PIRCS: a link context based search on the Web

In this paper, we propose a personal information retrieval system using cooperative servers (PIRCS), which performs link context based searches with personalized retrieval settings. PIRCS works efficiently on cooperative Web architecture Kobayashi Aki et al. (2002), and searches take less time. The system is developed for use with search services that treat the Web as if it were simply a distributed network storage system without a central manager. We call these systems "super-distributed" database systems. A client of a super-distributed database retrieves information from many distant nodes, and the node information is updated autonomously. To improve the processing efficiency, the client can pass on a part of the retrieval process to other agents in the network. We aim at constructing such a super-distributed database by using the cooperative Web architecture, we developed where each Web content on an advanced Web server (AWS) knows certain linked contents, and where the AWS can pass on the link information to a client. The client, a Web browser with a search function (WBS), constructs a logical Web space for an information search with the link information. Target information is retrieved from that space by the WBS. We developed a prototype of the PIRCS super-distributed retrieval application on our cooperative Web. The WBS has an intuitional user interface, which shows the link space of the retrieved contents as a graph. In the PIRCS, the final retrieval process is done with a search engine built into the user agent (WBS). This architecture provides an optimum design for balancing the retrieval process between the server and the client. Moreover, as the retrieval is based on up-to-date information, it is more efficient than that of conventional retrieval tools, which gather content information beforehand.