Problem-solving methods and cooperative information agents

Cooperative Information Agents and modern information systems in general have to access large amounts of information distributed across multiple heterogeneous sources. A great challenge of such systems is to evolve by adding new information sources or adapting the existing components for different domain knowledge. We propose the UPML framework as a methodology to build Information Agents by reusing a library of problem-solving components that are defined in a domain-independent manner. Moreover, the UPML language is used as an Agent Capability Description Language (ACDL) suitable to configure an agent-based application. From this approach, a new application is built by linking the components of the library with a particular domain and a collection of heterogeneous information sources. Adaptability and dynamic configuration of such a system is achieved by reasoning about the UPML specification of the agent capabilities. Independence of the domain and semantic interoperability are achieved by using ontologies and bridges (mappings between ontologies), while independence from the information sources is based again in the use of ontologies to overcome semantic heterogeneity and wrappers to achieve syntactic interoperability.