Designing a Super-Peer semantic Network based on Hierarchical Clusters

Abstract

In super-peer semantic network, when a new peer joins the network, the peer will use the semantic feature of local share documents to select super-peers. In traditionally networks, client-peers use the semantic feature of clusters in peers to select the super-peers to connect. But the clusters are fixed, general, single-level, and not effective to represent the features of all documents. And, while selecting super-peers to connect, the client-peer cannot select some relative clusters in different level accord to the semantic categories of super-peers semantic group. In this paper, a Super-Peer semantic Network based on Hierarchical Clustering Tree (HCTSPN) is presented. First, an improved hierarchical clustering algorithm for organizing the share documents in a client-peer is proposed. Secondly, the method of constructing a super-peer semantic network based on hierarchical clustering tree and search mechanism are proposed. Finally, the experiments are implemented, and the results show that search efficiency and retrieval quality are improved in the network.