Using the Semantics Inherent in Sitemaps to Learn Ontologies

The Semantic Web is a collection of components that work together so that a machine is able to process and understand information. In order for this vision to be fully realized, formal standards for representing and interpreting data must be determined and enforced. These formal standards include the Resource Description Framework (RDF) and machine processible ontologies. On the one hand, crafting an ontology manually is a difficult, time-consuming, and labor-intensive task. On the other hand, an automatically generated ontology may suffer from poor quality, including poor semantics. We are proposing a semi-automated approach in ontology development, an approach that limits the disadvantages of each above mentioned method when used in isolation. The Semantic Web is an extension of the current web. The current web itself has rich semantics, though not in a machine-readable format, and an important part of the current web is sitemaps. HTML sitemaps are useful in helping users to easily navigate websites, and they have rich taxonomic information. Thus, sitemaps could be of significant value in ontology development. Our approach evaluates related sitemaps by focusing on mining semantic information from these sitemaps and then using this information to help create ontologies. Given a set of HTML sitemaps from one domain, the proposed approach will first identify taxonomic relations to start building an ontology and then enhance the ontology by enrichment and refinement. The approach has been evaluated on several domains. The empirical experiments indicate that the approach is effective.