Information Technology Reference
In-Depth Information
thus the first successful “knowledge-intensive” AI system because it automated
the decision-making and problem-solving processes of experts in a field.
Feigenbaum looked at other domains where this approach could be
applied. With Bruce Buchanan and Edward Shortliffe, he developed the MYCIN
expert system to diagnose and treat blood infections ( Fig. 13.5 ). Using about 450
rules developed from interviews with experts, MYCIN performed better than
many doctors. The success of DENDRAL, MYCIN, and other expert systems led
to an overenthusiastic rush to produce commercial systems in the 1970s and
1980s. Although the high hopes of the pioneers were not fully realized, knowl-
edge-based expert systems are still used for applications ranging from straight-
forward help-desk and technical support to manufacturing and robotics. For
narrow, well-defined problems, expert systems can be successful. However, a
major limitation of this rule-based approach to knowledge is that these systems
do not generalize well to larger, broader problems. In addition, the develop-
ment and capture of the knowledge rules are very labor intensive and usually
very specific to the case at hand. Because almost nothing in real life is simply
true or false in the way that abstract logic requires, for any commonsense rule
about the world there must also be a large number of exceptions.
The creation of taxonomies and classifications dates back to the 300s B.C. ,
when Aristotle wrote his Organon , a collection of his works on logic. This included
a section on categories that we would now call a type of ontology , the study of
what kinds of things exist. It was the Swedish biologist Carolus Linnaeus who
invented our present system of biological classification in the 1700s ( Fig. 13.6 ).
Computer scientists have borrowed the word ontology from the philosophers to
describe a structural framework for organizing knowledge. An ontology speci-
fies a set of concepts within a domain that a computer can use to reason about
objects in the domain and about the relationships between them. AI researchers
have long believed that useful ontologies are essential for effective AI systems.
One response to this need is therefore to expand the knowledge base of the
computer by producing a comprehensive vocabulary of all of the important con-
cepts in a given domain, including the objects in the domain and the properties,
relations, and functions needed to define the objects and specify their actions.
One of the most ambitious ontology projects is the Cyc project, started
by Douglas Lenat in 1984. The name Cyc is a shortened form of encyclopedia .
The project is an attempt to build a knowledge base containing much of the
everyday, commonsense knowledge of a human being. Typical pieces of knowl-
edge represented in the database are statements such as “Every tree is a plant”
and “Plants die eventually.” After more than twenty-five years, Cyc's knowl-
edge base contains more than one million assertions, rules, or commonsense
ideas. However, its creators estimate that it will need more than one hundred
times that many entries before Cyc can begin to learn for itself from written
material.
The DBPedia ( Fig. 13.7 ) project has taken a different approach and uses a
method called crowdsourcing , soliciting content from a large group of people, to
extract structured data from Wikipedia. DBPedia's 2012 release contained an
ontology with more than two million concepts together with about one hundred
facts per concept. The researchers hope that the Cyc and DBPedia projects will
help realize Tim Berners-Lee's vision of the Semantic Web, in which machines
Fig. 13.5. The MYCIN project was an
expert system designed to diagnose and
treat blood infections. It was developed
at Stanford by Edward Shortliffe with
Bruce Buchanan and Ed Feigenbaum.
B.13.7. Ed Feigenbaum received
the Turing Award for his work in
expert systems in 1994. He is often
known as the “father of expert sys-
tems.” He also served as chief sci-
entific adviser for the U.S. Air Force
and received their Exceptional
Civilian Service Award in 1997.
 
Search WWH ::




Custom Search