Medical Ontology Learning Based on Web Resources

In order to deal with heterogeneous knowledge in the medical field, this paper proposes a method which can learn a heavy-weighted medical ontology based on medical glossaries and Web resources. Firstly, terms and taxonomic relations are extracted based on disease and drug glossaries and a light-weighted ontology is constructed, Secondly, non-taxonomic relations are automatically learned from Web resources with linguistic patterns, and the two ontologies (disease and drug) are expanded from light-weighted level towards heavy-weighted level, At last, the disease ontology and drug ontology are integrated to create a practical medical ontology. Experiment shows that this method can integrate and expand medical terms with taxonomic and different kinds of non-taxonomic relations. Our experiments show that the performance is promising.