Building a Large Knowledge Base Semi-Automatically

Wedescribe a knowledgeengineeringapproachby which conceptualknowledgeis extractedfroman informal,semanticallyweakmedicalthesaurus(UMLS)and automatically convertedinto a formallysounddescriptionlogics system. Ourapproachconsists of four steps: concept definitions are automaticallygeneratedfromthe UMLS source,integrity checkingof taxonomicandpartonomic hierarchiesis performedbythe terminologicalclassifier, cyclesandinconsistenciesare eliminated,andincremental refinementof the evolvingknowledgebase is performed bya domainexpert. Wereport onexperimentswitha terminologicalknowledgebase composedof 164,000conceptsand76,000relations.

• Type: Conference
• Year: 2001
• Where: FLAIRS
• Authors: Udo Hahn, Stefan Schulz
• Keywords: Artificial Intelligence, FLAIRS 2001, Integrity Checkingof, Taxonomicandpartonomic Hierarchiesis Performedbythe, UMLS Source