Automatic Alignment and Extraction of Bilingual Domain Ontology for Medical Domain Web Search

This paper proposes an approach to automated ontology alignment and domain ontology extraction from two knowledge bases. First, WordNet and HowNet knowledge bases are aligned to construct a bilingual universal ontology based on the co-occurrence of the words in a parallel corpus. The bilingual universal ontology has the merit that it contains more structural and semantic information coverage from two complementary knowledge bases, WordNet and HowNet. For domain-specific applications, a medical domain ontology is further extracted from the universal ontology using the islanddriven algorithm and a medical domain corpus. Finally, the domain-dependent terms and some axioms between medical terms based on a medical encyclopaedia are added into the ontology. For ontology evaluation, experiments on web search were conducted using the constructed ontology. The experimental results show that the proposed approach can automatically align and extract the domain-specific ontology. In addition, the extracted ontology also shows its promising ability for medical web search.