Date of Award

1-1-1992

Degree Type

Thesis

Degree Name

Master of Applied Science (Computer studies)

Faculty

Faculty of Science, Technology and Engineering

First Advisor

Mr. Tim Roberts

Abstract

This research studies currently available maintenance methodologies for expert system knowledge bases and taxonomically classifies them according to the functions they perform. The classification falls into two broad categories. These are: (1) Methodologies for building a more maintainable expert system knowledge base. This section covers techniques applicable to the development phases. Software engineering approaches as well as other approaches are discussed. (2) Methodologies for maintaining an existing knowledge base. This section is concerned with the continued maintenance of an existing knowledge base. It is divided into three subsections. The first subsection discusses tools and techniques which aid the understanding of a knowledge base. The second looks at tools which facilitate the actual modification of the knowledge base, while the last secttion examines tools used for the verification or validation of the knowledge base. Every main methodology or tool selected for this study is analysed according to the function it was designed to perform (or its objective); the concept or principles behind the tool or methodology: and its implementation details. This is followed by a general comment at the end of the analysis. Although expert systems as a rule contain significant amount of information related to the user interface, database interface, integration with conventional software for numerical calculations, integration with other knowledge bases through black boarding systems or network interactions, this research is confined to the maintenance of the knowledge base only and does not address the maintenance of these interfaces. Also not included in this thesis are Truth Maintenance Systems. While a Truth Maintenance System (TMS) automatically updates a knowledge base during execution time, these update operations are not considered 'maintenance' in the sense as used in this thesis. Maintenance in the context of this thesis refers to perfective, adaptive, and corrective maintenance (see introduction to chapter 4). TMS on the other hand refers to a collection of techniques for doing belief revision (Martin, 1990) . That is, a TMS maintains a set of beliefs or facts in the knowledge base to ensure that they remain consistent during execution time. From this perspective, TMS is not regarded as a knowledge base maintenance tool for the purpose of this study.

Share

 
COinS