JPH0221335A - Back-up device for formation of knowledge base - Google Patents

Abstract

PURPOSE:To realize the automatic conversion of code trains of a production rule by showing the pairs of words and codes included in the sentences in a data base and showing the knowledge needed for generation of a complete production rule in a knowledge base for completion of the production rule. CONSTITUTION:A sentence interpreting device 1 stores the name, the conditional part and the conclusion part of a production rule inputted via an input device 13 into a 1st working memory 2 and the conditional and the conclusion part of the production rule into a 2nd working memory 3 respectively. In this case, the form of IF added with the rule name is stored into the function name parts 24 and 34 corresponding to the conditional parts 22 and 32 of both memories 2 and 3 respectively. At the same time, the form of THEN added with the rule name is stored into the function name parts 25 and 35 corresponding to the conclusion parts 23 and 33 of both memories 2 and 3 respectively. Thus it is possible to convert the production rule into a code train that can be applied to a computer.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a knowledge base construction support device used when constructing a knowledge base.

[Prior Art] Conventionally, this type of knowledge base construction support device has not been specially configured, and an operator converts production rules expressed in text into code that can be used by a computer.

[Problems to be Solved by the Invention] Conventionally, the work of constructing a knowledge base has been entirely done manually by humans, which has had the following problems.

(1) When converting written production rules into codes that can be used by a computer, workers need to study drawings, books, etc., and it takes an enormous amount of time to build a knowledge base.

(2) When converting written production rules into code that can be used by a computer, there is a risk that workers will misinterpret the rules, which may lead to building an incorrect knowledge base. .

Therefore, an object of the present invention is to provide a knowledge base construction support device that can convert production rules into a code string that can be used by a computer.

Means for Solving 1 Problem] The present invention has taken the following measures in order to solve the above problems and achieve the objective.

(1) A database of pairs of words and codes in sentences, which are necessary when converting written production rules into codes that can be used by a computer, is created and used as a dictionary.

(2) Complement production rules by creating a knowledge base of the knowledge necessary to construct complete production rules.

[Effect] By taking the above measures, the following effects are achieved.

(1) Written production rules can be automatically converted into code strings.

(2) Even if the written production rule is an incomplete sentence, this sentence can be complemented and converted into a complete sentence.

[Example] Figures 1 to 17 are diagrams showing an example of the present invention.
The figure is a block diagram showing the configuration of the knowledge base construction support device. The knowledge base construction support device includes a text interpretation device 1,
A first working memory 2, a second working memory 3, a third working memory 4, a word dictionary iF5,
a synonym dictionary 6, a sentence interpretation rule base 7, a function group database 8, a total i'lPI point database 9, a knowledge base editing device] 0, a first memory 11, a second memory] 2, and , an input device 13, and an output device 14.

FIG. 2 is a diagram showing the configuration of the first working memory. The first working memory 2 includes a rule name section 21 that stores a rule name of a production rule, a condition section 22 that stores a condition section of a production rule, a consequence section 23 that stores a consequence section of a production rule, and a consequence section 23 that stores a consequence section of a production rule. It is composed of a function name section 24 and a function name section 25 that store the function names of the section 22 and the consequence section 23.

FIG. 3 is a diagram showing the configuration of the second working memory. The second working memory 3 includes a condition part 32 that stores a condition part of a production rule, a consequence part 33 that stores a consequence part of a production rule, and a function that stores function names of the condition part 32 and the consequence part 33. It is composed of a name part 34 and a function name part 35.

FIG. 4 is a diagram showing the configuration of the third working memory. 41 is a word identification name.

Next, the operation of the knowledge base construction support device will be explained while appropriately showing the contents of the information stored in the first working memory 2, the second working memory 3, and the third working memory 4.

[Step 1] The output device 14 guides the operator to input the rule name 1 condition part and consequence part of the production rule.

[Step 2] The operator inputs written production rules as shown below into the input device 13.

Rule name: RULE1 Condition part: Pressurizer release valve, or Pressurizer safety valve opening operation Consequences: 2 Pressurizer relief tank water level.

Increase in temperature and pressure [Step 3] The text interpretation device 1 stores the rule name 1 condition part and the consequence part of the production rule input through the input device 13 in the first working memory 2, and stores the production rule condition part and the result part in the first working memory 2.
The resultant part is stored in the second working memory 3. At this time, the function name part 24.32 corresponding to the condition part 22.32 of the first working memory 2 and the second working memory 3.
34 is the form in which the rule name is attached to the IF, that is, r
IF-Rule Name" is stored. Similarly, the resultant part 2 of the first working memory 2 and the second working memory 3
3.The function name part 25゜35 corresponding to 33 contains THEN
is stored with a rule name attached to it, that is, rTHEN-rule name.

FIG. 9(a) is a diagram showing the contents of the information stored in the first working memory 2 after performing the processing in step 3, and FIG. 9(b) is a diagram showing the contents of the information stored in the first working memory 2 after performing the processing in step 3. 3 is a diagram showing contents stored in the second working memory 3. FIG. IF-RULE in function name part 24.34
I is stored in the function name section 25.35, and THEN-RULEI is stored in the function name section 25.35.

FIG. 5 is a diagram showing the structure of the word dictionary. The word dictionary 5 includes a word section 51 that stores words used to construct a knowledge base, and a word identifier section 52 that stores word identifiers that are paired with the words in the word section 51.

The word identification name section 52 stores word identification names indicated by a to g, which have the following meanings.

a: A component name related to a physical name.

b: Name of component related to work information.

C: Sent from each measuring device It is a measurable quantity.

d: Operation information.

e: Status information.

f: means rorJ.

g: means randJ.

[Step 4] The text interpretation device 1 performs the following processing on the text stored in the condition section 32 of the second working memory 3.

(1) The parts of the sentence that are not enclosed in parentheses are compared starting with the word with the longest character in the word part 51 stored in the word dictionary 5.

When a word in the word portion 51 matches a word in the sentence, the matched word in the sentence is enclosed in parentheses. The number of parentheses () in the sentence is calculated, and the word identification name 52 of the matching word is stored in the area of the third working memory 4 corresponding to this number.

FIG. 10(a) is a diagram showing the contents of the production rule stored in the working memory 2 after the above processing has been performed.

FIG. 10(b) is a diagram showing the contents of information stored in the third working memory 4 after performing the above processing. The word identification name 52 that is a pair with "nigashiben" in the word section 51 of the word dictionary 5 is rbJ, and rbJ is stored in the third working memory 4.

Further, as a result of comparing all the words in the word dictionary 5 with the sentence, if there is no matching word, the processing from (3) onwards is performed.

(2) Perform the processing in (1) above using the word in the sentence that matches in the processing in (1) as the starting point.

FIG. 11(a) shows the condition section 3 of the second working memory 3.
11(b) is a diagram showing the contents of the information stored in the second working memory 3 after performing the processes (1) and (2) above for all words in the second working memory 3. For all the words in the condition part 32 of the working memory 3 of
) is a diagram showing the arrangement of word identification names stored in the third working memory 4 after performing the process of (2).

(3) Of the sentences stored in the condition section 32 of the second working memory 3, the parts not enclosed in parentheses are deleted.

FIG. 6 is a diagram showing the structure of a synonym dictionary. The synonym dictionary 6 is composed of a non-standard word section 61 and a standard word section 62, and non-standard words stored in the non-standard word section 61 are used in written production rules. In such cases, it serves as a dictionary used to convert non-standard words into standard words.

(4) Compare each word enclosed in parentheses stored in the condition section 32 of the second working memory 3 with the word stored in the non-standard word section 61 of the synonym dictionary iF6. . If there is a match, the word in the standard language section 62 is replaced with the word in the standard language section 62 that is paired with the word in the non-standard language section 61. This verification work is performed from before the text in question. Figure 12 is a diagram showing the contents of the production rule stored in the second working memory 3 after the processes (3) and (4) above have been performed. It is.

In the condition part 32, the process (3) above results in the condition shown in FIG. 11(a).
) in the condition section 32 has been deleted. Also,
As a result of the process (4) above, the non-standard word "nigashiben" in the condition section 32 shown in FIG. 11 is replaced with the standard word "relief valve".

The sentence interpretation rule base 7 stores rules for supplementing incomplete production rules stored in the second working memory 3 and the third working memory 4 into complete sentences. This rule (a)～(
This will be explained in d).

(b) 41 of word identifier a immediately before the word of word identifier C
If there is no word, the word with the word identifier a that is the shortest before the word with the word identifier C is copied immediately before the word with the word identifier C.

(b) (1i) If there is no word with the word identifier d immediately after the word with the word identifier s, the word with the word identifier d that is the shortest distance after the word with the word identifier s is replaced with the word with the word identifier d. Copy immediately after.

(c) If there is no word with word identifier e immediately after the word with word identifier C, the word with word identifier e which is the shortest distance after the word with word identifier C is placed immediately after the word with word identifier C. Copy to.

(d) A group in which the sentence starts with the word with the word identification name a and ends with the word with the word identification name d or the word identification name e is considered to be a simple sentence. If there is no word with the word identification name f or word identification name g between the simple sentences, then the word identification name f or g that is the shortest after the simple sentence.
Copy the words between two simple sentences. Furthermore, if there is no word identification name f or word identification name g after the simple sentence, the word with word identification name g, that is, (and) is inserted between the two simple sentences.

(5) Based on the rules stored in the sentence interpretation rule base 7, the arrangement of the single 35 columns stored in the condition section 32 of the second working memory 3 and the word identification names 41 of the third working memory 4 Completion is performed.

FIG. 13(a) shows the second working memory 3 after processing based on the above rules (a) to (d) is performed on the condition part 32 of the second working memory 3 shown in FIG.
FIG. 13(b) is a diagram showing the arrangement state of word strings, and FIG. It is a figure which shows the arrangement|sequence of the word identification name 41 of the 3rd working memory 4 which changed accordingly. The arrangement of the word identifiers 41 shown in FIG. 11(b) is the same as the arrangement of word strings in the second working memory 3 shown in FIG. 12. Said (b)
Through this process, the sixth rdJ from the front in the third working memory 4 shown in FIG. 11(b) becomes the second rbJ from the front.
After that, a copy was inserted and became rabdfabdJ, and the first
The result is as shown in Figure 3 (b).

[Step 5] The text interpretation device 1 performs the same process as in step 4 described above on the text stored in the consequent section of the second working memory 3.

FIG. 7 is a diagram showing the structure of the function group database. The function group database 8 is composed of a word section 81 and a function name section 82. The word section 81 stores words representing state changes, actions, etc., and the function name section 82 stores codes that are paired with the words in the word section 81.

FIG. 8 is a diagram showing the structure of a database of 71,111 points in total. The total/I1 point database 9 includes a word part 91,
It is composed of a total of 1i1111 point name parts 92. The word part 91 stores words such as instrumentation device, sensor, etc., and the measurement point name of the measurement point base 92 includes the i of the word part 91.
Stores signal names related to the IT word. That is, the words in the word section 91, the measurement point name section device, 1...first memory, 2...second memory, 3...input device, 4...output device.

Claims (1)

[Claims] An input device for inputting production rules, a word dictionary storing words necessary for constructing a knowledge base and word identification names of these words, and pairs of standard words and non-standard words. Complete the production rule using the synonym dictionary that stores the synonym dictionary, the sentence interpretation rule base that stores the rules for making the production rule into a complete sentence, and the word dictionary, synonym dictionary, and sentence interpretation rule base. A sentence interpretation device that complements the sentences, a function group database that stores words that express state changes or actions, etc., and function names that pair with these words, and words such as instrumentation devices and sensors, and words such as instrumentation devices and sensors. a knowledge base editing device that uses a measurement point database storing pairs of signal names such as , and the function group database and measurement point database to convert word strings complemented by the sentence interpretation device so that they can be used by a computer; a first memory for storing information converted by the knowledge base editing device; a second memory for storing production rules inputted by the input device; and the text interpretation device and the knowledge base editing device. A knowledge base construction support device comprising: a working memory used for processing operations; and an output device for outputting information from the knowledge base editing device.