Realization of Semantic Search Using Concept Learning and Document Annotation Agents.
Behrouz H. Far, Cheng Zhong, Zilan Yang, Mohsen Afsharchi · Software Engineering and Knowledge Engineering · 2009
Currently, search systems are based on commitment to a common ontology. In the real world, it is preferred to enable Web repositories to exchange information freely while keeping their own ontology. This helps contents providers to represent the information independently in the repositories at the expense of bringing complexity to the communication and negotiation. To solve the communication complexity problem we present (1) a method for semantic search supported by ontological concept learning, and (2) a prototype multi-agent system that can handle semantic search while encapsulating complexity of such process from the users. The method introduces a spiral search process and a layered structure of semantic interoperability. Agents, which conduct semantic search on behalf of users, deploy ontologies to organize documents in their corresponding repositories. Through a detailed experiment we will show that agents can improve their search capability by learning new concepts from each other, and consequently, provide better search results to the users. Index Terms — multi-agent system, semantic search, ontology, concept learning, interoperability, annotation.