JPH03296140A - Knowledge information processing system - Google Patents

Abstract

PURPOSE:To advance an inference in order to obtain a conclusion by deciding automatically the difference of conditions between the rule knowledges even though the complete coincidence is obtained between the accumulated rule knowledges and a given prior fact and correcting the rule most approximate to the prior fact. CONSTITUTION:A prior fact is inputted and stored in a prior fact storage means 1. Then the related rule knowledge is taken out of a rule knowledge storage 2 and collated with the contents of the storage 1. The trade-off relation satisfied between the rule knowledges is detected based on the contents of the storage 1. If a condition part is not completely coincident with the contents of the storage 1 at execution of an inference, the conditions are eased and added to the rule that is not conflicting with the prior fact based on the result of a trad-off relation detector 3. Then the rule knowledge is corrected and the inference is carried on. Thus it is possible to advance the inference and a proper conclusion is obtained even though the conditions of the accumulated rule knowledges are not completely coincident with the contents of given fact.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a knowledge information processing system typified by an expert system that supports intelligent work such as consultation.

[Conventional technology]

A very high level of specialized knowledge is required to reach conclusions based on complex and diverse facts. Acquiring such specialized knowledge requires a long period of education and experience, and training skilled professionals requires a huge amount of money and time. Therefore, knowledge information processing devices have been developed that store the knowledge of skilled experts as rules and seek conclusions that fit various facts so that anyone can solve problems on the same level as a skilled expert. .

In conventional knowledge information processing systems, problem-solving methods are expressed as rule knowledge, and based on each given fact, the accumulated rule knowledge that completely satisfies the conditions is sequentially applied. By drawing conclusions while
I've been trying to solve the problem.

[Problem to be solved by the invention]

Conventional knowledge information processing systems advance inference based on the complete match between the conditions of accumulated rule knowledge and the content of given facts.

Therefore, the accumulated rule knowledge had to be completely consistent with the given facts.

The purpose of the present invention is to automatically determine the difference in conditions between the rule knowledge and correct the rule that is closest to the prerequisite fact, even if the accumulated rule knowledge does not completely satisfy the given prerequisite fact. The object of the present invention is to provide a knowledge information processing system that can proceed with inference and draw a conclusion.

[Means to solve the problem]

The knowledge information processing system of the present invention includes a premise fact storage means for storing premise facts to be processed, a rule knowledge storage means for storing rule knowledge for deriving a conclusion, the premise fact storage means and the rule knowledge. trade-off relationship detection means for determining the difference between the stored contents and the storage means; It is comprised of a condition relaxing means for modifying knowledge, and an inference means for drawing a conclusion from the modified rule knowledge obtained by the condition relaxing means.

[Effect]

In the present invention, facts serving as a premise are input and stored in a premise fact storage device. Next, the relevant rule knowledge is retrieved from the rule knowledge storage device and compared with the contents of the premise fact storage device. Here, a trade-off relationship established between rule knowledge is found based on the contents of the prerequisite fact storage device.

When making inferences, if the condition part does not completely match the contents of the premise fact storage device, based on the results of the trade-off relationship detection device, for rules that do not contradict the premise facts,
By relaxing and adding conditions, it is modified as fireable rule knowledge and inference is continued. In conclusion, the process is finished by presenting the final result of the inference and the conditions modified during the inference process.

〔Example〕

The present invention will be described below with reference to the drawings. In FIG. 1 showing an embodiment of the present invention, prerequisite facts input by a user are stored in a prerequisite fact storage device 1. As shown in FIG. The stored prerequisite facts are compared with a rule knowledge storage device 2 that stores expert know-how in the form of rules, and trade-off relationships between rules are stored in a trade-off relationship detection device 3. If none of the target rule knowledge matches the prerequisite facts, the condition relaxing device 5 modifies the rule knowledge. As a result, the inference mechanism 4 performs a comparison based on the given prerequisite facts and the revised rule knowledge, stores the revised contents of the condition part of the adopted knowledge in the revised condition holding device 6, and draws a conclusion. The result is stored in the conclusion holding device 7.

Expert know-how is described in the form of a rule: ``If conditional part), then execute (action part).'' For example, one of the pieces of knowledge in electronic circuit design is ``If the operating speed is 10 nanoseconds and the power consumption is 2.
0 mW is sufficient), then it can be expressed as (use electronic circuit block 'A'°)'. Similarly, if the operating speed is 15 nanoseconds and the power consumption is 12 mW, then
Suppose that the rule ``Use ``)'' is also given. Here, the prerequisite facts are that the operating speed is about 9 nanoseconds and the speed is 15 m.
If a circuit specification that achieves power consumption of W is given, a conclusion cannot be reached in a conventional knowledge information processing system, but according to the present invention, conditions regarding operating speed are relaxed and conditions regarding power consumption are assumed. Accordingly, the circuits of electronic blocks "A'" and "B'" are adopted as candidates. In this case, the trade-off relationship detection device 3 finds that there is a trade-off relationship between operating speed and power consumption, and concludes that electronic circuit block "A" with less trade-off relationship is appropriate for the preconditions. be able to.

FIG. 2 shows the steps of knowledge information processing to perform inference using the knowledge expressed in this way. In addition, P] ~ in the same figure
P7 indicates each step of flower 1~. In Pl, the facts that are the premise of the process are initialized. continue,
In P2, the condition part of the rule is compared with the stored rule knowledge, and if a plurality of rule knowledge are targeted, a trade-off relationship between the conditions is determined. If the condition is not satisfied in P3, the condition part of the rule knowledge is relaxed and added in P4. As a result, P5
In this step, it becomes possible to execute the operation part of the appropriate rule knowledge, maintain the modification conditions regarding the adopted rule knowledge (P6), and update the prerequisite facts (Pl).

If no termination command is written in the action section of the rule knowledge, the process returns to Pl and processes the next rule knowledge. By repeating these operations, it is possible to reach an appropriate conclusion based on the prerequisite facts.

〔Effect of the invention〕

As described above, according to the present invention, based on the given prerequisite facts, it is possible to quickly reach an appropriate conclusion using accumulated rule knowledge, just like a skilled expert. Even users with low skill level can perform tasks equivalent to those of skilled professionals.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention, and FIG. 2 is a flowchart illustrating the operation of the embodiment. ]...Premise fact storage device, 2...Rule knowledge storage device, 3...Trade-off relationship detection device, 4...Inference mechanism, 5...Condition relaxation device, 6...Modified condition retention device Device, 7... Conclusion holding device.

Claims (1)

[Claims] A premise fact storage means for storing premise facts to be processed, a rule knowledge storage means for storing rule knowledge for deriving a conclusion, and a difference in memory content between the premise fact storage means and the rule knowledge storage means. a trade-off relationship detection means to be sought; a condition relaxation means for modifying the rule knowledge by referring to the stored contents of the prerequisite fact storage means and the rule knowledge storage means and the difference determined by the trade-off relationship detection means; A knowledge information processing system comprising: inference means for drawing a conclusion from the modified rule knowledge obtained by the condition relaxation means.