Research Of P2P Information Retrieval

P2P technology has become the hot spot of research recently. It has good fault- tolerance ability and scalability and otherwise. Resources locating is a key issue of P2P research. Although P2P network has been used in a variety of applications successfully those years, How to search efficiently and scalability in large-scale P2P overlay, which is still an open problem. Flooding search mechanisms is adapt by Unstructured P2P network adopts,It has high stability and support fuzz search ,but it has low efficiency and bad extensibility and locating the rare resource difficulty. Structured P2P network which bases distributed hash table(DHT) can provide good search efficiency and scalability, moreover,it is more suitable for P2P information retrieval in large-scale network. While DHT-based structured P2P networks fail to support flexible multi-keyword search.The key technologies of the P2P information retrieval is introduced in this paper and it emphatically analyzes the core mechanism of DHT-based structured P2P network, Moreover it introduces some methods for support fuzz search in DHT-based structured P2P network which is divided into two main categories: keyword search and semantic search.And then, analyzing the muti-keywords search in structured P2P information retrieval. In order to cope with such problem of bandwidth consumption. A indexing framework of the set of keywords strategy is proposed. It adopting truncate posting lists associated with indexing features to a constant size and extend the set of indexing features with carefully chosen feature item sets of keywords combination. Theoretical analysis and experimental results shows it guarantee an acceptable bandwidth consumption, the system has good scalability.Finally, According to semantic search in structured P2P retrieval, an effective information retrieval of structured P2P is designed, Technologies of Vector Space Model and Local Sensitive Hashing are adapt. The basic idea is to place data of semantically close files into same peer nodes with high probability. It speeds up search process. Experimental results shows the validity of this system.