The Automatic Extension of a Knowledge Base Through Natural Language Expressions




Since both natural language processing systems and expert systems work with knowledge in some form, they have similar knowledge representation issues. For example, representation problems such as dealing with uncertainty, redundancy, and prototypical knowledge must be addressed in both natural language processing systems and expert systems.

Techniques that have been applied in these two major areas of research may be combined to develop expert systems that are capable of extending their own knowledge bases by making use of machine-readable natural language sources such as dictionaries, encyclopedias, and technical manuals. Such a self-extending expert system would be taking advantage of "real world" sources of common knowledge typically used by experts in the domain for which the system is defined. Thus such a system would be better able to model the behavior of a human expert than currently existing systems.

This paper describes current research being done in the development of knowledge bases capable of extending their knowledge through machine-readable natural language sources. The emphasis is on the development of a flexible knowledge representation medium and sufficient natural language processing tools to provide this self-extending capability via large-scale sources such as dictionaries, encyclopedias, thesauri, and technical references.