A knowledge-based approach to domain-specialized information agents

Finding specific information on the Web can be difficult and time-consuming due to the Web's tremendous size and the current speed of network connecti ons. Information agents that search the Web on behalf of users are invaluab le. Since information needs are diverse, there should be tools for users to build their own agents that specialize in finding the information they wan t. These tools should be usable by non-programmers. This paper describes AR IS, a shell for constructing information agents without coding. Our approac h is akin to expert system shells, where the user supplies the knowledge fo r each specialist agent, and a common engine utilizes the agents' knowledge . An agent's search is guided by knowledge about how Web sites in a given d omain are typically structured. We also report on our experiences with two agents built using ARIS.