Finding Conceptual Models to Assist Validation

Debbie Richards and Paul Compton

Many approaches to knowledge based systems (KBS) development attempt to build complete systems that are mostly considered final before they are put into production. Verification and Validation is seen an activity performed before the system is installed and little consideration is given to incremental KA, maintenance or validation of the system after installation. We take a situated view of knowledge as something that is created to suit the situation and that knowledge evolves as contexts change. Therefore we do not regard KA and maintenance as two distinct phases. For actual KA and maintenance we use a technique, known as ripple down rules (RDR) that is simple, yet reliable. RDR provides online validation of rules by ensuring that no new rule results in a previous rule giving the wrong conclusion. We seek to further improve the validation capabilities of RDR by retrospectively finding the models inherent in our knowledge base. We have added tools based on Formal Concept Analysis to RDR to assist the user with KA by showing the user whether the new rule and/or conclusion are consistent with existing concepts in the knowledge base and where they fit into the subsumption lattice of concepts.

This page is copyrighted by AAAI. All rights reserved. Your use of this site constitutes acceptance of all of AAAI's terms and conditions and privacy policy.