Knowledge-base Consistency Maintenance in an Evolving Intelligent

Advisory System

Priyamvadha Thambu?

Department of Computer Science

Iowa State University

Ames, IA 50011

Vasant Honavar

Dept. of Computer Science

Iowa State University

Ames, IA 50011

Thomas Bartay

Dept. of Industrial & Manufacturing Systems Engg.

Iowa State University

Ames, IA 50011

April 16, 1993

Abstract

1 In most real-world applications, knowledge bases grow and change over time. This paper examines two solutions to the problem of knowledge base consistency maintenance as new knowledge is acquired and assimilated in an evolving knowledge base. One of the proposed solutions makes use of symbolic inference methods to identify potential inconsistencies that might result from the addition of new knowledge in the form of rules; the other uses a hybrid symbolic-connectionist system that can be re-trained using a database of examples whenever new knowledge needs to be assimilated. The proposed solutions are discussed in the context of an intelligent advisory system designed to help identify probable cases of discrimination in public housing in Iowa and several other states in the midwestern United States.

Keywords: consistency checking, expert systems, machine learning, knowledge acquisition, neural networks.

1 Introduction

In the context of this paper, we use the term knowledgebased systems to mean those systems that use knowledge encoded in some form (e.g., rule-based systems, decision trees and artificial neural networks). Traditionally, the construction of a knowledge base (often

?Current address: Inference Corporation, Stamford, CT 06905.
ySupported by Iowa Civil Rights Commission
1This paper was published in Proceedings of FLAIRS, Fort Lauderdale, Florida (1993).

refered to as knowledge engineering) has been carried out by interviewing experts in the particular domain and painstakingly translating their problem-solving expertise into an appropriately structured set of rules or decision trees. Knowledge bases constructed in this manner are typically used with inference engines that apply the general rules in the knowledge base to a problem-specific set of facts presented to the system to derive useful conclusions (e.g., in fault diagnosis, legal decision-making).

A knowledge-base is inconsistent if a contradiction can be deduced from consistent set of input data. In most real-world applications, the domain of the knowledge-based systems evolve over time. It is therefore important for knowledge-based systems to be able to acquire new knowledge and assimilate it with the old, effectively resolving any inconsistencies that might arise in the process. For example, a system designed to provide medical advice has to stay current as new knowledge becomes available in the field of medicine. Such a system cannot simply add new rules to its existing knowledge base because the new knowledge can potentially yield conclusions that might conflict with those that could be derived using the old rules (in other words, domain knowledge may be non-monotonic). An ideal system must be able to acquire and update its rules automatically while learning to solve new problems.

Given the current state of knowledge based systems technology, this is a very difficult task. In practice, the task of keeping the knowledge base current is left to knowledge engineers, who must, using the advice provided by domain experts determine the changes nec-