned in these papers, and examined the clustering structure of the network. Specifically, we found that the discipline and function of the packages can partly explain the largest clusters. The present study offers the first large-scale analysis of R packages’ extensive use in scientific research. As such, it lays the foundation for future explorations of various roles played by software packages in the scientific enterprise.