Analyzing Social Media Relationships in Context with Discussion Graphs

We present discussion graphs, a hyper-graph-based representation of social media discussions that captures both the structural features of the relationships among entities as well as the context of the discussions from which they were derived. Building on previous analyses of social media networks that focus on the strength of relationships between entities, our discussion graphs explicitly include contextual features such as who is participating in the discussions, when and where the discussions are occurring, and what else is being discussed in conjunction. There are two contributions of this work. First, we extend standard hyper-graph representations of networks to include the distribution of contexts surrounding discussions in social media networks. Second, we demonstrate how this context is useful for understanding the results of common graph measures and analyses, such as network centrality and pseudo-cliques, when applied to the analysis of textual social media content. We apply our framework across several domains captured in Twitter, including the mining of peoples' statements about their locations and activities and discussions of the U.S. 2012 elections