Document Retrieval from Multiple Collections by Using Lightweight Ontologies

This paper proposes the use of lightweight ontologies to describe documents collections. A meta harvesting protocol is used to access the documents. Ontologies are constructed semi-automatically by means of data mining techniques and document clustering algorithms. They are represented in Extensible Markup Language. Resource description framework is used to formalize a unique interpretation of the semantic of their elements. We describe a scenario where software agents use the ontologies to support information retrieval tasks from multiple collections and we show some experimental results