OntoMiner: Bootstrapping and Populating Ontologies from Domain Specific Web Sites

HTML documents, which are designed primarily for human consumption. The presence of such legacy documents makes embracing the Semantic Web vision difficult.2 Thus, we need scalable solutions to automatically transform legacy HTML to Semantic Web documents. Recent work describes algorithms that automatically annotate HTML documents with semantic labels.3 Unfortunately, constructing the domain ontologies that drive these algorithms is human intensive. Bootstrapping and populating large, rich, and up-to-date domain ontologies that organize the most relevant concepts, their relationships, and instances (which correspond to members of concepts) can considerably enhance the scalability of transforming legacy HTML into Semantic Web documents. Our system, OntoMiner, offers automated techniques for creating such ontologies based on a small collection of relevant Web sites. Users can then employ the ontologies to create a rich set of labeled examples that a supervised machine-learning system such as WebKB can use.4