Information Agents That Learn to Understand Each Other Via Semantic Negotiation

A key issue in Distributed Applications, that widely use Information Agents for implementing several typologies of services, is that of making reciprocally understandable the meaning of terms contained in the exchanged messages, in those cases where agents use different, heterogeneous ontologies. A possible way for facing this issue is offered by the semantic negotiation, a framework in which agents try to understand each other by negotiating the semantic of the terms. Several models and protocols of semantic negotiation have been proposed in the last years. However, most of these approaches are not able to support semantic negotiation without requiring agents either to share knowledge or to use a global common ontology, and none of them provides a semantic negotiation protocol that allows the whole agent community to contribute to the semantic understanding process between each agent pair. In this work, we propose the HIerarchical SEmantic NEgotiation (HISENE) protocol, based on the idea that an agent a should be able to partition the set of the other agents on the basis both of their personal expertise of the application domain, as well as on the particular capability that each of them shows in understanding a. We also give an implementation of the proposed protocol in the standard Java Agent DEvelopment Framework (JADE).