JPH02113338A - Knowledge acquisition control system for knowledge base system - Google Patents

Abstract

PURPOSE:To improve the quality of a knowledge base and to shorten the time for the structuring of a knowledge base system by providing a decision part which decides the contradiction and deficiency of knowledge. CONSTITUTION:A rule table 52 is generated and updated according to the contents of the knowledge base 2 and a decision pattern table 51 is generated according to the acquired knowledge of a knowledge acquisition part 6 before the knowledge is acquired by the knowledge acquisition part 6. A decision part 4 refers the added knowledge with the knowledge in the knowledge base 2 indicated by the rule table 52 to decide whether or not there is contradiction. Then when the contradiction is found, that is reported to the knowledge acquisition part 6 and when not, the addition of the knowledge is reported to a knowledge base control part 3 and the rule table 52 is updated to calculate a space that the knowledge in the knowledge base 2 does not cover, thereby deciding and reporting deficient knowledge to the knowledge acquisition part 6. Consequently, the knowledge is verified without performing inference actually by an inference execution part 7, to the knowledge base having no contradiction can be structured in a short time.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a knowledge acquisition control method for a knowledge base system.

[Conventional technology]

In a knowledge-based system, the knowledge of experts is
HEN Conclusion" is stored in a rule format called "HEN Conclusion," which is checked against separately given initial conditions. Inference processing is performed to draw a conclusion by repeatedly checking the new conclusion and conditions. For example, in fault diagnosis, the relationship between symptoms and failure location The location of the failure is determined by making inferences based on the rule description that indicates this and the state of the diagnostic target at the time of failure.

An example of the configuration of a conventional knowledge base system is shown in FIG. The knowledge base system 1 receives knowledge expressed in a tree structure or table from an expert through a terminal 8, and the knowledge acquisition unit 6, which inputs data to be inferred, uses the received knowledge to define the knowledge base system. The information is converted into a knowledge expression method such as a rule format, and sent to the knowledge pace management section 3. The knowledge pace management unit 3 manages the knowledge base 2 using the received knowledge retention/modification function. Knowledge base 2 accumulates the knowledge of experts. The inference execution unit 7 uses the contents of the knowledge base 2 and the terminal 8
Make inferences based on data input from.

[Problem to be solved by the invention]

There are many duplications and deficiencies in conditions and conclusions in the knowledge input by experts, and it is important to construct complete knowledge when acquiring knowledge.

However, in the past, in order to confirm inconsistencies and missing knowledge between knowledge input by experts, it was necessary to actually operate the system using the inference control unit 7 on the previously created knowledge base. There wasn't.

The purpose of the present invention is to store the knowledge of experts as a knowledge base, and the purpose of the present invention is to determine conditions and conclusions between the contents of the knowledge base and the contents of the knowledge base when storing the knowledge of experts as a knowledge base. It is believed that the quality of the knowledge base can be improved by pointing out inconsistencies between acquired knowledge and existing knowledge, and by pointing out incompleteness of the knowledge base by determining lack of knowledge based on the content of the knowledge base. The aim is to shorten the time required to build a knowledge base system.

[Means to solve the problem]

In order to achieve the above object, the present invention includes a knowledge base, a knowledge base management unit that manages the knowledge base, a knowledge acquisition unit, and an inference execution unit, as well as a determination unit that determines contradictions and deficiencies in knowledge, and a knowledge base management unit that manages the knowledge base, a knowledge acquisition unit, and an inference execution unit. A rule table expressing the space covered by the rules written on the base, a judgment pattern for contradictions, etc., and a judgment pattern table for determining the operation at the time of judgment are provided.

[For production]

The rule table is created and updated according to the contents of the knowledge base. The determination pattern table is created by the knowledge acquisition unit according to the knowledge acquired by the knowledge acquisition unit before the knowledge acquisition process. The knowledge acquisition unit requests the determination unit to, for example, determine the contradiction of additional knowledge. The judgment unit compares the additional knowledge with the knowledge in the knowledge base shown in the rule table,
Determine whether there is a contradiction, etc., and if there is a contradiction, notify the knowledge acquisition department of that fact, and if there is no contradiction, notify the knowledge pace management department of the addition of knowledge, then N will update the rule table,
Insufficient knowledge is determined by calculating the space that is not covered by knowledge in the knowledge base, and the knowledge acquisition unit is notified of this fact.

〔Example〕

An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 shows a configuration diagram of an embodiment of a knowledge base system according to the present invention. This knowledge base system 1 includes a knowledge base 2 that accumulates knowledge of experts, a knowledge base management section 3 that manages the knowledge base 2. In addition to the knowledge acquisition unit 6 that acquires knowledge from the terminal 8 and the inference execution unit 7 that performs inference based on the knowledge base 2, there is also a determination unit 4 that determines the contradiction between the acquired knowledge and the knowledge received from the knowledge acquisition unit 6. Equipped with: Table 5 includes a judgment pattern table 51 and a knowledge base 3 for expressing judgment patterns such as spear 14 and determining processing at the time of judgment.
It is made up of a rule table 52 that expresses the knowledge within.

The knowledge notified from the knowledge acquisition unit 6 is composed of a condition (IF) and a conclusion (THEN), and an example in which the condition and conclusion are composed of predicates connected with "and" will be described below. explain.

First, the knowledge acquisition unit 6 sets a judgment and a processing method at that time in the judgment pattern table 51 regarding a pattern such as a contradiction indicated by an inclusive relationship between a condition and a conclusion between two rules. FIG. 3 is an explanatory diagram of the determination pattern table 51. FIG. 3 (1) shows the inclusion relationship between conditions or conclusions.

FIG. 3 (2) shows a judgment pattern table 51 for diagnostic problems in which the conclusion consists of one predicate.
This is an example. FIG. 3(3) is an example of the shortage table 51. Figure 3 (3) shows a conceptual diagram of the space covered by insufficient knowledge. FIG. 4 shows an example of knowledge input to the knowledge acquisition section 6 in a diagnostic problem. Additionally, FIG. 4 also shows additional knowledge that will be used for subsequent explanations (FIG. 4 (2)).

The rule table 52 uses an expression method (referred to as a "positional cube") in which the dimensions of each rule are determined for each predicate constituting the condition and conclusion, and the position is determined for each predicate pattern. FIG. 5 shows positional cubes for each of the knowledge shown in FIG. 4. Figure 5 (1)
Figure 5 (2) is an example of a rule table expressed as if the knowledge in Figure 4 (1) was already in the knowledge base, and Figure 5 (2) is converted when the knowledge in Figure 4 (2) is input. Shows an example of knowledge.

The determination unit 4 converts the knowledge acquired by the knowledge acquisition unit 6 into a positional cube representation (FIG. 5 (1)), and combines the positional cube stored in the rule table 52 and the above-mentioned converted positional cube. By comparing, a contradiction is determined and missing knowledge is extracted from the positional cube stored in the rule table 52. Note that the knowledge acquisition unit 6 supplements the fact that the predicates that are not described in the acquired knowledge are in one of the situations.

FIG. 6 is a specific example of processing in the determining unit 4, and shows how additional knowledge is determined using the positional cube shown in FIG. The determination process is performed by performing a bit-AND between the positional cube in the rule table 52 and the positional cube to be added to check for common conditions and conclusions, and then comparing the pattern of the inclusion relationship with the determination pattern table 51. Based on the results, a judgment is made regarding inconsistencies, etc. (Fig. 6 (1)). Also, the universe (
Insufficiency is determined by subtracting all the positional cubes in the rule table (performing Sharp calculation) from the positional cubes (all of which are 1) (Figure 6 (2)).

〔Effect of the invention〕

As is clear from the above description, the knowledge acquisition control method of the present invention provides the following effects.

(1) By determining the relationship between knowledge acquired from the knowledge acquisition unit and already acquired knowledge,
Since knowledge can be verified without actually performing inference in the inference control unit, a consistent knowledge base can be constructed in a short time.

(2) A complete knowledge base can be constructed in a short time because it is possible to determine whether the knowledge stored in the knowledge base is lacking at the knowledge acquisition stage.

(3) As shown in the specific example, if the acquired knowledge is connected with "and", patterns such as contradictions can be grouped into 9 types, and the 1st judgment section determines which pattern the acquired knowledge is associated with. Since the applicability is formally determined, contradictory patterns within the knowledge base can be determined based on the meaning determined by the knowledge acquisition unit that uses this, and a flexible system can be configured for determination.

[Brief explanation of drawings]

Fig. 1 is a block diagram showing an example of the knowledge base system according to the present invention, Fig. 2 is a block diagram showing an example of the configuration of a conventional knowledge base system, and Fig. 3 is a diagram showing the relationship between conditions and conclusions. FIG. 4 is a diagram showing an example of a pattern table and deficiency determination. FIG. 4 is a diagram showing an example of rule description representing knowledge. FIG. 5 is a diagram showing an example of a rule table. FIG. 6 is a diagram of a determination unit to which the present invention is applied. FIG. 3 is a diagram showing a specific example of a processing result. 1...Knowledge base system, 2...Knowledge Heas,
3...Knowledge base management section, 4...Judgment section, 5...
Table, 51... Judgment pattern table, 52.
...Rule table, 6...Knowledge acquisition section, 7...
Inference control unit, 8...terminal. Foil 1 Figure 2 Figure 4 Condolence 4 Figure 5 Figure 3 Figure (2) Gi 1 Konoha O turn child -717 Cs'; 3.4 (1) Senzume Gin =) Ushishiba Hakudate ■) Et (■) Karidama Satoko @ Kanai mo - Chakake Cz fee 80 kito

Claims (1)

[Claims] (1) A knowledge base that accumulates the knowledge of experts, a knowledge base management unit that manages the knowledge base, a knowledge acquisition unit that acquires the knowledge necessary for the knowledge base, and the contents of the knowledge base and input from the outside. A knowledge base system comprising: an inference control unit that makes inferences based on data that is input to the knowledge base; A determination unit is provided that monitors the relationship between rules previously input to the knowledge base and rules newly added through the knowledge acquisition unit, and the determination unit determines missing knowledge based on the contents of the rule table. The knowledge acquisition unit is notified, and based on the contents of the rule table and the rules newly added through the knowledge acquisition unit, the relationship between the added rules is checked and discrepancies are determined according to the determination pattern table, and if there is a discrepancy, the above A knowledge acquisition control method for a knowledge base system, characterized in that a knowledge acquisition unit is notified, and if there is no contradiction, knowledge is added to the knowledge base management unit and a rule table is updated.