Gastroenterology Ontology Construction Using Synonym Identification and Relation Extraction

Ontology plays an increasingly important role in knowledge management and the semantic Web. However, ontology cannot perform well in realistic diagnosis reasoning unless it contains timely and accurate medical information and its individual items display all attributes of the categories they belong to. In this paper, we present a method that extracts synonyms along with concepts and their relationships for gastroenterology ontology construction. Specifically, we reuse the existing ontology as the basis for ontology completion. In addition, we conduct synonym identification through a combined application of global context features, local context features, and medical-specific features, and incorporate dependency information into deep neural networks for relation extraction. The extracted information is merged for ontology completion. Experimental results demonstrate that the proposed synonym identification and relation extraction method achieves the best performance compared with state-of-the-art methods and also builds a more complete ontology compared with existing gastroenterology disease ontologies. Our results are reproducible, and we will release the source code and ontology of this work after publication: https://github.com/shenyingpku/gastrointestinal_owl.