Mapping between Ontologies in Agent Communication

One of the most important issues in the area of agent communication is the understanding of message content meaning. The meaning is captured in message ontology. In the "open" multi-agent systems we cannot guarantee that all agents share the used ontologies completely. To enable communication, messages must be translated from one ontology to the other one. We propose an algorithm for establishing a mapping between ontologies and describe how the mapping is used for translation of messages between different ontologies. The approach relies on a possibility to point to objects in the world that agents share. The process of refining the mapping is iterative and knowledge gained at any step can be used for translation. The proposed algorithm permits a certain information loss (e.g. when the source ontology is more detailed than the target one) but always provides a consistent translation.