Hyperedge Communities and Modularity Reveal Structure for Documents

Words and their association in documents could be represented by hypergraph in a standard way. Communities of words often overlap. Accepting a community should have more internal than external connections, we view every hyperedge as a vertex, establish a network for hyperedges by their similarity, and use the method of modularity to find their communities. The example here shows that email address' communities, generated by pulling back hyperedges communities, naturally incorporate overlap and reveal hierarchical organization.