Efficient query routing for information retrieval in semantic overlays

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