Effectiveness of UMLS semantic network as a seed ontology for medical domain ontology building.

Abstract

The goals of building this ontology are to determine the effectiveness of UMLS in predicting the extracted concept relationships and consequently enrich the seed ontology based on the extracted concept and semantic relationships. The UMLS semantic network is an extensive knowledge source that is useful in the medical domain. However, it was found that information extraction using UMLS concepts and semantic types were not sufficient to predict the relationships. This is due to the fact that only half of the semantic relationships extracted hold true based on UMLS. In addition, some semantic types in UMLS were found to be too general and its definitions are too broad. Therefore, subclasses are created to have a more specific categorization of two of the semantic types. In general, the effectiveness of the UMLS knowledge source could be further enhanced with the above findings. The generated ontology is hoped to be used as a domain knowledge base for medical digital library application. In addition, it would provide faster retrieval of relevant and useful information for treatment of colon cancer.