mod-ki message 0788
************************************************************************
Prolog Forum UNI/ETH Zurich WS 1991/92
Wednesday, Nov. 6, 1991, 12.30 at room 21-G-74 (below the mensa)
University of Zuerich-Irchel, Winterthurerstr. 190
Lawrence Chachere
CUI, Universite de Geneve
Advantages of a Logic-Based Representation for
Knowledge Acquisition in Expert Systems
Knowledge acquisition is the bottleneck of expert system development.
One major component of this bottleneck is the difficulty associated with
refining information within a partially constructed knowledge base.
Thus, many recent research efforts have focused on the automation of
knowledgebase refinement techniques. The development of MINERVA by the
Knowledge Base Systems group at the University of Illinois was one such
research effort.
MINERVA is a Prolog reimplementation and enhancement of HERACLES,
the expert system shell of NEOMYCIN. The reimplementation was guided
by experience gained from the ODYSSEUS apprenticeship learning program
for HERACLES. MINERVA has an advantage over the older shell because it
accommodates inference engines for learning, in addition to the general
inference engine for problem solving.
It is emphasized that the improved shell is the result of the
restructuring of knowledge. No new knowledge was added to the system. In
this talk, I will describe how the Prolog representation of the shell
facilitates two machine learning techniques, both of which have already
produced significant results for acquiring medical domain knowledge. I
will also discuss how meta-interpreting, a special feature of Prolog,
facilitates the development of inference engines.
This work was done at the University of Illinois in Urbana.
For more information:
Markus Fromherz Tel.: +41-1-257 4313 / 4311
Institut fuer Informatik E-Mail: from...@ifi.unizh.ch
Universitaet Zuerich-Irchel
Winterthurerstrasse 190
CH-8057 Zuerich
************************************************************************