Authors: Jacques Calmet, Sebastian Jekutsch, Peter Kullmann, Joachim Schü.

Abstract We present KOMET, an architecture for the intelligent integration of heterogeneous information sources. It is based on the idea of a mediator, which is an independent software layer between an application and various knowledge sources which need to be accessed. We present an especially suitable logic-based language for encoding typical mediation tasks like conditional preference strategies, schema integration or data inconsistency resolution. Using annotated logic, KOMET is able to perform various common types of reasoning, such as probabilistic, fuzzy, paraconsistent and certain types of temporal and spatial reasoning. In combination with an extensible type system and the embedding of external knowledge sources as constraint domains, our mediation language offers a rich framework, which not only facilitates access to structured information, but as well supports unstructured and semi-structured information. A number of examples show the practical application of our approach.