Information Retrieval on Socialized Peer-to-Peer Environment

博士 === 國立中正大學 === 資訊管理所 === 95 === Due to significant advantages of Peer-to-Peer (P2P) environment, Information Retrieval (IR) on P2P environment (P2PIR) is indeed potential to overcome the scalability problems of traditional approach to implement Internet Search Engine. However, for current approac...

Full description

Bibliographic Details
Main Authors: Hsiang-Yuan Hsueh, 薛祥元
Other Authors: Shi-Ming Huang
Format: Others
Language:en_US
Published: 2007
Online Access:http://ndltd.ncl.edu.tw/handle/71121868703239108710
Description
Summary:博士 === 國立中正大學 === 資訊管理所 === 95 === Due to significant advantages of Peer-to-Peer (P2P) environment, Information Retrieval (IR) on P2P environment (P2PIR) is indeed potential to overcome the scalability problems of traditional approach to implement Internet Search Engine. However, for current approaches, including overlay network construction by node-oriented and document-based mechanisms, the trade-off between scalability of overlay network construction among documents in peers and effectiveness of semantic gap elimination between users and resources is always a problem. It leads to a semantic gap between users’ query and search interpretation within P2P environment. Retrieving resources which are interesting, useful, relevant, and valuable to users is always challenge for researchers and vendors of IR. Under the research assumption and hypothesis that users tend to search with concept-based or interest-based manners, in this study, a socialized approach to systematically construct P2P environment is proposed for concept-based Information Retrieval activities. Constructing artifact, the authors proposed mechanism for feature discovery in single peer. Based on Zipf’s Law, the feature representatives as personal characteristics of single peer can be approximately extracted for document clustering. On the other hand, determining correlation among resources shared by single peer, documents can be properly aggregated as individual concept groups. To verify the feasibility of proposed artifact in this study, a set of specialized analyzers and standard measurement are developed for further simulation experience based on clear-defined verification goals. In simulation experience, a logical P2PIR environment is constructed by template data set which can properly reflect the large-scale P2PIR environments. Measuring factors defined in simulation experience, it can be obtained that the proposed artifact is effective and feasible in concept-based feature discovery and information retrieval in large-scale P2PIR environments. After the simulation experience, the authors also discussed some phenomenon discovered in simulation experience, including semantic and prestige gaps in concept-based P2PIR environments. Summarily in this study, feasibility of implementation for socialized P2PIR environment is demonstrated in this study. Verifying the technical effectiveness of proposed artifact, it is expected that socialized approach proposed in this study can bring benefits for both vendors and consumers in different types of IR system.