Cluster the nodes in the graph into "topic regions". By (pairwise) calculation of the VSM similarity of URIs, relations can be assigned weights, and the nodes can be clustered.