Research On Query Expansion And Search Technology Based On Partially Decentralized P2P Network

The initial application of Peer-to-Peer (P2P) is to do the file Sharing, allowing any end-users (Peer) exchange files through the Internet. After a few years development, it has become the dominating application type of the internet traffic. P2P system's ability to support a large number of users has begun to show its technical advantages: it can rapidly deploy a powerful and massive distributional application with fairly low cost.Two crucial issues need to be addressed in P2P System: resource search and resource delivery. With its distributional storage characteristics, it has become relatively easy to implement a scalable system for resource delivery. The fundamental difficulty is how to find the necessary resources from the right peer, i.e., resource search. Unfortunately, the existing P2P's information search mechanisms all have a major weakness: the basis on which the recourse search is from depends on the mechanical match between the key words from the users'inquiries and the string words from the shared information. Due to the complexity of human language and the complicated difference from assorted users'education background and living habits, P2P system can not fully understand the users'intention resulting in ineffective inquiries.Based on an in-depth study on the various existing P2P search technology , this thesis chooses the partially decentralized topology as networking mode,and based on JXTA platform,designes and implements a concept-based query expansion P2P search prototype. With the goal of accuracy, efficiency, scalability and load balancing, it is the first time that WordNet, a semantic dictionary being introduced into the P2P searching network, combining advanced P2P network architecture and modern information retrieval techniques. WordNet is used to disambiguate keywords from the users'initial inquiries and then complete the semantic expansion, which can ultimately express the true intention of users'inquiries more accurately and comprehensively. More importantly, this prototype takes the advantage of the existence of super-peer from the partially decentralized topology network, caches the clients'source index on the super-peer and optimizes the forwarding of query message to satisfy the requirement of efficiency. In addition, it adopts some load balancing strategy to avoid the possibility of overloading of the pop peer in the system. Hopefully, this paper can provide some useful illustration for the future development of search technology.