Proceedings of the Third Conference on Artificial Intelligence Applications
Breaking the Primitive Concept Barrier
February 23, 1987
This paper addresses a fundamental trade-off that exists in knowledge representation languages between maintaining the integrity of knowledge bases, and ease of knowledge acquisition. Systems that use a classifier (such as KL-ONE) do a good job of maintaining knowledge base integrity but make it difficult to add new knowledge. An interactive classifier is a means of easing the knowledge acquisition process while still maintaining the integrity of the knowledge base. However, the presence of primitive concepts in the knowledge base makes automatic classification error-prone and interactive classification tedious. In this paper, we discuss adding a definitional component to a KL-ONE-like knowledge base, which greatly re reduces the number of primitive concepts in the knowledge base and significantly enhances interactive classification.
Downloads: 18 downloads