JPH06259256A - Inference control method for expert system - Google Patents

Abstract

PURPOSE:To improve efficiency by hierarchizing inference control and minimizing an inference control range by dividing an inference process into worlds by inference functions and providing control rule sets in the respective worlds. CONSTITUTION:An inference process 2 is composed of an inference interface world 3 divided from a function plane and plural inference function worlds 41-4N. Concerning the inference control inside the respective inference function worlds 41-4N, Plural rule sets 62. in the worlds themselves are controlled through respective world memories 63 by providing respective control rule sets 61 and the inference is individually and simultaneously controlled inside the worlds. Concerning the inference control among the inference function worlds 41-4N, a control rule set 32 in the inference interface world 3 controls the control rule sets 61 in the inference function worlds through global memory 5 and totally controls the respective inference function worlds 41-4N. Thus, the inference can be efficiently performed and the inference process can be easily designed/ manufactured.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an inference control method for an expert system.

[0002]

2. Description of the Related Art An expert system has a knowledge base for storing expert knowledge, an inference engine, and an explanation mechanism, and the inference engine makes inferences using rules or frames with respect to external inputs such as user questions. Get the answer by.

The software structure of the expert system is generally divided into a peripheral process and an inference process,
The peripheral process performs processing such as system operation, data input / output, data management, setting change, drawing, calculation, printing, and engineering (maintenance), and the inference process performs knowledge inference processing.

[0004]

Since the expert system generally performs knowledge inference processing, the load of inference processing is heavy from the viewpoint of the entire system, and the processing speed becomes worse in a system that cannot perform efficient inference processing. In particular, when the number of inference rules or inference frames increases, the efficiency of inference processing becomes a problem in system evaluation and practical application.

Therefore, conventionally, there is a method of selecting an inference method such as forward inference or backward inference depending on an inference target, but it has been difficult to increase the efficiency of inference when the knowledge base becomes large.

An object of the present invention is to provide an inference control method that enables efficient inference processing.

[0007]

In order to solve the above-mentioned problem, the present invention provides an inference environment space in the inference process in an expert system in which a software configuration is a peripheral process and an inference process, and the inference process is performed by the inference process. Is divided into an inference interface world and a plurality of inference function worlds, a control rule set is provided in each world, and the control rule set of each inference function world collectively controls the rule sets in the self world, The world control rule set is characterized in that the plurality of inference function worlds are collectively controlled by a common access via a global memory.

[0008]

[Action] The inference process is divided into worlds according to the inference function,
In addition, by providing a control rule set in each world, the inference control is separated into the inference control in the world and the inference control between the worlds, and the inference control is hierarchized and the inference control range is minimized to be efficient. Get reasoning.

[0009]

1 is a software block diagram of an expert system showing an embodiment of the present invention. The peripheral process 1 includes software elements that perform processing such as data management as in the conventional case, and has a pair inference process interface to realize an interprocess data communication function with the inference process 2.

[0010] Inference process 2, the inference interface world 3 divided from the functional surface, composed of a plurality of inference functions World 4 ₁ to 4 _N, high inference environment space of possible independent interfere with each other only the global memory 5 (world)
Configured as.

[0011] Inference interface World 3 executes through the interface 3 ₁ a communication process between the peripheral process 1, inference environment for overall control of the inference function World 4 ₁ to 4 _N with the control rule set 3 ₂ Become a space. The world memory 3 ₃ is used for exchanging control data between the interface 3 ₁ and the control rule set 3 ₂ .

[0012] The inference function World 4 ₁ ~4 _N will become inference environment space making inferences corresponding to the function of each of the case of roughly the inference function for each function, each control rule set 6
The inference control for each function is performed by the control of each rule set 6 ₂ in the own world by _{1 and} the transfer of control data during that period is performed via each world memory 6 ₃ .

The global memory 5 is used for each world 3, 4
_{It is a} common memory that exists outside ₁ to 4 _N and can be commonly accessed from multiple worlds.

The inference control method in the inference process 2 in this embodiment will be described. For the inference control in each inference function world 4 ₁ to 4 _N , by providing each control rule set 6 ₁ , a plurality of rule sets 6 ₂ in the own world are controlled via each world memory 6 _3, and Control inference individually and collectively.

Inference function world 4₁~ 4_NReasoning control between
Is a control in the inference interface world 3
Rule set 3₂Is an inference machine via the global memory 5.
Each control rule set 6 in Noh world₁Control
And each inference function world 4₁~ 4 _NControl over.

That is, the control rule set 3 ₂ in the inference interface world 3 integrally controls each inference function world 4 ₁ to 4 _N, and each control rule set 6 ₁ in the inference function world 4 each rule in each world. This is an inference control method in which the set 6 ₂ is individually controlled collectively.

[0017]

As described above, according to the present invention, the inference environment space in the inference process is divided into an inference interface world and a plurality of inference function worlds, and a control rule set is provided in each world. The control rule set of the functional world collectively controls the rule sets in the self world, and the control rule set of the inference interface world collectively controls the plurality of inference function worlds by common access via the global memory. , Has the following effects.

(1) By dividing the inference process into worlds, the independence of each function can be improved.

(2) The inference control can be hierarchized by separating it into the inter-world inference control and the in-world inference control.

(3) The inference control range can be minimized by providing a control rule set in the world.

(4) Efficient inference can be performed by the above items, and at the same time, the inference process of the expert system can be easily designed, manufactured, and maintained.

[Brief description of drawings]

FIG. 1 is a software configuration diagram of an expert system showing an embodiment of the present invention.

[Explanation of symbols]

1 ... around the process 2 ... inference process 3 ... inference interface World 4 _1, 4 _N ... inference function World 5 ... global memory

Claims (1)

[Claims] 1. An expert system in which software configuration is a peripheral process and an inference process, and inference processing is performed in the inference process, an inference environment space in the inference process is divided into an inference interface world and a plurality of inference function worlds. A control rule set is provided in the world, the control rule set of each inference function world collectively controls the rule sets in the self world, and the control rule set of the inference interface world stores the plurality of inference function worlds in the global memory. An inference control method for an expert system, which is characterized by performing a centralized control by a common access via.