Interoperability and semi-structured data in an open web-based agent information system

The World Wide Web is a vast resource of information and services that continues to grow rapidly. How to find information providers and how to integrate information agents in such an open environment are new challenges. In this paper we present an open multi-agent system, SportsAgents, for information gathering from the World Wide Web. In this system middleware agents are introduced to solve the interoperability problem among information agents. Novels in SportsAgents are an intelligent matchmaking algorithm for inferring the service relationships among information agents, and a Naive Bayesian model for information agents cooperation forming is given. Information agents use score patterns to extract semi-structured sports results. All these makes the system open and scalable to new applications and flexible to organise agents cooperation.