[comp.ai.shells] testing a rule-based system

sweeney@cs.washington.edu (Christine Sweeney) (02/08/91)

I have the task of testing a medical ai system which is
rule-based.  We have written our own inference engine
which does forward and backward chaining on a set of
rules.  The whole thing is in lisp.  

Does anyone know of ways to test such a system thoroughly
other than writing a bunch of tests for *many* individual
hypothetical cases?  

Can anyone recommend any books or papers on the subject of
testing and verifying ai-based systems?

Thank you.  If I recieve enough interesting responses I
will post a collection of them,

Christine Sweeney

sweeney@cs.washington.edu

ntm1169%dsacg3.dsac.dla.mil@dsac.dla.mil (Mott Given) (02/12/91)

From article <7516@uklirb.informatik.uni-kl.de>, by sweeney@cs.washington.edu (Christine Sweeney):
> Can anyone recommend any books or papers on the subject of
> testing and verifying ai-based systems?

You might try looking at some of these references:

AI EXPERT, Feb. 1990,  "A tool for detecting expert system errors"

IEEE Expert, Winter 1987, "Validating expert system performance"

IEEE Expert, Spring 1988, "A software engineering tool for expert system design"

IEEE Expert, June 1990, "Validating expert systems"

A paper by the authors Todd Stock and others (whose title I do not have)
  presented at the Fourth Conference on Artificial Intelligence for Space 
  Applications, Huntsville, AL, Nov 15-16, 1988 describes an Expert System
  Validation Assistant.

Hall, David L.; Heinze, Daniel T.
 "The use of simulation techniques for expert system test and evaluation"  
 ISA Transactions v28 no1  p19-22 '89
  
Culbert, Chris; Riley, Gary; Savely, Robert T.
 "An expert system development methodology that supports verification and 
 validation"
 ISA Transactions v28 no1  p15-18 '89
 
"Technique for knowledge base verification" by Zhang, Du and Nguyen, Doan
  from the conference  IEEE International Workshop on Tools for Artificial 
  Intelligence: Architectures, Languages and Algorithms;
  in  Fairfax, VA, on Oct 23-25, 1989

-- 
Mott Given @ Defense Logistics Agency ,DSAC-TMP, P.O. Box 1605,
Bldg. 27  Section 1, Systems Automation Center, Columbus, OH 43216-5002
INTERNET: mgiven@dsac.dla.mil         UUCP:  ...osu-cis!dsac!mgiven
Phone:  614-238-9431    AUTOVON: 850-9431     FAX: 614-238-3214

timm@runx.oz.au (Tim Menzies) (02/14/91)

In article <7516@uklirb.informatik.uni-kl.de> sweeney@cs.washington.edu (Christine Sweeney) writes:
>I have the task of testing a medical ai system which is
>rule-based.  We have written our own inference engine
>which does forward and backward chaining on a set of
B
>rules.  The whole thing is in lisp.  
>
>Does anyone know of ways to test such a system thoroughly
>other than writing a bunch of tests for *many* individual
>hypothetical cases?  
>

An approach to verifying  completeness and consistentcy  in a rule-based
expert system

Suwa. M, Carlisle Scott A, Shortliffe E.H. , Technical report,  Dept of
Computer Science, Stanford University,  August 1982.

bernus@batserver.cs.uq.oz.au (Peter Bernus) (02/20/91)

In article <7516@uklirb.informatik.uni-kl.de> sweeney@cs.washington.edu
(Christine Sweeney) writes:


>Does anyone know of ways to test such a system thoroughly
>other than writing a bunch of tests for *many* individual
>hypothetical cases?


see
Mark Bezem: Consistency of Rule Based Systems
CWI report CS-R8736

(I think he is bezem@cwi.nl but the postal address is
definitely 

%I Centrum voor Wiskunde en Informatica
%A Kruislaan 413
%A 1098 SJ Amsterdam
%C the Netherlands

fax: +31-20-592.4199
)

Peter Bernus

jjacobs@uunet.uu.net (Jeffrey Jacobs) (02/26/91)

Expert System Testing, J.M. Jacobs & C. Chee, bd Systems, 357 Van Ness
Way, SUite 110, Torrance, CA 90501, (213)618-8798.  Ask for Christine
Chee.

Jeffrey M. Jacobs
ConsArt Systems Inc, Technology & Management Consulting
P.O. Box 3016, Manhattan Beach, CA 90266
voice: (213)376-3802, E-Mail: 76702.456@COMPUSERVE.COM