Information Technology Reference
In-Depth Information
Finally, we summarize the advantages and disadvantages of using production
rules through the work of Reichgelt (
1991
). The advantages are:
• Naturalnessofexpression:Productionruleshaveprovedparticularlysuccessful
in building expert systems. One of the main reasons for this has been the natural-
ness with which expert knowledge can be expressed in the terms of production
rules.
• Modularity:Thearchitectureofaproductionsystemsupportsaverystructured
knowledge base. First, “permanent” knowledge is separated from “temporary”
knowledge. Production rule systems contain both a rule base, in which the more
permanent knowledge resides, and a working memory, which contains the tem-
porary knowledge describing the problem the system is currently working on.
Second, the different rules are structurally independent. Third, the interpreter is
independent from the knowledge that is encoded in the rule base and working
memory. The advantages gained from this modularity are that it is easy to con-
struct, maintain, and debug the knowledge base.
• Restrictedsyntax:Productionruleshaveaveryrestrictedsyntax.The main ad-
vantage is that it becomes feasible to write a program that can read and/or modify
a set of production rules. It is also useful in generating natural language explana-
tions.
• Theproblem-solvingprocess:Productionrulesdeterminewhattodonextby
examining the representation of the present state of the problem solving process
in working memory. This particular feature gives important advantages for the
overall problem-solving process. The system can quickly focus on a hypothesis
that looks particularly promising without being forced to do so at a premature
stage.
• Explanation:Productionruleshavebeenclaimedtofacilitatetheconstructionof
programs that can explain their reasoning.
The disadvantages of the production system are:
• Inefficiencyinthecaseoflargerulebases:Therearetwopossiblesourcesof
inefficiency for large rule bases. First, determining the conflict set for a large rule
base might become a very time-consuming process. Second, once the conflict
set is determined, and turns out to contain a lot of rules, conflict resolution can
require a lot of computational power. Some work has been done in this area, such
as the RETE matching algorithm (Forgy
1982
), and the use of meta-rules (Davis
1980
).
• Limitedexpressability:Theexpressionofnegativeanddisjunctiveknowledgeis
difficult in the THEN-part of rules.
• Lackofformality:Thereisalackofformalityinthedescriptionsofproduc-
tion rules and of the reasoning processes that they use. It is not, therefore, clear
whether one can sustain the claim that rule bases can be constructed incremen-
tally. Without this capability, a lot of the attractive features of production rules
would disappear.
Search WWH ::
Custom Search