Collaborative indexing over networked information resources by distributed agents

Abstract

Various resource discovery tools appeared to help users locate and retrieve information resources over the Internet. The amount of information resources is increasing dramatically, and there is a growing demand for the capability of dealing with the large scale of information resources. Two metaphors are being used for resource discovery: navigation and search. Although a few indexing systems exist to support search over each navigation tool, there is no single indexing system for global search over entire information space by supporting various navigation tools. The difficulty of global search system is caused by three factors: (1) the large scalability of the global information space, (2) heterogeneity of resource organizing structure and resource format by diverse tools, and (3) a large part of knowledge on where to start search depends on user``s heuristics. Most of existing indexing systems periodically execute global exhaustive search over the entire set of resources to collect index data. This can cause excessive load to the servers. Also, the organizing information of each navigation tool is not incorporated in the index data. Mechanical indexing over resources is also not enough. Thus indexing systems need to gather information from the network users`` knowledge as well as mechanical index data to categorize which information sources are useful for which topic. These problems lead to the idea of distributed indexing. Each index system should focus on a specific topic or a specific community, and a user should be able to find out a right indexing system among a large number of distributed indexing systems. This thesis proposes an agent model for global indexing by partial search on selected resource domains. The feasible set of information is selected by trading between distributed agents. Each trading agent has its own topic andkeep metainformation on the topic with weight values. The organizing information of navigation tools are represented by weight...