P1: IML/FFX P2: IML/FFX QC: IML/FFX T1: IML
WL040-Schalkoff WL040/Bidgolio-Vol I WL040-Sample.cls July 16, 2003 15:52 Char Count= 0
Rule-Based and Expert SystemsRule-Based and Expert Systems
Robert J. Schalkoff,Clemson UniversityIntroduction 237
The Production System Paradigm and
Rule-Based Systems (RBS) 237
Overview 237
Features of Rule-Based Production Systems 238
Theoretical and Computational Aspects of
Rule-Based Systems 238
The Logical Basis of Rule-Based Inference 238
The Concept of Chaining and Inference
Directions 239
Potential Complexities in Chaining 239
The Inference Engine (IE) 239
Rule-Based Expert Systems Development 240
The Appeal of Expert Systems 240
Expert System Examples 240
Expert System Challenges and Limitations 240Expert System Development 240
Reasoning With Uncertainty 241
Rule-Based Expert Systems and Intelligent
Agents 241
Selected Internet Applications 242
Web-Based Technical Support 242
Electronic Commerce: Recommender Systems 243
Online Portfolio Selection 243
Network Monitoring 243
OSHA Compliance Monitoring and Advising 243
CommonRules 244
Conclusion 244
Glossary 244
Cross References 244
References 244
Further Reading 245INTRODUCTION
One of the most widely used models of knowledge rep-
resentation and manipulation is the production system.
Rule-based systems may be thought of as a subclass of
production systems. Production systems are conceptu-
ally simple and, when implemented as shells in “canned”
form, may be developed with a minimum of specialized
programming. Examples of such system shells are OPS5,
CLIPS, and Corvid (see “OSHA Compliance Monitoring
and Advising” later in the chapter).
The term “expert system,” although seemingly a
catchall in current jargon, is used to indicate a subset of
production systems that are restricted to specific task do-
mains. Many expert systems (ESs) are implemented using
the rule-based paradigm, in which knowledge is encoded
in “if–then” form.
The explosive growth of the Internet has created a large
application environment for rule-based expert systems.
Many of these systems are packaged inside so-called in-
telligent agents. Applications include recommender sys-
tems for e-commerce, online portfolio selection, network
monitoring, and expert advisors for OSHA (U.S. Occu-
pational Safety and Health Administration) compliance
monitoring. In addition, there are efforts to standardize
the communication of business-related rules across the
Web. These are described more fully in Selected Internet
Applications at the end of the chapter.THE PRODUCTION SYSTEM PARADIGM
AND RULE-BASED SYSTEMS (RBS)
Overview
Figure 1 shows a somewhat simplistic and generic
rule-based production system (Schalkoff, 1990) consis-
ting of the following:- A database of information or knowledge (e.g., facts).
- A set of productions (e.g., rules) that modify the existing
database and whose applicability is conditioned on the
current database. - A control mechanism or rule interpreter that deter-
mines the applicability of the rules in the context of
the current database and selects the most appropri-
ate rule(s) through a process known as conflict resolu-
tion.
Productions in production systems are specified by a
set of condition–action pairs. Specification of conditions
in the form of “if” statements and actions via “then” yields
the familiar rule-based system representation. Production
systems may also be thought of as a subset of pattern
directed systems, systems in which production applica-
tions are driven by input (or initial) data patterns. Produc-
tion system operation is based on a “production cycle” of
searching for applicable rules, selecting a particular appli-
cable rule, using (firing) the rule, and repeating the cycle.
Firing of a rule usually results in some modification (e.g.,
addition or deletion of facts) of the current database.
Rule-based systems provide a natural means to express
situation–action heuristics in problem solving. They are
also a natural means to express observed human behav-
ior. Thus, a rule-based paradigm is a natural choice for
expert system implementation. Rule-based expert systems
are typically based on if–then (implication)–based repre-
sentations of knowledge obtained from expert query and
applied in narrowly defined problem domains.
Following is an example of a rule from the DENDRAL
(Lindsay, Buchanan, Feigenbaum, & Lederberg, 1980) ex-
pert system. DENDRAL is one of the oldest ES, having
been developed in 1964 (in the programming language
LISP). A typical production in DENDRAL (Harmon/King,
1985) is the following:237