Knowledge Representation and Automated Reasoning

Abstract

In this paper we present some leading activities in Artificial Intelligence at the IBM Paris Scientific Center.

DAEDALUS, an expert system for trouble shooting in analog electronics, is a prototype of 2nd generation expert systems. It uses a qualitative reasoning system, in the sense that the mathematical modelization of systems in chemistry, economy, etc, generally does not allow a qualitative explanation of their behaviour as an expert would do in term of variation, sign and order of magnitude of numerical parameters. The study in qualitative reasoning currently conducted at the Center leads to the axiomatization of this concept, to the definition of the semantics of a formal language about the qualitative representation of the parameters of the mathematical model describing the system, and to the implementation of the associated theorem prover.

The underlying expert system shell of DAEDALUS uses an object oriented representation system and the qualitative reasoning module: it was necessary to use this technique to access parameters not really accessible in real situations and more generally, to emulate the qualitative reasoning approach of human experts when dealing with partially quantitative domains.

But one of the major problems is the interface of expert system. By implementing most of the basic conceptual graph manipulation algorithms and an inference process on conceptual graph, this package for natural language semantics can be used as part of an intelligent data base query system or as an intelligent interface to expert systems.