JPH0358230A - Rule executing device in production system - Google Patents

Abstract

PURPOSE:To improve the execution efficiency of a rule by extracting a rule in which a conclusion part is the same and a condition part has an inclusive relation from a production rule group, determining a rule to be executed from therein, and reconstituting and executing a rule group from this rule and a rule in which a condition part is in a non-inclusive relation. CONSTITUTION:At the time of executing a production rule, plural production rules in which conclusion parts are the same and condition parts have an inclusive relation are extracted from a rule group by an inclusive rule extracting means 2, and a production rule to be executed is determined from therein by an execution rule selecting means 3. Subsequently, a rule group is reconstituted from one production rule selected from these plural production rules and a production rule in which it is decided that the condition part is in a non- inclusive relation by a rule extracting means, and each production rule for constituting this reconstituted rule group is executed by a rule group executing means 5. In such a way, the execution efficiency of the rule can be improved.

Description

[Detailed description of the invention]

[Industrial Application Field] The present invention relates to a rule execution device for a knowledge base system (hereinafter referred to as a "production system") in which knowledge is expressed as production rules. [Conventional technology] Conventionally, knowledge is divided into IF (condition part) - THEN (conclusion part)
Expressed as a production rule (hereinafter simply referred to as a "rule") in the form of a rule, when a certain event corresponds to the conditional part or conclusion part of the rule, the rule is executed by the rule execution device to obtain a specific inference result. The production system is known. In such a production system, for example, a rule that includes a certain event in its condition part is selected by a rule execution device, and the selected rule is executed to produce one inference result (conclusion). The rule execution device further selects a rule whose condition part includes this conclusion and executes it. Finally, one or more inference results are obtained. The confidence level of each conclusion is defined in the conclusion section which is the inference result of each rule, and the inference result is evaluated based on this confidence level. By the way, as shown in FIG. 6, a plurality of rules have the same conclusion part, and each condition part has a logic m (AN
When there is an inclusive relationship with each other due to D), all rules R1 to R4 having such a relationship are executed, and the confidence CF of the conclusion T is obtained by synthetically calculating the confidence CF of the rules R1 to R4. . That is, the event (A, B, C, D
) as a condition part, rule R4, rule R3, rule R2, and rule R1 are executed sequentially for each event A, B, C, and D. Then, the confidence level CF of the conclusion part T when the rules R1 to R4 hold is determined by the following composite calculation: CP-1-(1-0.9) (1-0.8) (1-0
．． 7) (1-0.8). However, even when the condition part is (A, B, C, D), such a rule execution device cannot execute rules R2 to R2 that have an inclusive relationship with the condition part of rule R1 that has this condition part.
Since it is executed up to R4, the confidence of the rule's event has to be recalculated as described above, resulting in a value that deviates from the predetermined confidence, so we do not know the exact confidence of the inference result. I can't. Further, the rule builder must set the confidence level by taking into consideration that the confidence level of each rule is determined by a synthetic calculation, and setting the confidence level is extremely difficult. [Problem to be Solved by the Invention] Therefore, in the conventional rule execution device, all rules in which the event in the conclusion part of the rule is the same and the condition part has an inclusive relationship are executed, so the efficiency of rule execution is low. In addition, the confidence level of the conclusion part will be a value different from the predetermined value, making it impossible to perform accurate evaluation, and when setting the confidence level of the conclusion part of each rule, This poses a problem in that it places a heavy burden on the rule builder, since it is necessary to take into account that the reliability of the rule is calculated synthetically. The present invention has been made in view of the above circumstances, and it is possible to improve the efficiency of rule execution, to accurately evaluate the conclusion part of each executed rule with a predetermined confidence level, and to improve the efficiency of rule execution. It is an object of the present invention to provide a rule execution device that can significantly reduce the burden on a builder. [Structure of the Invention] [Means for Solving the Problems] In order to solve the above problems, the present invention provides a method for solving the above-mentioned problems by selecting products that have the same conclusion part and an inclusion relationship from a rule group that is a collection of production rules. an inclusive rule extraction means for extracting a rule; an execution rule selection means for determining a production rule to be executed from among a plurality of production rules extracted by the inclusive rule extraction means; and a production rule determined by the execution rule selection means. and a production rule whose condition part has a non-inclusive relationship and was not extracted by the inclusive rule extracting means. The present invention is configured to include a rule execution means for executing a group of rules.

[Effect]

By taking the above-mentioned measures, the present invention extracts a plurality of production rules from a group of rules that have the same conclusion part and have an inclusion relationship in the condition part when a production rule is executed, and The production rule to be executed from among the extracted production rules is then determined by the execution rule selection means. Then, a rule group is reconstructed from one production rule selected from a plurality of production rules whose condition parts have an inclusive relationship and a production rule whose condition part is determined to have a non-inclusive relationship by the rule extraction means. , each production rule constituting the reconfigured rule details is executed by the rule group execution means. Therefore, the confidence of the conclusion part of each executed production rule is not calculated jointly but becomes a predetermined value, and the inferred conclusion part can be evaluated accurately. [Example] Hereinafter, length examples of the present invention will be described with reference to the drawings. FIG. 1 is a diagram showing functional blocks of a rule execution device according to an embodiment of the present invention. In this rule execution device, a rule having a certain event in a condition part or a conclusion part is extracted by an inclusive rule extracting means 2 from a rule base 1 in which a plurality of production rules are stored. The inclusive rule extracting means 2 extracts the second rule from among the plurality of extracted rules.
As shown in the figure, rules (1) to (
4> Extract. Although Figure 2 shows only the case where the event in the conclusion part is T, in reality the event in the conclusion part is T.
Similar extraction is performed for other rules as well. The execution rule selection means 3 selects the rule (1) with the strongest inclusiveness from among the rules (1) to (4) extracted by the inclusive rule extraction means 2 as the rule to be executed. Here, the judgment of inclusiveness is based on the conditional part with the largest number of events,
Or, if the number is the same, it is determined that the combination with fewer valid combinations has strong inclusiveness. For example, if the condition part of the rule is an inclusion relationship such as rules (5) to (7) shown in FIG.
), rule (6) has the largest number of events, rule (5),
Since rule (7) has the smallest number of valid combinations, rule (5) is selected as the execution rule.The rule group reconstruction means 5 extracts the rules selected by the execution rule selection means 3 and the included rule extraction means 2. A rule group is constructed from the rules determined to have a non-inclusive relationship.Then, the rule group execution means 5 executes each rule of the reconfigured rule group by the rule group reconfiguration means. According to the rule execution device, it extracts multiple rules in which the event in the conclusion part is the same and the event in the condition part is in an inclusive relationship, selects the rule with the strongest inclusiveness among them, and executes the selected rule and each other. Since the rule group is reconfigured from rules with non-inclusion relationships and this rule group is executed, multiple rules with the same conclusion part event and an inclusion relationship with condition part events can all be executed as before. Instead, only the rules that should be executed are executed. Therefore, the efficiency of rule execution can be increased, and when each rule is executed, the certainty of the event in the conclusion part is As a result, the inference result when the rule is executed can be evaluated accurately.In addition, the composite calculation of confidence can be performed between the included rules. Therefore, the confidence level of the rule can be set more easily than in the past, and the burden on the rule builder can be reduced.The present invention is not limited to the above embodiment.For example, as shown in FIG. As shown in FIG.
It is possible to manually select an execution rule from among the included rules extracted in step 1. That is, for each rule extracted by the inclusive rule extracting means 2, the inclusive rule editing means 6 edits the inclusive rule for each event in the conclusion part of each rule. Then, each edited inclusive rule is displayed on the display section of the execution rule manual selection means 7, which is a console or the like, and an arbitrary rule from among the inclusive rules is selected and executed from the input section manually. According to such a rule execution device, it is possible to have flexibility in determining execution rules, and even in a rule group consisting of several thousand rules, rules that have an inclusive relationship can be checked on the display of a console, for example. Therefore, it can be used as a support tool when constructing new rules. [Effects of the Invention] As detailed above, according to the present invention, it is possible to increase the efficiency of rule execution, and to accurately evaluate the conclusion part of each executed rule with a predetermined confidence level. I can do it,
Furthermore, it is possible to provide a rule execution device that can significantly reduce the burden on the rule builder.

[Brief explanation of drawings]

Fig. 1 is a functional block diagram of a rule execution device according to an embodiment of the present invention, Fig. 2 is a diagram showing an inclusive rule in which the events in the conclusion part are the same and the car image in the condition part is in an inclusive relationship, and Fig. 3 4 is a diagram illustrating the strength of the inclusive relationship of the conditional part. FIG. 5 is a functional block diagram of a rule execution device that manually selects an execution rule. is a diagram showing the structure of a knowledge base. 1... Rule base, 2... Inclusive rule extraction means,
3... Execution rule selection means, 4... Rule group reconstruction means, 5... Rule group execution means.

Claims (1)

[Claims] Inclusive rule extraction means for extracting production rules whose conclusion parts are the same and whose condition parts are in an inclusive relationship from a rule group that is a collection of production rules; an execution rule selection means for selecting a production rule to be executed from among the plurality of production rules selected; and a non-inclusive relationship between the production rule selected by the execution rule selection means and the condition part not extracted by the inclusion rule extraction means. A production system comprising: a rule group reconstruction means for reconstructing a rule group from production rules in the production rules; and a rule group execution means for executing the rule group reconstructed by the rule group reconstruction means. A rule execution device in a system.