Ontology Discovery for the Semantic Web Using Hierarchical Clustering

Abstract

According to a proposal by Tim Berners-Lee, the World Wide Web should be extended to make a Semantic Web where human understandable content is structured in such a way as to make it machine processable. Central to this conception is the establishment of shared ontologies, which specify the fundamental objects and relations important to particular online communities. Normally, such ontologies are hand crafted by domain experts. In this paper we propose that certain techniques employed in data mining tasks can be adopted to automatically discover and generate ontologies. In particular, we focus on the conceptual clustering algorithm, COBWEB, and show that it can be used to generate class hierarchies expressible in RDF Schema. We consider applications of this approach to online communities where recommendation of assets on the basis of user behaviour is the goal, illustrating our arguments with reference to the Smart Radio online song recommendation application.