Antecedent. See premise. [Top]

Attribute. A variable that takes on values that might be numeric, text, or logical (true/false). Attributes store the factual knowledge in a knowledge base. [Top]

Backward chaining. The process of determining the value of a goal by looking for rules that can conclude the goal. Attributes in the premise of such rules may be made subgoals for further search if necessary. [Top]

Breadth first search. A search strategy that examines all rules that could determine the value of the current goal or subgoal before backtracking through other rules to determine the value of an unknown attribute in the current rule. [Top]

Certainty processing. Allowing confidence levels obtained from user input and rule conclusions to be combined to increase the overall confidence in the value assigned to an attribute. [Top]

Certainty factor. A measure of the confidence assigned to the value of an attribute. Often expressed as a percentage (0 to 100%) or probability (0 to 1.0). 100% or 1.0 implies that the attribute's value is known with certainty. [Top]

Clause. One expression in the If (premise) or Then (consequent) part of a rule. Often consists of an attribute name followed by a relational operator and an attribute value. [Top]

Conclusion. See consequent. [Top]

Confidence factor. See certainty factor. [Top]

Consequent. The Then part of a rule, or one clause or expression in this part of the rule. [Top]

Control information. Elements of a knowledge base other than the attributes and rules that control the user interface, operation of the inference engine and general strategies employed in implementing a consultation with an expert system. [Top]

Depth first search. A search strategy that backtracks through all of the rules in a knowledge base that could lead to determining the value of the attribute that is the current goal or subgoal. [Top]

Domain. A specific problem environment for which knowledge is captured in a knowledge base. [Top]

Expert system. A domain specific knowledge base combined with an inference engine that processes knowledge encoded in the knowledge base to respond to a user's request for advice. [Top]

Expert system building tool. See expert system shell. [Top]

Expert system shell. A suite of software that allows construction of a knowledge base and interaction with this knowledge base through use of an inference engine. [Top]

Expertise. Specialized domain knowledge, skills, tricks, shortcuts and rules-of-thumb that provide an ability to rapidly and effectively solve problems in the problem domain. [Top]

Firing a rule. A rule fires when the if part (premise) is proven to be true. If the rule incorporates an else component, the rule also fires when the if part is proven to be false. [Top]

Forward chaining. Applying a set of previously determined facts to the rules in a knowledge base to see if any of them will fire. [Top]

Goal. A designated attribute: determining the values of one or more goal attributes is the objective of interacting with a rule based expert system. [Top]

Inference. New knowledge inferred from existing facts. [Top]

Inference engine. Software that provides the reasoning mechanism in an expert system. In a rule based expert system, typically implements forward chaining and backward chaining. strategies. [Top]

Knowledge base. The encoded knowledge for an expert system. In a rule-based expert system, a knowledge base typically incorporates definitions of attributes and rules along with control information. Knowledge base format is specific to the implementing expert system shell or other software. [Top]

Knowledge engineering. The process of codifying an expert's knowledge in a form that can be accessed through an expert system. [Top]

Premise. The if part of a rule: represents an hypothesis. [Top]

Primary goal. A goal for which a value is sought during an expert system consultation. Primary goals are terminal objectives for the consultation: once their value are found, or a determination is made that they cannot be found, the consultation ends. [Top]

Production rule. See rule. Rules are called production rules because new information is produced when the rule fires. [Top]

Relational operator. Conditions such as is equal to or is less than that link an attribute name with an attribute value in a rule's premise to form logical expressions that can be evaluated as true or false. In the example logical expression:

Rule. A statement of the form: if <x> then <y> else <z>. The if part is the rule premise, and the then part is the consequent. The else component of the consequent is optional. The rule fires when the if part is determined to be true or false. Here is an example rule:
If the credit rating is good and the amount of the sale is less than 10000
Then the decision is accept the sale
Else the decision is reject the sale
[Top]

Shell. See expert system shell. [Top]

Subgoal. An attribute the becomes a temporary intermediate goal for the inference engine. Subgoal values need to be determined because they are used in the premise of rules that can determine primary goals. [Top]