JPH0275024A - Knowledge input method for expert system - Google Patents

Abstract

PURPOSE:To easily grasp the relation between rules by arranging the opinions serving as the elements of a rule condition part in a vertical row together with the rule hypothesis parts arranged in a horizontal row of a table respectively and adding the marks to the intersecting points between those vertical and horizontal rows for production of rules. CONSTITUTION:The opinions serving as the elements of a rule condition part are arranged in a vertical or horizontal column together with the rule hypotheses arranged in the other column respectively. Thus a rule table is produced and displayed on a screen 5. A mark is added to the intersecting point between the column of the table corresponding to the opinion of the rule to be produced and the column corresponding to the rule hypotheses via a man-machine interface MI. Then a rule describing part 6 detects the mark of the table, and the hypotheses and opinions are picked up out of the column related to the mark for production of a rule where the AND conditions of the relevant opinion are defined as a condition part. The produced rule is stored in a rule memory 1. Thus the rules are produced based on the opinions and hypotheses displayed on the screen 5 via a table or a tree structure. In such a way, the correspondence between the hypotheses and opinions can be easily grasped.

Description

DETAILED DESCRIPTION OF THE INVENTION A. Field of Industrial Application The present invention relates to a knowledge input method when constructing an expert system, which is one of the applied fields of artificial intelligence.

B1 Summary of the Invention The present invention provides a method for inputting knowledge of an expert system that performs inference by an inference mechanism using rules in a rule storage unit, in which each finding, which is an element of a conditional part of a rule, is arranged in a vertical column, for example. A table is created in which the hypothesis parts of rules are arranged in horizontal columns, and the intersection points are marked to create rules, making it easy to understand the relationships between rules. It is.

C0 Conventional technology One of the application fields of artificial intelligence is the expert system.This system incorporates the knowledge of experts into the computer F using knowledge engineering methods, and creates the same level of ability as the experts. The aim is to realize a system that has To explain a production system as an example of such an exuberant system, first, the knowledge necessary to solve a problem is shown in Figure 4.
This is stored in advance as a rule in the rule storage unit l via the man-machine interface Ml 17, and necessary information is written into the working memory unit (working memory) 2 via the man-machine interface Ml, for example. , the inference mechanism 3 selects a rule that matches the Moeki information from the rule group in the rule storage unit l, and applies that rule.

The rule is ``If the temperature is below T℃ -L, close the valve.''
It is expressed in the form IP-THEN~, where the IF~ part is called the condition part, and the THEN~ part is called the hypothesis (execution part). In the inference process, the data in the working memory unit 2 is updated in accordance with the application of rules, and finally when there are no more rules to apply.

Alternatively, the inference ends when a prespecified state is reached.

In order to store rules in the rule storage unit when constructing such a system, for each rule, the rule name, condition part, and hypothesis must be written according to the rule description grammar, for example, rule name IF A=B THEN I=1+1. The input was through a man-machine interface.

Problems to be solved by the B1 invention However, when entering rules in this way,
One problem is that as the number of rules increases, it becomes difficult to understand and organize the relationships between the rules. For this reason, incorrect description 1. It takes a long time and a lot of effort to detect the rules that have been created.

An object of the present invention is to solve such problems.

Means for Solving the IC3 Problem The first invention is [Arranging findings, which are elements of the condition part of a rule, in one column of vertical or horizontal columns, and arranging the hypotheses of the rule in the other column. Display the rule table created by 2 on the screen (2. Mark the intersection of the column of the table corresponding to the findings of the rule to be created and the column corresponding to the hypothesis of the rule via the man-machine interface. ]7. The rule description unit detects the mark in the table, picks out the hypothesis and findings in the column related to the mark, creates a rule with the AND condition of the finding as the condition part, and stores it in the rule storage unit. The second invention is characterized in that [findings that are elements of the conditional part of a rule and hypotheses of the rule are arranged separately on the screen, and the findings of the conditional part of the rule to be created are A line is drawn through the man-machine interface so that a tree structure is created from the hypothesis of the rule for all the rules, and the rule description section creates rules based on the tree structure and stores them in the rule storage section. IF, Embodiment FIG. 1 is a block diagram showing an example of an apparatus for carrying out the method of the present invention, and the same reference numerals as in FIG. 4 indicate the same parts. A rule table is constructed in the internal memory 4 by inputting data from the interface Ml, and this is displayed, for example, on the CRT screen 5.This table arranges findings, which are elements of the condition part of the rule, in vertical columns, and also displays the rules. Hypotheses are arranged in the horizontal column.As for the findings and hypotheses, assuming the rule ``If a certain value l is below the lower limit value, the pump l can be driven.'' [The part where a certain value 1 is just below the lower limit -1 corresponds to the finding, and the part that says "pump 1 is driven" corresponds to the hypothesis.

FIG. 2 is a diagram showing an example of a rule table, and in this example, there is provided a column 3 in which sentences of findings are digitized.

51"-84 in the table corresponds to the values 1-4, respectively. To enter a rule after creating such a table, enter the column corresponding to the finding in the condition part of the rule and the column corresponding to the hypothesis of the rule. Mark all the intersection points with the side that does If so, mark O in the column 1, 3, 5, or 7 in the column corresponding to hypothesis l.The rule description part 6 picks up the marks from the table and sets the AND condition of the marked finding as the condition part. A rule is created and stored in the rule storage section l.

Further, in the present invention, the entire rule may be displayed on the screen 5 shown in FIG. 1 as a tree structure as shown in FIG. 3. In this case, the findings and rules are arranged separately on the screen, and the man-machine interface, DICE M I Draw a line through. The O mark in the middle lower row of Figure 3 corresponds to the findings, the ○ mark in the middle row corresponds to the intermediate hypothesis, and the ○ mark in the -L row corresponds to the hypothesis. The line extending to the mark means that the latter holds true due to the former AND condition, and if there is an or (or) specification, it means that the OR condition holds -1). .

Note that the intermediate hypothesis is introduced to make the overall structure easier to understand, and to indicate the state in which the findings hold, for example, [Findings 1 and 2 hold simultaneously.] Thereafter, the rule description unit 6 creates rules based on this tree structure and stores them in the rule storage unit 1.

G1 Effects of the Invention According to the present invention, findings and hypotheses, which are elements of the condition part of a rule, are displayed on the screen in a table or tree structure, and rules are created based on this display. Since the correspondence between rules can be easily grasped, it is easy to understand the relationship between rules, and therefore, incorrectly written rules can be detected reliably in a short time, and the entire knowledge can be easily understood.

[Brief explanation of the drawing]

Fig. 1 is a configuration diagram showing a device used in an embodiment of the present invention, Fig. 2 is an explanatory diagram showing a state where a rule table is displayed on the screen, and Fig. 3 shows a state where a tree structure is displayed on the screen. The explanatory diagram, FIG. 4, is a structural diagram showing a conventional example. ... rule storage section, 2 ... working memory section, 3 ... inference mechanism, 4 ... internal memory, 5 ... screen, 6 ...
Rule description section. Configuration diagram of the device used in the example by two people 1... Rule storage section 2, working memory section 3... Inference mechanism 4... Internal memory 5... Screen 6... Rule description section 3rd Temporary screen display diagram of tree structure
hypothesis theory
Commentary See See See See See See See ql 2345678910 Figure 4 Configuration diagram showing a conventional example

Claims (2)

[Claims] (1) A method in which a rule consisting of a conditional part and a hypothesis is stored in a rule storage part, and the knowledge of an expert system that uses this rule to perform inference by an inference mechanism is inputted. A rule table created by arranging the rules in one of the vertical or horizontal columns and the rule hypotheses in the other column is displayed on the screen, and the table corresponding to the findings of the rule to be created is displayed. The intersection of the column and the column corresponding to the hypothesis of the rule is marked via the man-machine interface, and the rule description section detects the mark in the table and picks up the hypothesis and findings in the column corresponding to the mark. A knowledge input method for an expert system, characterized in that a rule is created having the AND condition of the finding as a condition part, and the rule is stored in a rule storage part. (2) In a method of inputting knowledge to an expert system that performs inference by an inference mechanism using the rules in the rule storage section, findings, which are elements of the conditional part of the rule, and the hypothesis of the rule are arranged separately on the screen. At the same time, lines are drawn through the man-machine interface to form a tree structure based on the hypothesis of the rule for all findings in the condition part of the rule to be created, and the rule description part is based on that tree structure. A knowledge input method for an expert system characterized by creating rules and storing them in a rule storage unit.