Node Interest Based Information Retrieval Mechanism Research And Implementation In Peer-To-Peer Network

Peer-To-Peer (P2P), an innovational network technology, is regularly coming intopeoples live and industry. Its utilization is more prevalent in several fields, such asfile-sharing, distributed computing and cooperation work. The information amount risesrapidly with the intense increment of both users' quantity and network's scale and caused aserious problem, difficulty in searching, locating and retrieving resource. To conquer thisproblem, information retrieval in P2P network is a pivotal technique. A good strategy not onlyincreases hit ratio, but also serves for users in studying relevant information to improve theperformance in later search process. The ability to decrease node load and network cost isanother goal a good search algorithm seeks. Consequently, the information retrieval in P2Pnetwork is a valuable research direction.Due to the inability to control the whole resource in network, P2P information retrievalhas to stick in the mud of users' repeating work, changing keywords and the absolutekeyword matching. Lacking sufficient and desirable results is understandable in this situation.Our work is focusing the phenomenon of frequent keyword changing operation, presenting aninformation retrieval mechanism based on keyword relationships and node interests. Throughthe mechanism, the relationships between keywords are studied; the node possible interestsare discovered. In later search process, based on the keyword relations, expansion of theoriginal query keyword increases the hit possibility, based on the node interests, limitation ofsearch scope reduces the network cost and improves the search efficiency. Besides, in order toenhance the performance, the rapid table construction and feedback mechanism is appended.At last, author presented a desirable experimental result, a reasonable comparison and provedthe validity and efficiency of the algorithm.Traditional information retrieval is based on keyword, apparently, which is not perfectlysuitable for people's increased need. Standing on this point, a Contend-Based informationretrieval algorithm is presented, which is mainly used for audio file retrieval, namedcluster-based information retrieval strategy. The audio feature extraction is the first step, mainpursuit of this part is including the implementation of Short-Time Fourier Transform (STFT)algorithm and several feature extraction operation (entropy, centroid, centroid ratio,bandwidth, silence ratio, energy ratio, and location of minimum and maximum energy). Thelater work are with the feature vector extracted before performing vector calculation, building the local file cluster, constructing node cluster among interest contacts. After pretreatment, aconcrete search mechanism is illustrated. To achieve a better performance, two improvingmethods are proposed, dynamic CIT update and optimized cluster-cluster node. At the end ofthis part, the experimental result and relevant charts are listed, with the discussion andanalysis, the conclusion that the cluster-based information retrieval strategy is practical andaccurate in audio file retrieval is proved strongly.