No abstract

No abstract

No abstract

No abstract

A fundamental problem in peer-to-peer networks is how to locate appropriate peers efficiently to answer a specific query request. This paper proposes a model in which semantically similar peers form a semantic overlay network and a query can be routed or forwarded to appropriate peers instead of broadcasting or random selection. We apply Latent Semantic Indexing (LSI) in information retrieval to reveal semantic subspaces of feature spaces from documents stored on peers. After producing semantic vectors through LSI, we train a support vector machine (SVM) to classify the peers into different categories based on the extracted vectors. Peers with close categories are defined as semantic similarity and form a semantic overlay. Experimental results show the model is efficient and performs better than other non-semantic retrieval models with respect to accuracy. In addition, our approach improves the recall rate nearly 100% while reducing message traffic dramatically compared with Gnutella.

No abstract

No abstract

No abstract