Ontology-learning Supported Sematic Search Using Cooperative Agents

In this paper we present (1) a method for semantic search supported by ontological concept learning; and (2) a prototype multiagent system that can handle semantic search and encapsulate the complexity of such process from the users. Agents which conduct semantic search on behalf of a user, deploy ontologies to organize structured and unstructured documents in their corresponding repositories. The ontology for each repository is individualized and commitment to a common ontology is not required. The agents can improve their search capability by learning new concepts from each other. This method thus allows agents dynamically establish common grounds on concepts known only to some of them. The concept learning is realized by analyzing positive and negative examples from other agents, and/or taking votes in case of conflicts in the received knowledge by involving other agents again. Index Terms — multi-agent system, semantic search, ontology, concept learning, semantic interoperability.