Profiling user interactions on online social networks.

Over the last couple of years, there has been signi_cant research e_ort in mining user behavior on online social networks for applications ranging from sentiment analysis to marketing. In most of those applications, usually a snapshot of user attributes or user relationships are analyzed to build the data mining models, without considering how user attributes and user relationships can be utilized together. In this thesis, we will describe how user relationships within a social network can be further augmented by information gathered from user generated texts to analyze large scale dynamics of social networks. Speci_cally, we aim at explaining social network interactions by using information gleaned from friendships, pro_les, and status posts of users. Our approach pro_les user interactions in terms of shared similarities among users, and applies the gained knowledge to help users in understanding the inherent reasons, consequences and bene_ts of interacting with other social network users.