S2P2P: Semantic Search in Unstructured Peer-to-Peer Networks

In this paper, we present a semantic-driven k-walkers-based search scheme, called S2P2P, for data information dissemination and query routing in unstructured peer-to-peer (P2P) networks. In S2P2P, each peer maintains its observation on the semantics of received queries (demands) and data information (supplies), as well as a local view on network topology. On top of this, each peer, in line with forwarding a query, disseminates its known data information to a selected set of remote peers by taking advantage of query piggybacked data. For routing a query, each peer, instead of merely introducing an immediate neighbor or remote peer, suggests a query routing path containing a sequence of peers with expertise on the similar topic of query. This is achieved by a path suggestion heuristics that iteratively applies Dijkstra's algorithm in a greedy manner. Each iteration manages to detect one more expert peer and augments the current path suggestion with the shortest path from its tail to the detected expert peer. The comparative experimental evaluation shows that S2P2P outperforms a semantic flooding based search strategy in terms of search precision and recall. In addition, our evaluation reveals that S2P2P is as least as robust against the network dynamics than the semantic flooding approach.