| Ontologies are formal representations of knowledge. They typically are comprised of a database of terms and information about these terms. This information may include descriptions of how these terms are related to each other. Such information, if available, can be used to determine how similar, or closely related, two terms are. Graph-based data mining is the technique of finding interesting patterns in a database represented as a graph. This process can be improved if we have some additional information on the various terms occurring in the database.; This study presents Subdue, a graph-based data mining technique and its use in conjunction with ontologies to better the process of knowledge discovery. We introduce a distance metric that can be used to quantize the proximity of two terms according to the information provided by an external ontology. In this work we integrate two ontologies, WordNet and CYC, with the Subdue graph-based data mining system. We validate the results of this integration with a series of experiments to examine the effects of ontology-based graph match on resulting graph match cost, cluster goodness, and the quality of discoveries generated by Subdue. |
