Database and Expert System Technology - Information Systems Reengineering, Integration and Normalization - page 71

be inferred from the nodes to which they are directly linked, without the need for a
search through a large database.
Human-defined links are used to capture heuristic knowledge such as “Learning
Skill GENERATEs Knowledge” (see Fig. 2.13 ). Relationships like these enrich the
network by providing additional paths.
Flexibility is a major advantage of this representational scheme. New nodes and
links can be defined as needed. The lack of any formal semantics and difficulties
handling exceptions are the major disadvantages. A system that was built using
semantic networks cannot generally distinguish between instances and classes. For
example, “Jack is a human” represents an instance, while “Human is a Mammal”
represents a class. This disadvantage has meant that semantic networks have limited
success for large knowledge representation systems.

2. Production Rule Systems
Production rules were previously used in automata theory, formal grammars, and
the design of programming languages, before being used in psychological model-
ing and expert systems. In the expert system literature, they are sometimes called
“condition-action rules,” “situation-action rules,” “premise-conclusion rules,” or
“if-then rules.” The syntax of production rules include two parts: the IF-part and the
THEN-part. For example:

condition
situation
premise
if

If C1 and
..... and
Cn

Then A1 and
..... and
Cn

actions
conclusion
then

When the IF-part is true (i.e., conditions C1 and… and Cn are true), the THEN-part
(i.e., perform actions A1 and…. and An) is executed.
A production rule system is a system to effectively manage production rules.
Roughly speaking, a production rule system consists of:

• Asetofrulescalledproductionrules
• Aworkingmemorythatcanholddata,goals,orintermediateresults
• Aruleinterpreterthat decides how and when to apply the rules, and which rules
to apply

The working memory holds a number of facts relevant to the particular problem
to which the production system is being applied. These facts are used by the in-
terpreter to drive the production rules, in the sense that the presence or absence of
data elements in the working memory will “trigger” some rules, by satisfying their
activation patterns.