Comparative analysis of scientific papers collections via topic modeling and co-authorship networks

Abstract

In this paper, the authors present an approach to benchmarking the collections of scientific journals based on the analysis of co-authorship graphs and a text models. The main methodical result is Comparative Topic Modeling (CTM) technique. The application of time series to the metrics of co-authorship graphs allowed trends in the development of author collaborations in scientific journals to be analyzed. A text model was created using machine learning methods. The content of journals was classified to determine the degree of authenticity both in various journals and their issues. Experiments was conducted on the archives of two journals in the field of Rheumatology. The authors used public data sets from the SNAP research laboratory at Stanford University to benchmark the co-authorship network metrics. The application of the research results is improving editorial strategies for development of co-authorship collaborations and scientific content excellence.