The Internet Encyclopedia (Volume 3)

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

238 RULE-BASED ANDEXPERTSYSTEMS

Figure 1: Basic production system architecture.

If

the molecule spectrum has two peaks at masses

x1 andx2 such that

x 1 +x 2 =M+ 28 AND

x 1 −28 is a “high” peak AND

x 2 −28 is a “high” peak AND

at least one of thex1orx2 is high,

Then

The molecule contains a ketone group.

Another sample rule, taken from the online technical sup-

port example in the “Web-Based Technical Support” sec-

tion is this:

if “symptom” = “Label Designer hangs on converting

databases dialog”

and (“BackTrack Version”<=“4.10”)

and “second edition”=“yes”

and (operatingsystem=“Windows 98 Second Edition”

or operatingsystem=“Windows 98”)

then problem=“Win 98se”

Figure 2 shows a more realistic rule-based system ar-

chitecture, which takes into account practical concerns

such as user interfacing and structuring and partitioning

of the fact and rule databases.

Features of Rule-Based Production Systems

Advantages

Characteristic of systems employing productions in rule

form are the following generally positive features:

Figure 2: More complete rule-based production system

structure. Includes structured rule and facts databases and

user interface.

1. Expressibility (rules allow the expression of knowledge
   in a form understood by human experts and at the
   same time sufficiently quantitative for symbolic ma-
   nipulation by the machine).


2. Ease of modification of the database (facts and or rules
   may be added or removed); in addition, the develop-
   ment and implementation of systems that are highly
   modular (e.g., see Figure 2) is straightforward.


3. Ease of exploring the current knowledge base con-
   tained in the system (i.e., the encoding of information
   is in a readable form).


4. Flexibility of processing (the inference mechanism(s)
   may be chosen to suit the problem).


5. Ease in following the inference mechanism (the or-
   der in which rules were employed may be recorded
   and traced for an “explanation” of the system’s conclu-
   sions).


6. Standardization in terms of a knowledge representa-
   tion and inference approach.


7. Availability of off-the-shelf software for implementa-
   tion.

Disadvantages

Disadvantages of productions in rule form include the fol-

lowing:

1. Inability to predict system behavior for a given prob-
   lem. (This characterizes many AI computations.) In
   other words, there may be many solutions, no solution
   or a unique solution, and the only way to establish this
   is via IE search.


2. Inability to force a specific production sequence com-
   pared with imperative programming. (Note some ex-
   pert system shells allow manipulation of inference en-
   gine parameters to accomplish this.)


3. Lack of suitability for all applications.


4. Lack of ability to implement directly “deep reasoning”
   and “common sense.” Note that this is a general repre-
   sentational issue, not an inherent shortcoming of the
   rule-based production system paradigm.

THEORETICAL AND COMPUTATIONAL

ASPECTS OF RULE-BASED SYSTEMS

The Logical Basis of Rule-Based Inference

Generation of new facts or verication of a goal fact set in

a rule-based production system proceeds by linking the

“then” parts of rules to the “if” parts of other rules and

proceeding until the goal statement is proven to be true

or no new facts may be produced. This process is called

chaining and is accomplished under the control of the

inference engine (IE).

Consider a simplistic rule of the (abstract) form:

Ifpthenq,

wherepis denoted the rule antecedent andqis denoted

the consequent. More formally,pandqare statements in

logic, connected by implication (the symbol→denotes

the implication connective), and this is normally written