Knowledge Representation and Reasoning with Definitional Taxonomies

Weida, Robert Anthony

We provide a detailed overview of knowledge representation issues in general and terminological knowledge representation in particular. Terminological knowledge representation, which originated with KL-ONE, is an object-centered approach in the tradition of semantic networks and frames. Terminological systems share three distinguishing characteristics: (1) They are intended to support the definition of conceptual terms comprising a "terminology" and to facilitate reasoning about such terms. As such, they are explicitly distinguished from assertional systems which make statements of fact based on some terminology. (2) Their concepts are arranged in a taxonomy so that the attributes of a concept apply to its descendants without exception. Thus, the proper location of any concept within the taxonomy can be uniquely determined from the concept‘s definition by an automatic process known as classification. (3) They restrict the expressiveness of their language to achieve relatively efficient performance. We first survey important general issues in the field of knowledge representation, consider the semantics of concepts and their interrelationship, and examine the intertwined notions of taxonomy and inheritance. After discussing classification, we present a number of implemented terminological systems in detail, along with several hybrid systems which couple terminological and assertional reasoning components. We conclude by assessing the current state of the art in terminological knowledge representation.



More About This Work

Academic Units
Computer Science
Department of Computer Science, Columbia University
Columbia University Computer Science Technical Reports, CUCS-047-91
Published Here
March 17, 2012