Detection of a core-periphery structure in bipartite user-content networks based on modularity and Stochastic Block Model.

In this paper, we study the patterns of interaction of the general audience with a sample of Italian Twitter accounts representing various organizations. We model the interaction between the audience and the content produced by each account as a bipartite network in which the nodes are on the right side the tweets of the official account and on the left side the users in the audience, and edges represent direct interactions such as retweets and replies. First, we show how these networks behave very differently from the usual null models for bipartite networks with highly inhomogeneous degrees such as the bipartite configuration model, uniform bipartite graph and the 'soft' configuration model. In particular, we show that these networks exhibit a bipartite core-periphery organization using a Degree Corrected Stochastic Block Model. Finally, through a bipartite modularity optimization, we study how this core-periphery organization reflects in the community structure within each network and find that most communities are made up of one or a few star graphs around either a few core users or a few core tweets.