Skip to Main content Skip to Navigation
Theses

Lézard : acquisition des connaissances et gestion de l'incertitude dans un système expert de seconde génération

Abstract : Most medical expert systems use advanced knowledge representation modes, allowing the description of complex pathophysiological models. The knowledge thus declared is called "deep knowledge". Yet, no matter how complex, such models do not always provide the knowledge needed for a diagnostic task. Empirical or surface knowledge (``shallow knowledge'') must necessarily be integrated into the system. We propose here an approach that allows the acquisition of surface knowledge from deep knowledge: the latter has a complex semantic content that allows several valid logical interpretations. The translation of this deep knowledge into surface knowledge forces to retain only the semantically valid interpretations. This decoupling then allows to have a simplified interpretation procedure insofar as it only depends on the logical framework of the surface knowledge. In our system, the expert is questioned during the process of acquiring surface knowledge from deep knowledge. Heuristic meta-knowledge is used to guide his interrogation. The information provided by the deep knowledge is thus completed in a very precise way at the level of the very granular surface knowledge; this completion makes it possible to solve in a satisfactory way the problems of uncertainty, contributing to the construction of an operational knowledge base. This base is then used by an inference engine of the "forward chaining" type operating within the framework of a non monotonic logic. This engine relies on a coherence maintenance system providing the necessary information to discriminate possible reasonings. Heuristic strategies are then used to choose the right reasoning. The essential Lizard system, in which these mechanisms are implemented, is applied to the domain of acute abdominal pain.
Complete list of metadata

https://tel.archives-ouvertes.fr/tel-03560682
Contributor : Jean Charlet Connect in order to contact the contributor
Submitted on : Friday, February 25, 2022 - 10:46:52 PM
Last modification on : Thursday, August 4, 2022 - 5:21:50 PM
Long-term archiving on: : Thursday, May 26, 2022 - 6:08:35 PM

File

CharletTHESE1989.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : tel-03560682, version 1

Citation

Jean Charlet. Lézard : acquisition des connaissances et gestion de l'incertitude dans un système expert de seconde génération. Intelligence artificielle [cs.AI]. Université Pierre & Marie Curie - Paris 6, 1989. Français. ⟨tel-03560682⟩

Share

Metrics

Record views

34

Files downloads

7