The exchange of retrieval knowledge about services between agents
Recently, several work has been done in hybrid case-based reasoning (CBR) systems [Ah98] [Le99]. Hybrid CBR systems are case-based systems that are combined or integrated with either one or more other AI methodologies (for example neural networks or rule-based systems) [Le99]. Some systems combine agents with case-based reasoning: Models of the users in form of profiles are managed as cases [MLR02] or as sets of cases [CCH02]. [WL98] describe how CBR can be used for situation assessment in a multi-agent soccer system; [RA98] performs e-mail filtering with an agent using CBR as a service. In this article, we present a hybrid system of assistant agents using textual case-based reasoning (TCBR) for a smart retrieval of services. The service library of an agent is regarded a case base. As agents are allowed to exchange services, the exchange of parts of their retrieval knowledge - namely information entities and local similarity relationships - has to be discussed and implemented. The integration of the received tiny case retrieval nets with the own retrieval structures is a crucial task of our work described here. Slightly different duplicates of information entities and similarity relationships have to be handled as well as inconsistently assigned strings. In the future, semantic aspects of the harmonization of merged background knowledge could be taken into account as it has been done in the ontology community.
Full Text: PDF