Increased system reliability through improved validation of knowledge in expert systems with single domain expert development teams.

摘要

Current software engineering techniques provide a solid foundation for the validation and verification of computer applications. They ensure that delivered applications are of higher quality and have a reduced number of “bugs” due to developer error. These techniques do not address “bugs” that are inherent in the knowledge, requirements, or specifications as they are received from the experts. 30% of the “bugs” occurring in software development originate in the Requirements Phase (Jones, 1991). Baseline data indicates that a third of these, or 10% of the “bugs” in software development projects, are due to errors in the knowledge or specifications received from the expert. This dissertation describes a technique, Knowledge Validation, which enhances knowledge acquisition for expert system development or specification and requirements gathering in traditional software development projects.; For projects with multiple experts participating on the development team, techniques such as use of the Delphi Algorithm (Brown et al. 1969; Shafai, 1989) are available to reduce these expert induced errors. This technique focuses on projects where there is a single expert participating on the development team.