Research On Key Technologies Of Information Retrieval In Unstructured Peer-to-Peer Network

This dissertation focuses on the problems of information retrieval in the unstructured peer-to-peer (P2P) network. At first, this dissertation analyzes systematically the current key technologies and their shortages in the field of unstructured P2P network. Then a novel information retrieval model in unstructured P2P network based on mobile agent has been proposed. This model makes full use of the autonomy and heterogeneity of mobile agent, which can reduce overfull network connections and responding in the traditional information retrieval model. This model introduces lexical chain into the information retrieval of P2P network to improve the pheromone of ant colony algorithm by information recommendation, and speed up P2P query routing by using ant colony algorithm based on information recommendation and InformPSO algorithm respectively. These core issues in the field of information retrieval in the unstructured P2P network have great theoretical significance and value of practical applications. The main contributions of this paper are summarized as below:(1) This dissertation proposes a novel information retrieval model in unstructured P2P network based on mobile agent. In the traditional unstructured P2P network, the message mechanism of information retrieval requires the sender and receiver keep connection until the responding happens, which is bandwidth-intensive. To overcome the shortages of traditional model, the proposed model makes full use of the autonomy and heterogeneity of mobile agent that can reduce the overfull network connections and responding.(2) An information retrieval method in unstructured P2P network based on lexical chain has been introduced, which implements the semantic retrieval under the unstructured P2P network. The proposed work makes full use of the WordNet, a famous lexical database of English, and constructs vectors by lexical chain instead of traditional vectors composed of index terms to represent texts. Further more, the method of calcuating lexical chain has been proposed to resolve the polysemy and synonym problems, and process the relationship between words to improve the ability of vectors to express topics. To understand user's query more accurately, the lexical chain query template updating algorithm based on evolutionary algorithm has been designed. After using lexical chain, the system can accept semantic and various granularity query from user in the unstructured P2P network.(3) This dissertation proposes an advanced ant colony algorithm based on information recommendation, the proposed model collects more information of peers in the P2P network and make query routing more accurately than traditional P2P network. Because query routing in the unstructured P2P network is similar to the process of foraging ants, the ant colony algorithm can simulate query process very well. In order to obtain more information of peers, the proposed model takes information recommendation as a fact of pheromone updating in ant colony algorithm, so that the model realizes bidirectional information exchanging instead of one directional query in the traditional P2P network. It can speed up information exchanging and leave behind the recommender peer's information to the past peers during the process of recommendation so as to reduce network congestion and improve query performance in P2P network.(4) An improved PSO algorithm, called InformPSO, has been proposed. The performance of this algorithm has been verified from theory and simulated experiments and applied in the query routing of unstructured P2P network to improve the accuracy of query. At the beginning of system running, the ant colony algorithm works to speed up routing, during the process each peer will collect other peers'information and construct a limited P2P network which includes the peers known by the constructer. After a period of time, the peers obtain many other peers'information and the query routing can be considered as the shortest path problem, the InformPSO algorithm can find a best path in each peer when an agent visits this node and direct the agent.After all, to improve the performance of information retrieval, semantic search and query routing in the traditional unstructured P2P network, a novel information retrieval model in unstructured P2P network based on mobile agent has been proposed. This model uses lexical chain, implements automatic information recommendation, and improves query routing by using ant colony algorithm and advanced PSO algorithm. The simulated experiments indicate that our methods can improve the effect and be expected to bring practical value.