Dynamic Data Fusion Using An Ontology-Based Software Agent System
Mark Elmore, Thomas E. Potok, Frederick T. Sheldon · 2008
Developing a knowledge-sharing capability across distributed heterogeneous data sources remains a significant challenge. Ontology-based approaches show promise by resolving heterogeneity, if the participating data owners agree to use a common ontology (i.e., a set of common attributes). Such common ontologies offer the capability to work with distributed data as if it were located in a central repository. This knowledge sharing may be achieved by determining the intersection of similar concepts from across various heterogeneous systems. However, if information is sought from a subset of the data sources, there may be concepts common to the subset that are not included in the full common ontology, and therefore are unavailable for knowledge sharing. We offer a novel ontology-based software agent approach that provides flexible and dynamic fusion of data across any combination of the participating data sources to maximize knowledge sharing. The software agents generate the largest intersection of shared data across any selected subset. This approach maximizes knowledge sharing by dynamically generating common ontologies over the data sources of interest. Data provided by five national laboratories was used to validate this approach by defining a local ontology for each. In this experiment, the ontologies are used to specify how to format the data using XML to make it suitable for query. Consequently, software agents provide the ability to dynamically form local ontologies from the data sources. In this way, the cost of developing these ontologies is reduced while providing the broadest possible access to available data sources.