Mediation of information sharing in cooperative federated database systems: Ontologies, mediators, and data mining

A Cooperative Federated Database System (CFDBS) is an information sharing environment, in which units of information to be shared are substantially structured and participants are actively involved in information sharing activities. The problem of our focus is to build a common ontology for the purpose of assisting information sharing in the CFDBS context. We introduce the concept and mechanism of the Dynamic Classificational Ontology (DCO), which is a collection of concepts and inter-relationships to describe and classify information exported by participating information providers.; A DCO consists of a Base Ontology and a Derived Ontology. The Base Ontology is an application-specific collection of concepts and their relationships that are agreed among participants and used by them for description and classification of shared information. The Derived Ontology contains additional knowledge of associations among some concepts in the Base Ontology, which helps better understanding of the classification. It is derived from the Base Ontology and the collection of exported information. The process of deriving concept associations is closely related to mining generalized association rules which has attracted much attention recently. We have developed an efficient algorithm for this process, and showed its superior performance over other algorithms for our library application Medline.; The advantages of our approach to managing a common ontology is that participants are not required to reach a total agreement on the ontology, but it is allowed to change as the usage of involved concepts changes. In particular, the DCO maintains domain specific and dynamically changing classification schemes, by contrast with fixed hierarchical classifications. We will show how the DCO can be utilized by mediators for effective information sharing in the CFDBS environments.