Abstract:
Knowledge-base design requires a thorough analysis of the concepts to be represented.
If the analysis is incomplete or inaccurate, the resulting knowledge base may contain
arbitrary restrictions, inconsistent data, or limitations that make future extensions
impossible. Conceptual analysis is even more critical for sharing knowledge bases
between heterogeneous systems. Analyses that are adequate for independent systems
may cause incompatibilities when the systems share data. This paper discusses the
kinds of problems that require a careful conceptual analysis and the difficulties
caused by an incorrect or incomplete analysis. It also shows how independently developed
knowledge bases can be related to one another by reanalyzing and redefining the basic
concepts and relations. Such analysis is essential for sharing knowledge between systems
developed for different purposes, such as relational and object-oriented databases
or expert systems and natural language processors.
Keywords:
Conceptual Analysis - Conceptual Graphs - Natural Language Processing - Knowledge
Base Design
