Directed Networks of Online Chats: Content-Based Linking and Social Structure

Online chats are recently shown to result in long term associations among users, represented by a directed weighted network, similar to dialogs in online social networks. We consider the persistent network which emerges from user-to-user communications found in the empirical dataset from IRC Ubuntu channel. The structure of these networks is determined by computing topological centrality measures, link correlations and community detection, and by testing validity of the "social ties" hypothesis. To unravel underlying linking mechanisms, we further study type of messages exchanged among users and users with Web bots, and their emotional content, annotated in the texts of messages. We find that the ranking of the users according to the frequency of their messages obeys Zipf\´s law with a unique exponent for each message type. Furthermore, the specific hierarchical structure of the network with a strong core as well as its social organization are shown to be closely related with the most frequently used message types and the amount of emotional arousal in them.