Using Non-Primitive Concept Definitions for Improving DL-based Knowledge Bases

Medical Terminological Knowledge Bases contain a large number of primitive concept definitions. This is due to the large number of natural kinds that are represented, and due to the limits of expressiveness of the Description Logic used. The utility of classification is reduced by these primitive definitions, hindering the knowledge modeling process. To better exploit the classification utility, we devise a method in which definitions are assumed to be non-primitive in the modeling process. This method aims at the detection of: duplicate concept definitions, underspecification, and actual limits of a DL-based representation. This provides the following advantages: duplicate definitions can be found, the limits of expressiveness of the logic can be made more clearly, and tacit knowledge is identified which can be expressed by defining additional concept properties. Two case studies demonstrate the feasibility of this approach.

• Type: Conference
• Year: 2004
• Where: DLOG
• Authors: Ronald Cornet, Ameen Abu-Hanna
• Keywords: Automated Reasoning, Concept Definitions, DLOG 2004, Large Number, Primitive Concept Definitions