Emergence and structure of Cybercommunities

Abstract

We study topology of bipartite networks representing high-resolution data of the online communications of users on Blogs and similar Web portals. User communities occurring in connection with certain popular posts, movies etc., are detected by spectral analysis of these networks. Due to indirect nature of the online interactions between users, further information about the structure of the communities is inferred by text analysis of the related comments. We employ the emotion classifier based on machine-learning methods and trained for this type of data, to determine the emotional contents of text of each post and comment within a given community. Combined with the network theory, in this way we are able to unravel the role of emotion expressed in the text for the patterns of user behavior, which leads to the emergence of collective states with the appearance of communities, and their internal structure and evolution. All data are fully anonymized. No information about user IDs are given.