Extracting communities of interests for semantics-based graph searches

We propose an algorithm that divides a graph, whose vertices are users and whose edges represent shared interests, into subgraphs - communities of interests (COIs). Our algorithm differs from conventional modularity-based community extraction because it introduces a mechanism to harmonize the number of users in each COI. It also attaches semantic tags to the relationships between COIs or between users based on taxonomies of the content items of interest. Our algorithm allows users to identify new interests quickly and accurately when searching their interests in the graph.