On-line Learning Of Domain Ontologies From The Web

As a formal, explicit specification of a shared conceptualization, Ontology places an important role in many areas, including knowledge engineering, electronic commerce, natural language processing, et al. However, for a long time, Ontology is constructed by knowledge engineer and domain experts manually, which is time-consuming, error-prone and inclined to be subjective. As a result, the concept of Ontology Learning is proposed, which seeks to learn Ontology from various data sources automatically or semi-automatically. Up to now, Web is the biggest wide-ranging information source, but it lacks necessary semantic information for computer automatic processing which is the aim of the Semantic Web in the future. Therefore, it’s an urgent and significant task to mine potential useful information from the Web at present. The relative work and contributions mainly include:(1) Firstly, perform a comprehensive analysis of related theories of Ontology and Ontology Learning involving relevant problems, technologies and status and so on. Then a novel architecture of on-line learning of domain Ontologies from the Web is proposed.(2) Semantic focused crawler is adopted for collecting domain related Web resources putting special emphasis on discussion principles of computing semantic theme relativity and conducting semantic ambiguity elimination based on Ontology.(3) Afterwards, give a deep and comprehensive description on how to learn domain concepts from the Web, ranging from concept terms, data properties to concept instances.(4) Finally, considering the distinguishing features of the Web, explore methods of learning taxonomies and relations between domain concepts from multiple perspectives covering hyperlinks, web categories and free texts, aiming at overcoming the one-sidedness of applying single method.