In many applications of peer-to-peer networks, such as information sharing and media streaming, a peer is much more likely to communicate with other peers with similar interests. Therefore, to optimize the structure of these systems, a good strategy is to form communities composed of peers with similar interests. In this paper, we propose a method to automatically identify a user?s interests, and a distributed algorithm to discover communities based on the interests identified. We provide empirical results to demonstrate the correctness and efficiency of our algorithm.