A Web-Based Method for Ontology Population
The Semantic Web, proposed by Berners-Lee, aims to make explicit the meaning of the data available on the Internet, making it possible for Web data to be processed both by people and intelligent agents. The Semantic Web requires Web data to be semantically classified and annotated with some structured representation of knowledge, such as ontologies. This chapter proposes an unsupervised, domain-independent method for extracting instances of ontological classes from unstructured data sources available on the World Wide Web. Starting with an initial set of linguistic patterns, a confidence-weighted score measure is presented integrating distinct measures and heuristics to rank candidate instances extracted from the Web. The results of several experiments are discussed achieving very encouraging results, which demonstrate the feasibility of the proposed method for automatic ontology population.