Agents for information gathering

Abstract

With the growing number of information sources available through networks, the problem of how to combine distributed, heterogeneous information sources is becoming increasingly critical. The available sources include traditional databases, flat files, knowledge bases, programs, and so forth. The solution to this problem is to provide access to the large number of information sources by organizing them into a network of information agents. Each agent provides expertise on a specific topic by drawing on relevant information from other information agents. To build such a network, we need an architecture for a single agent that can be instantiated to provide multiple agents. Our design is based on our previous work on the SIMS system (Arens et al., 1993; 1996), an information mediator that provides access to heterogeneous data and knowledge bases. We need to consider several aspects that are critical for any agent-based system: agent organization, knowledge of an agent, communication language and protocol, query processing, and learning capabilities. We briefly discuss each of them in turn.