Feature selection and complex networks methods for an analysis of collaboration evolution in science: an application to the ISTEX digital library

Nicolas Dugué (1) , Ali Tebbakh (1, 2) , Pascal Cuxac (2) , Jean-Charles Lamirel (1)

1 SYNALP - Natural Language Processing : representations, inference and semantics
2 INIST - Institut de l'information scientifique et technique

In this paper, we aim to give insights about the self-organization of scientific collaboration. To that aim, we describe a new framework to monitor the evolution of a collaboration graph that models the co-authorship of research papers authors. We use community structure of the network as a high-level description of its self-organization and thus consider the evolution of the communities across time. To monitor this evolution, we describe a diachronic analysis method based on the extraction of prevalent nodes for each community. We apply this approach on data issued from the ISTEX project, a scientific digital library that contains so far more than 16 million documents and present some preliminary results and visualizations.

Mots clés: Co-authorship evolution; Diachronic analysis; Communities; Dynamic graphs; Feature selection; Complex networks

Domaines: Réseaux sociaux et d'information [cs.SI]; Apprentissage [cs.LG]; Intelligence artificielle [cs.AI]; Web

Nicolas Dugué, Ali Tebbakh, Pascal Cuxac, Jean-Charles Lamirel. Feature selection and complex networks methods for an analysis of collaboration evolution in science: an application to the ISTEX digital library. ISKO-MAGHREB 2015, Nov 2015, Hammamet, Tunisia. ⟨hal-01231791⟩