Research topic flows in co-authorship networks

Abstract In scientometrics, scientific collaboration is often analyzed by means of co-authorships. An aspect which overlooked and more difficult to quantify the flow expertise between authors from different research topics, an important part progress. With Topic Flow Network (TFN) we propose a graph structure for analysis topic flows their respective fields. Based on multi-graph model, our proposed network accounts intratopic as well intertopic flows. Our method requires construction TFN solely corpus publications (i.e., author abstract information). From this, topics are discovered automatically through non-negative matrix factorization. The thereof derived allows application social techniques, such common metrics community detection. Most importantly, it large, macroscopic scale, i.e., topic, microscopic certain sets authors. We demonstrate utility TFNs applying two comprehensive corpora altogether 20 Mio. spanning than 60 years in fields computer science mathematics. results give evidence that Networks suitable, e.g., topical communities, discovery fields, and, most notably, flows, transfer expertise. Besides that, opens new directions future research, investigation influence relationships