social software: automatic relationship clustering is a very interesting article over on d2r. It discusses using graph clustering techniques to identify groups within a social network. Diego suggests that adapting this to use FOAF ought to be straight-forward.
As I noted in my critique of XFN plain old foaf:knows
is a fairly limited relationship. It’s be more interesting to discover other kinds of relationships based on event attendance, collaboration, common interests, etc. This technique ought to be applicable to that kind of exploration as it deals only with nodes and edges; it’s just a matter of presenting the right nodes and edges to the algorithm.