Robotics Reference
In-Depth Information
then converted into rules, of the form
If
this is true
then
that is likely to be the case.
The whole collection of rules form the
knowledge-base
or
rule-base
.The
knowledge-base in an expert system encapsulates many heuristics—little
rules of good judgment, of plausible reasoning and of good guesswork—
that characterize expert-level decision making. An entire rule-base can
be thought of as a model of the expertise of the best practitioners in the
field.
Knowledge engineering is the process of extracting relevant knowl-
edge from human experts and reducing the resultant body of knowledge
to a precise set of facts, usually in the form of these heuristic rules. The
knowledge engineer gets at and transforms appropriate information from
the minds of experts, or from topics, journals or any other source of ex-
pertise, into some manageable form. The skill of the knowledge engineer
lies in his ability to act as a translator between, on the one hand, the hu-
man experts who are being debriefed in order to extract their expertise,
and, on the other hand, the knowledge-base of the expert system itself.
The knowledge engineer elicits knowledge from the expert, encodes it
in rules in the format required by the knowledge-base, and refines it in
collaboration with the experts until an acceptable level of performance
has been achieved by the system.
Many statements that may not at first look like rules can easily be
formulated as rules, for example,
the statement
“All mammals have fur”
converts to
“
If
something is a mammal
then
it has fur”
while the statements
“Lockjaw is fatal”
and
“A patient with lockjaw will die”
both convert to
“
If
the disease is lockjaw
then
the patient
will die”.
Part of the knowledge engineer's job depends on being able to recognize
how such statements can be translated into the syntax of the rule-base.
Clearly it is important that the knowledge engineer has good com-
munication skills and that the expert is able to express his knowledge
clearly to the engineer. A typical problem faced by knowledge engineers
is that experts are often unable to volunteer useful information about
