JPH01270141A - Inference system for expert system - Google Patents

Abstract

PURPOSE:To obtain inference with high functionality using an addition procedure by setting the start condition of the addition procedure, and performing the inference using the addition procedure based on the condition. CONSTITUTION:In an expert system of frame system having the addition procedure, the inference is performed by giving the attribute and the start conditional equation of each slot on a frame, and enabling the specification and the feasibility/infeasibility of start-up of the addition procedure to be selected from the conditional equation and the attribute at the time of referring to a slot value. Therefore, it is possible to set the start-up and the specification of the addition procedure from the attribute and the start conditional equation in spite of the presence/absence of the slot value, and to perform the inference by finding the slot value corresponding to a various kinds of status. In such a way, it is possible to attain the high functionality of the slot value and the inference using the slot value.

Description

DETAILED DESCRIPTION OF THE INVENTION A. Field of Industrial Application The present invention relates to an Exverton stem, and more particularly to an inference method using an additive procedure.

B1 Overview of the Invention The present invention improves the level of inference using the additional procedure by setting a starting condition for the additional procedure and performing inference using the additional procedure from this condition in a frame-based expert system having an additional procedure. This is intended to improve functionality.

C1 The conventional technical expert system is configured as shown in FIG. Knowledge Engineers (KE) utilize the knowledge of experts in each field.
) and store them in the knowledge base l, and the inference system 2 obtains answers to questions from the user by inference using the knowledge in the knowledge base l.

The inference system 2 has a description language 3 such as LISt) and PROLOG, uses a man-machine interface of a knowledge base editor 4 and a debugger 5 to store knowledge in the knowledge base 1, and uses a working computer for inference. While securing data such as knowledge base areas, rules, facts, etc. necessary for the work process in the memory 6, the inference engine 7 performs a mathematical inference.

In addition, an explanation function 8 is provided in the man-machine interface with the user to accept the intention of the question and display the inference procedure.

Here, the knowledge expression of the knowledge base l has an IF conditional section>
There are production rule expressions in the THEN (conclusion part) format and frame expressions in which abstract concepts and entities are expressed as frames in a hierarchical network. A fused hybrid system is known.

Also, known inference methods include forward inference leading from the condition part of a rule to the conclusion part, and backward inference leading from the conclusion part of the rule to the condition part.

In addition to such knowledge rules and inference methods, efficiency, speed, and functionality of inference can be improved by various methods such as conflict resolution means, working memory manipulation methods, meta-rule creation methods, and certainty factor introduction.

Here, the conventional method of knowledge rules and inference using frame representation will be explained in detail. FIG. 3 shows the frame structure of a vehicle, in which classes B1, B2, etc., such as small cars and light passenger cars, are included as sub-concepts of class A, which expresses the abstract concept of a passenger car. , B2 have instances CI, C2, C3, C4-Cn, which are concrete objects or concepts, respectively.

Each of these classes or instances is called a frame (or object), and each frame includes its attribute name (slot) and a description (value) indicating the attribute, as well as additional procedures (daemons) and procedural knowledge (methods). )
etc. are described.

For example, class A has 3 slots for car name, number of wheels, etc.
The number of wheels, 4, is described as a value that can be determined at the conceptual level of class A among these slots.

Additionally, the method describes a procedure for accumulating travel distance and fuel consumption using a procedure called movement. Further, the additional procedure describes an attribute-specific procedure that is automatically activated when the fuel gauge scale is unknown. ``Furthermore, in class B1, slots and values such as passenger capacity and fuel efficiency are determined, and in instance CI, substantive information such as car name, color, number, etc. is determined.

In such a frame representation, class slots serve as variables that can be referenced by all instances of the class, and additional procedures and class methods become procedural programs that can be referenced by all instances of the class. That is, a lower frame can inherit the attributes of an upper frame, thereby eliminating redundant representations of data. For example, in instance C1, the passenger capacity, fuel efficiency, and number of wheels are written in that slot, and by referring to classes Bl and 8, which are superordinate concepts, the passenger capacity, fuel efficiency, and number of wheels can be known. I can do it. In addition, by adding additional procedures, for example, if the scale of the fuel gauge in Mr. A's ton malfunctions and the remaining fuel cannot be determined,
The remaining fuel can be inferred by referring to the class A additional procedure and multiplying the distance traveled since the last refueling by the fuel consumption and subtracting it from the fuel tank volume h1.

D1 Problem to be Solved by the Invention In the conventional frame system, when the value of a slot within a frame is undetermined (unassigned),
Refer to the frame. Then, when the slot value is undetermined or unknown in the upper frame, an additional procedure is automatically activated to continue the inference that estimated the value of the slot.

However, the activation of the additional procedure is limited to when the value of the slot is undetermined, and this results in a functionally simple procedure, which poses a problem that makes it difficult to make inferences in accordance with various situations.

SUMMARY OF THE INVENTION An object of the present invention is to provide an inference method that can improve the functionality of addition procedures for slot values.

In order to achieve the above object, the present invention provides a frame-based expert system having an additional procedure, in which each slot in the frame has an attribute for setting handling conditions for the additional procedure. In addition, an activation conditional expression that provides a condition for activating the additional procedure is set, and inference is made by determining the selection and activation of the additional procedure from the above-mentioned attributes and activation method in the slot reference, and the attribute is set regardless of the value of the slot. The activation and type of additional procedures can be arbitrarily set from the activation conditional expression and the slot value corresponding to various situations is determined and inference is performed.

F. Embodiment FIG. 1 is a frame configuration diagram showing an embodiment of the present invention. The figure shows the frame structure of a polygon, with lower instances 12 and 13 of ``polygon'' in class 11 having ``triangle'',
゛Prepare a square f etc. 1 for instance 12.13
As shown in Figure 2, there are one corner and two corner slots. 3
Give the angle type, etc., data type, value, default value for each slot.

Give attributes, activation condition expressions, and additional procedures.

Here, the attribute sets handling conditions for the additional procedure in the slot, and allows setting of read-only, write-only, and read/write, and activates the additional procedure according to these conditions.

For example, if the attribute is read-only, the additional procedure will not be activated when writing to the slot, but will be activated only when reading from the slot.

On the other hand, the activation conditional expression provides a condition for a value to be written when writing or rewriting a slot, and allows activation of the additional procedure only when this condition is satisfied. Also, when reading, the slot value is read when it satisfies the conditional expression.

In this way, the attribute and activation conditional expression for each slot are given to the frame, and the type of additional procedure and whether or not it can be activated can be selected from the conditional expression and the attribute when referring to the slot value.

Therefore, depending on the attribute, a slot value can be given a special meaning when writing or reading. For example, conversion of angular units can be performed automatically. In other words, when a user manually infers the angles of a polygon system in degrees in radians, if a slot has a read-only additional procedure, the user can convert the angles at the time of readout. Inferences that use radians can be made using automatic conversion using an additional procedure.

On the other hand, an additional procedure can be started only when a condition is met using a startup conditional expression, and compared to conventional methods, which were executed every time with an undetermined slot value. It eliminates the need to compare conditions within a procedure, reduces the number of unnecessary additional procedure launches, and makes detailed condition settings easier because the launch condition descriptions are separated by slot.

Effects of the G0 Invention As described above, according to the present invention, since the activation of the additional procedure is set by the attribute of the slot and the activation conditional expression, it is possible to improve the functionality of the slot value and the inference using this. effective.

[Brief explanation of the drawing]

FIG. 1 is a frame configuration diagram showing one embodiment of the present invention, and FIG.
The figure is a basic configuration diagram of an expert system, and FIG. 3 is a conventional frame configuration diagram. ! ...Knowledge base, 2...Inference system, ll...
- Class, 12.13... Instance.

Claims (1)

[Claims] (1) In a frame-based expert system that has attached procedures, each slot in the frame has an attribute that sets the handling conditions for the attached procedure, and an activation conditional expression that provides the conditions for starting the additional procedure is set, and the slot reference is An inference method for an expert system, characterized in that inference is performed by selecting an additional procedure and determining whether or not to start it based on the attributes and the start conditional expression.