Production Knowledge Model

  “IF condition, then action” 

The production model has the disadvantage that with the accumulation of a sufficiently large number (of the order of several hundred) products, they begin to contradict each other.

In the general case, the production model can be represented as follows:

Knowledge processing systems using a production model are called “production systems” . The composition of expert systems of the production type includes a rule base (knowledge), a working memory and a rule interpreter (solver) that implements a certain inference mechanism. Any production rule contained in the knowledge base consists of two parts: the antecedent and the sequential . Antecedent is a premise of a rule (conditional part) and consists of elementary sentences connected by logical ligaments "and", "or". The consequent (conclusion) includes one or several sentences that express either a certain fact or an indication of a specific action to be executed. Production rules are usually written in the form of antecedent-sequential.

Examples of production rules:

  IF A 
  "The engine will not start" 
  and 
  "Engine starter does not work" 
  THAT 
  "Problems in the power supply system of the starter" 

Any rule consists of one or several attribute-value pairs. In the working memory of systems based on production models, attribute-value pairs are stored, the truth of which is established in the process of solving a specific problem to a certain current point in time. The contents of the working memory is changed in the process of solving the problem. This happens as the rules trigger. The rule is triggered if, when comparing the facts contained in the working memory, with the antecedent of the rule being analyzed, there is a coincidence, and the conclusion of the triggered rule is entered into the working memory. Therefore, in the process of logical inference, the volume of facts in working memory, as a rule, increases (it can decrease if the effect of a rule is to delete facts from working memory). In the process of logical inference, each rule from the rule base can only work once.

There are two types of production systems - with "direct" and "reverse" conclusions. Direct conclusions implement the strategy “from facts to conclusions”. In the case of inverse conclusions, hypotheses of probabilistic conclusions are put forward, which can be confirmed or disproved on the basis of the facts entering the working memory. There are also systems with bidirectional outputs.

The main advantages of systems based on production models are associated with the simplicity of knowledge representation and organization of logical inference. The disadvantages of such systems include the following:

When developing small systems (dozens of rules), the positive aspects of the production models of knowledge are manifested, but with an increase in the amount of knowledge, the weaknesses become more noticeable.