Routing Policy Study On The Semantic Query Enabled P2P Network

With the advancement of researches of Semantic Web and Ontology, the requirements of semantic based information retrieval increased rapidly. On the other hand, the P2P network emerged as a great solution for file sharing and searching in distributed environments. To combine the technologies of Semantic Web and Ontology and P2P network shows to be valuable for P2P and semantic based information retrieval. How to use the semantic information to enhance the performance of the query routing in semantic based P2P network becomes a hot topic.Based on the study and analysis of the characteristics of semantic based information retrieval in P2P environments and reusing of other people's research results done in this area, this paper described a new P2P network architecture for semantic based information retrieval, which is called Semantic Clustering and Routing P2P Network (SCRN). The keyword vectors and ontology concept vectors are used as the features of documents and peers and peer groups and queries in SCRN. The peers which reposit similar documents are clustered into same peer group according to the similarities of peer vectors and peer group vectors. Every peer group has a super peer called group server, which creates and maintenances a route table storing other groups'vectors. The queries end users delivered are routed by server peers. It is done by means of the query and group vectors'similarity calculating and the next hops choosing.This paper discussed in details about the network topology initialization, the self-organized and automated network management, and the weight functions and similarity functions of vectors. A two layers search mechanism and two phases query method are argued to proceed the queries as also.The prototype system of SCRN was realized and experiments are done by deploying it in real distributed network PlanetLab and simulating by software. The results of the experiments proved SCRN architecture and the algorithmsproposed. Benefiting from using the semantic information in network topology building and query routing, the SCRN prototype system acquired a higher recall under the same restrict of query time and search scale, in comparison with the keywords based P2P network.