Visualizing Overlapping Latent Communities Using POI-Based Visualizations

Social network analysis and social network visualizations can provide a meaningful statistical and topological understanding of latent communities. However, the majority of current visualization approaches just represent sub-communities as clusters of closely related nodes in a node-link diagram and embed limitations to represent overlapping communities and multi-layer community structure frequently found from modern complex networks. We argue that visualizations based on points of interest can provide a better solution to represent overlapping latent sub-communities. We present two visualization systems, SuperVIBE and ContextForces, which implement this approach. These systems operate by creating two-dimensional latent spaces by means of grouping nodes using external variables not presented in the graph and by offering an interactive visualization to filter and map in these latent spaces. Understanding which latent groups are most central to a variety of topics and providing visual clues to the individuals critical to those groups provides a mechanism to explore and discover overlapping latent communities.