bdn@lln-cs.UUCP (Denis Ballant) (03/22/90)
What is an expert system (ES) ?
To arouse some discussion about this topic, I submit you
several common definitions (found in the literature).
I hope that this discussion will be valuable to have a better
comprehension of this often misunderstood term.
Denis Ballant.
bdn@info.ucl.ac.be
\ / /
/\ ~ o Y /
/ | |/ /^/ / /\ |/
/_/ _/\/ /_/_/ c' /
|
-=-=-=-=-=-=-=-=-=-/
1. 1st Definition [Clan 89]
Expert systems handle heuristics and uncertain knowledge;
their reasoning is qualitative and imprecise. In contrast to
quantitative models (involving mathematical laws as found in
physics), expert systems contain qualitative world models.
2. 2nd Definition [Mich 86]
An expert system is a computer system that achieves
performance comparable to a human expert at solving problems
in some task domain by utilizing a large amount of
domain-specific knowledge. Because of the substantial
amounts of knowledge required, the knowledge acquisition task
assumes major proportions.
3. 3rd Definition [Ole 86]
Les syst`emes experts se comportent comme des experts
humains. Ceux-ci ont r'euni, dans le domaine pour lequel ils
sont experts, des connaissances pr'ecises et s^ures. Gr^ace
`a celles-ci ils peuvent porter des diagnostics et r'epondre
aux questions qui portent sur ce domaine. Les syst`emes
experts fonctionnent de cette mani`ere. Ils comportent deux
'el'ements principaux: une base de donn'ees et un moteur
d'inf'erence. Ils correspondent respectivement aux
connaissances acquises, stock'ees dans la m'emoire de
l'ordinateur, et au raisonnement qui permet, `a partir des
informations communiqu'ees, de porter le diagnostic. Les
syst`emes experts assurent une gestion des connaissances. La
constitution de la base de donn'ees se r'ealise par la
consultation, l'interview d'experts humains qui communiquent
le savoir que l'ordinateur va enregistrer et ensuite
exploiter.
4. Misc: [Sowa 84]
Expert systems can be grouped in three major categories
according to the kinds of problem they address:
a. Classification systems. Ex: MYCIN
b. Design systems. Ex: R1
c. Decision support systems (DSS).
5. Misc: [Greg 88]
Three Expert System problem-solving philosophies:
a. shallow knowledge approach
b. model-based reasoning approach (Steels 1985)
Constraint: representation of behavioral models
of complex entities must be allowed by the
knowledge representation formalism.
c. structured reasoning approach
Make the structure of the reasoning process
explicit, and organize the reasoning process by
means of a comparison of elementary tasks
operating on specific blocks of knowledge.
Constraint: The partitioning of the KB into
role-specific components with selective access
must be allowed.
6. References:
--------------
[Sowa 84]
Sowa John F.,
Conceptual Structures -- Information Processing in Mind and Machine,
Addison-Wesley, 1984.
[Greg 88]
Gr'egoire Eric,
Evaluation of the expert system tools KEE and ART:
a case study,
Applied Artificial Intelligence, Vol. 2, Num 1,
R.Trappl (ed.), Hemisphere, New-York, 1988, pp.1--23.
[Clan 89]
Clancey William J.,
Viewing knowledge bases as qualitative models,
IEEE Expert, Summer 1989.
[Mich 86]
Michalski Ryszard S., Carbonell Jaime G., Mitchell Tom M.,
Machine Learning: An Artificial Intelligence Approach,
Vol.2, Morgan Kaufmann Publishers, Palo Alto, CA, 1986.
[Ole 86]
Ol'eron Pierre,
L'intelligence, Presses Universitaires de France,
Collection ``Que sais-je ?'', Num 210, 4th edition, 1986.