Speed up semantic search in p2p networks

Peer-to-peer architectures become popular in modern massively distributed systems, which are often in very large scale and contain a huge volume of heterogeneous data. To facilitate the information retrieval process in P2P networks, we consider semantic search approach, where syntax-based queries are shipped to peers based on semantic correlations. Motivated by an interesting experience in Web information retrieval, we propose a novel ontology-based scheme to measure similarity of peer interests accurately and consistently in a decentralized way, and group peers under a scalable hierarchical overlay network. Given queries, our approach either floods them within local peer groups or guides them towards remote groups based on the similarity of interests. Our work overcomes the limitations of the existing P2P hybrid-search approaches by avoiding costly data popularity measurement. Performance evaluation and comparison against baseline algorithms show that our approach provides a better solution for information retrieval in large-scale P2P networks.