D17: Expert Systems

Systems for representing expert knowledge in computers currently provide the most useful application of Artificial Intelligence techniques. An Expert System may be defined as a system which embodies knowledge of a specialised domain and can perform a task at a level of competence comparable to that of an expert in the domain. The technology of Expert Systems has been applied in diverse areas such as engineering, medicine, finance and economics with the principal aim of making human expertise widely and cheaply available. The aim of the course is to impart an appreciation of the nature of expertise and Expert Systems technology, and awareness of the various ways of representing and reasoning with knowledge and the practicalities of eliciting it.

Term Prerequisites Core For
2 N/A None

Taught By

John Campbell (.5)
John Washbrook (.5)

Syllabus

Introduction

· The nature of Expert Systems. Types of applications of Expert Systems; relationship of Expert Systems to Artificial Intelligence and to Knowledge-Based Systems.

· The nature of expertise. Distinguishing features of Expert Systems. Benefits of using an Expert System. Choosing an application.

· Theoretical Foundations.

· What an expert system is; how it works and how it is built.

· Basic forms of inference: abduction; deduction; induction.

· The representation and manipulation of knowledge in a computer. Rule-based representations (with backward and forward reasoning); logic-based representations (with resolution refutation); taxonomies; meronomies; frames (with inheritance and exceptions); semantic and partitioned nets (query handling).

· Basic components of an expert system. Generation of explanations. Handling of uncertainties. Truth Maintenance Systems.

· Expert System Architectures. An analysis of some classic expert systems. Limitations of first generation expert systems. Deep expert systems. Co-operating expert systems and the blackboard model.

· Building Expert Systems. Methodologies for building expert systems: knowledge acquisition and elicitation; formalisation; representation and evaluation. Knowledge Engineering tools .

Assessment

Weighting No. Exam Questions No. Courseworks
100% examination 5 N/A

Examination Rubric

Answer 3 questions out of 5. Time allowed: 2 hours; 30 minutes.

Reading list

Recommended Text: P Jackson, Introduction to Expert Systems, Addison Wesley, 1990 (2nd Edition) Currently the primary text recommended for purchase. It does not cover the course completely, and conversely contains much (a great deal, in fact) that is not covered. Background Reading: Elaine Rich, Kevin Knight, Artificial Intelligence, McGraw-Hill Inc, 1991 (2nd Edition) Much improved over the first edition, which is no substitute. Contains material covered in the course which is not in Jackson. Jean-Louis Lauriere, Problem Solving and Artificial Intelligence, Prentice Hall, 1990