JPH11161498A - Analysis method for knowledge information, knowledge information processing system and storage medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide basic information for easily setting a knowledge expression clearly indicating the semantic contents of knowledge by analyzing/evaluating a knowledge description while considering the equality or similarity of its semantic contents. SOLUTION: A knowledge element extracting part 13 of a processor 10 extracts a knowledge element from knowledge information 110 composed of knowledge elements described in a rule type expression or procedure type expression. Based on the equality with the description contents of the knowledge element extracted by the knowledge element extracting part 13, an equality analysis part 14 generates a knowledge element dictionary 120 by merging and arranging the knowledge elements and based on the similarity to the description contents of the knowledge element extracted by the knowledge element extracting part 13, a similarity analysis part 15 generates the knowledge element dictionary 120 by aligning and arranging the knowledge elements.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a method for analyzing knowledge information, a knowledge information processing system and a storage medium, and more particularly to a method for analyzing knowledge information and a knowledge information processing system suitable for analyzing and evaluating knowledge information. And a storage medium.

[0002]

2. Description of the Related Art Knowledge information processing technology or artificial intelligence technology is used for advanced scheduling planning such as a plan for creating a train schedule or a crew schedule in a transportation system, and a tour maintenance plan for an elevator. As a conventional method of processing knowledge information, for example, US Pat.
As described in the specification of Japanese Patent No. 8,044, it is known that knowledge is described by rules and inference is performed using the knowledge.

However, US Pat.
In the method described in the specification of No. 44, the description of the purpose and execution means for executing the inference and the description of another method when a problem occurs in the inference execution are all described as rules. However, it is difficult to understand and confirm the description, and a problem is likely to occur from the viewpoint of reliability.
In addition, when dealing with an object with a large number of combinations and complicated processing, such as the creation of a train schedule for trains and crew members, the number of rules to be described is enormous, and it is easy to develop into a problem of processing capacity. Met.

On the other hand, for example, Japanese Patent Application Laid-Open No. 1-928
As described in Japanese Patent Publication No. 38, in addition to the description of the rules, the purpose for using these rules, the achievement policy / means thereof, and the like are described as a strategy, and selected in accordance with the set purpose. An information processing system that executes rules based on a strategy and advances inference is known.

[0005] According to this method, it is possible to separate the purpose and strategy for executing the problem solving and the planning and the rules necessary for the execution, and to describe them in association with each other. It is possible to set a plurality of strategies for a goal in order of priority using a goal / strategy net in which correspondences between goals / strategies are connected by AND or OR, and to assign a plurality of sub-goals to each strategy. Because it can be set, it is possible to divide difficult problems into simple problems and to describe them while organizing them. Moreover, it is possible to change strategies according to the situation and to make a well-balanced plan that cooperates between lower-level objectives.

[0006]

However, in the method described in Japanese Patent Application Laid-Open No. 1-92838, a rule type expression or a procedural type expression is used for lower-level knowledge. Since it is described in a language, it does not reach a knowledge expression that clearly expresses the meaning of the knowledge, so that there is a problem that the efficiency of understanding, adding, or modifying the description is low. For this reason, a method has been adopted in which a knowledge expression that clearly expresses the semantic content of knowledge is associated with each description of a rule type expression or a procedural expression. However, in general, the number of rules and procedures to be described is large, and it is troublesome to associate each description with a knowledge expression that clearly expresses the meaning.
In addition, there is a problem that it is difficult to properly use a knowledge expression that clearly expresses semantic content according to the equivalence or similarity of the knowledge description contents. In addition, if the number of rules and procedures is enormous, set different expressions for rules and procedures with the same semantic content, or set the same or similar expressions for rules and procedures with different semantic contents. There was also a problem that caused confusion.

Therefore, the present inventors have previously described in Japanese Patent Application No.
No. 302644 proposes a method for analyzing lower-level knowledge in which a rule type expression or a procedural expression is adopted. This method is intended to translate the low-level knowledge and create an easy-to-understand specification.

However, it has been found that, when translating, it is possible to more accurately set a knowledge expression by using basic information relating to an appropriate knowledge element, that is, information corresponding to a dictionary of knowledge elements.

An object of the present invention is to analyze and evaluate a knowledge description in consideration of the equivalence and similarity of its semantic contents, and to facilitate setting of a knowledge expression that clearly expresses the semantic contents of the knowledge. To provide basic information.

[0010]

Means for Solving the Problems (1) In order to achieve the above object, the present invention provides an analysis of knowledge information for analyzing knowledge information composed of knowledge elements described in a rule type expression or a procedural expression. In the method, the knowledge elements constituting the knowledge information are extracted, and the knowledge elements are arranged and integrated based on the analysis and evaluation of the equivalence and similarity to the description contents of the extracted knowledge elements. A knowledge element dictionary in which the integrated knowledge elements are arranged is generated. With this method, the knowledge description is analyzed and evaluated in consideration of the equivalence and similarity of the semantic contents, and a knowledge element dictionary that clearly expresses the semantic contents of the knowledge is generated, thereby setting the knowledge expression. It can be made easier.

(2) In the above (1), preferably,
The above-mentioned integration of the knowledge elements is performed by unifying the expressions of the knowledge elements having the same semantic contents although the expression methods are different. With this method, the equivalent knowledge elements can be unified, and the knowledge expression can be easily set.

(3) In the above (1), preferably,
The arrangement of the knowledge elements is performed by organizing the knowledge elements based on the similarity of the expression method. According to such a method, similar knowledge elements can be arranged to facilitate setting of a knowledge expression.

(4) In the above (1), preferably,
A plurality of knowledge elements having a connection of semantic contents and a unity are treated as one knowledge element, and a knowledge element dictionary in which the above knowledge elements are arranged is generated. According to such a method, a set of knowledge elements can be easily arranged by organizing them as a single unit.

(5) In the above (1), preferably,
The extraction of the knowledge element is performed with reference to the already created knowledge element dictionary, and a new knowledge element is added to the already created knowledge element dictionary to generate a knowledge element dictionary. With this method, it is possible to easily set a knowledge expression using an existing knowledge element dictionary.

(6) In the above (1), preferably,
A new knowledge element dictionary is generated by changing the names of the knowledge elements constituting the generated knowledge element dictionary. With this method, the knowledge element dictionary can be edited to facilitate setting of the knowledge expression.

(7) In order to achieve the above object, the present invention relates to a knowledge information processing system having a processing device for analyzing and processing knowledge information composed of knowledge elements described in a rule type expression or a procedural expression. The processing device integrates the knowledge element based on equivalence to the description content of the knowledge element extracted by the knowledge element extraction unit that extracts the knowledge element from the knowledge information, and the knowledge element extracted by the knowledge element extraction unit. An equivalence analysis unit that generates a knowledge element dictionary by arranging the knowledge elements, and arranges and arranges the knowledge elements based on the similarity to the description content of the knowledge elements extracted by the knowledge element extraction unit. And a similarity analysis unit for generating a dictionary. With this system, knowledge descriptions are analyzed and evaluated in consideration of the equivalence and similarity of their semantic contents, and a knowledge element dictionary that clearly expresses the semantic contents of the knowledge is generated to set the knowledge expression. It can be made easier.

(8) In the above (7), preferably,
8. The knowledge information processing system according to claim 7, wherein the equivalence analysis unit registers a knowledge element obtained by integrating knowledge elements having the same description content in an integrated knowledge element table,
The knowledge element names of the knowledge elements having the same description contents are stored in the same knowledge element table, and the same knowledge element table is linked to the integrated element table so that the two are associated with each other.

(9) In the above (7), preferably,
The processing apparatus further includes a knowledge element editing unit that changes the name of the knowledge element that forms the generated knowledge element dictionary and edits the name of the knowledge element into a new knowledge element dictionary. With such a system, the knowledge element dictionary can be edited and the knowledge expression can be easily set.

(10) In order to achieve the above object, the present invention provides a knowledge element extracting unit for extracting a knowledge element from knowledge information described in a rule type expression or a procedural expression in a storage medium, An equivalence analysis unit that integrates and arranges the knowledge elements to generate a knowledge element dictionary based on the equivalence to the description contents of the knowledge elements extracted by the extraction unit, and an extraction unit that is extracted by the knowledge element extraction unit. A similarity analysis unit that arranges and arranges the knowledge elements based on the similarity to the description contents of the knowledge elements and generates a knowledge element dictionary is provided.

[0020]

DESCRIPTION OF THE PREFERRED EMBODIMENTS A method for analyzing knowledge information according to an embodiment of the present invention will be described below with reference to FIGS. First, a configuration of a knowledge information processing system using a method for analyzing knowledge information according to an embodiment of the present invention will be described with reference to FIG.

The processing device 10 is the main body of the information processing system, and includes a command receiving unit 11, a knowledge element extracting unit 13, an equivalence analyzing unit 14, a similarity analyzing unit 15, and a knowledge element editing unit 1.
6 and a display unit 18. Programs for controlling the functions of these units are stored in a main storage device (not shown) inside the processing device 10. This program is a portable medium such as an optical disk or floppy disk.
2 is installed in the main storage device of the processing device 10 using the medium drive 20.

The storage device 100 is a part of the main storage device of the processing device 10 or an external storage device connected to the processing device 10, and includes a plurality of procedure / rule type expression knowledge information 110 which is knowledge information,・ Rule type expression knowledge information 11
The knowledge element dictionary 120 and the same knowledge information 130 which are knowledge analysis result information obtained by analyzing 0 are stored.

The procedure / rule type expression knowledge information 110, which will be described later with reference to FIG. 2, is information in which knowledge necessary for inference is described in a programming language or a rule type language. This is a program for creating a patrol maintenance plan for elevators. As a programming language, for example, there is C language, COBOL, and the like. A rule-based language is a language based on rules uniquely created by the creator.

The knowledge element dictionary 120 is a table for storing knowledge elements extracted from the procedure / rule type expression knowledge information 110, the details of which will be described later with reference to FIG. The same knowledge information 130 is a table for storing a result of analyzing and evaluating the knowledge elements extracted from the procedure / rule type expression knowledge information 110, and details thereof will be described later with reference to FIG.

The input device 30 is connected to the processing device 10 and inputs instructions for analyzing and evaluating the procedural / rule-type expression knowledge information 110 and a keyboard and a pointing device for inputting conditions for the analysis and evaluation. Device. The display device 40 is connected to the processing device 10,
This is a device that displays the analysis and evaluation results of the knowledge element dictionary 120 and the like.

Next, the operation of each unit constituting the processing apparatus 10 will be described. The command receiving unit 11 is a processing unit that transfers control to the knowledge element extraction unit 13 or the equivalence analysis unit 14, the similarity analysis unit 15, and the knowledge element editing unit 9 in accordance with an instruction input from the input device 30.

The knowledge element extracting unit 13 is a processing unit that extracts knowledge elements constituting the knowledge description contents from the designated procedure / rule type expression knowledge information 110 and creates a knowledge element dictionary 120. The detailed operation will be described later with reference to FIG. The equivalence analysis unit 14 integrates, for the knowledge elements in the knowledge element dictionary 120 created by the knowledge element extraction unit 13, elements that are duplicated or elements in which the same content is separately expressed, and the like. And a detailed operation thereof will be described later with reference to FIG. The similarity analysis unit 15 classifies and arranges the knowledge elements in the knowledge element dictionary 120 corrected by the equivalence analysis unit 14 based on the similarity and relevance of the knowledge elements, and arranges the knowledge elements in the knowledge element dictionary 120. This is a processing section, and the detailed operation thereof will be described later with reference to FIG. The knowledge element editing unit 16 is a processing unit that accepts an editing operation such as setting of a knowledge element name representing a meaning content for each knowledge element in the knowledge element dictionary 120 created by the knowledge user. Will be described later with reference to FIG.

The display unit 18 includes a knowledge element extraction unit 13, an equivalence analysis unit 14, a similarity analysis unit 15, and a knowledge element editing unit 16.
Is a processing unit that displays the result of extraction, analysis, evaluation, and editing on the display device 40.

In FIG. 1, general software such as an operating system (OS) or a window system intervening between the storage device 100, the input device 30, the display device 40, and the above-described processing unit is omitted. I have.

Next, the flow of processing of the method for analyzing knowledge information according to one embodiment of the present invention will be described with reference to FIG. In the method of analyzing knowledge information according to the present embodiment, a knowledge element is extracted, analyzed, and evaluated for the procedure / rule type expression knowledge information 110 stored in the storage device 100 of FIG. The dictionary 120 is displayed on the display device 40.

In step 200, when a knowledge information name is input from the input device 30 shown in FIG.
The instruction receiving unit 11 of 0 receives this input. here,
The knowledge information name is a file name of the procedure / rule type expression knowledge information 110 stored in the storage device 100,
For example, in the case of a train schedule creation plan, "TrainDia
g-1 ". Knowledge information to be extracted / analyzed / evaluated for knowledge elements is stored in advance in the storage device 100 as procedure / rule type expression knowledge information 110.
In step 210, the command receiving unit 11 of FIG. 1 executes the procedure / rule type expression knowledge information 11 having the designated name.
0 is retrieved from the storage device 100 and stored in the main storage device in the processing device 10.

Here, a specific example of the knowledge description contents of the procedure / rule type expression knowledge information 110 stored in the storage device 100 will be described with reference to FIG. An example of the knowledge information shown in FIG. 3 is a description example based on a description rule of the C language, which is one of the general-purpose programming languages, and this knowledge information is an object of the analysis and evaluation according to the present embodiment.

The description content is the most general "IF to THE
N (〜ELSE) ”, a loop description such as“ FORＦ ”, and a process description such as value setting and subroutine call. Also, as shown outside the frame of FIG. 3, line numbers corresponding to each line of the knowledge description content are given. The description contents shown in FIG. 2 are for explaining a method of analyzing knowledge information according to the present embodiment described later, and the contents themselves have no special meaning.

The condition description is, for example, in “if (wtime == 20)” described in line number 11, “i”
“wtime == 20” in parentheses of “f” corresponds to this. Note that the condition description includes a single condition as in line number 11 and a compound condition in which a plurality of conditions are combined by AND or OR as described in line numbers 13 to 14. The loop processing description is, for example, in “for (i = 0; i <n; i ++)” described in line number 21, “i = 0, i <n; i
++ ". The processing description corresponds to, for example, “x = y + 20” described in the line number 12.

In this embodiment, since the description expression based on the description rules of the C language is used, AND
Is described with “&&”, OR is described with “||”, and inequality signs (<,>, ≦, ≧) are “<”, “>”, “<=”, “>”, respectively.
= ”, And the equal sign (=) and the inequality sign (≠)
= ”And“! = ”, And the end of each processing description is described with a semicolon“; ”.

Next, in step 220 of FIG. 2, the display unit 18 of the processing device 10 displays a process selection instruction via the display device 40, and responds to this request by the input device 30.
The processing options are input from. Processing options include:
"Create new dictionary", "Create with existing dictionary", and "Edit dictionary" are provided.

Further, in step 220, when an instruction of "creation of a new dictionary" is input from the input device 30, the knowledge element extraction unit 13 is activated. Step 220
, When an instruction of “creation using an existing dictionary” is input from the input device 30, the display unit 18 displays the display device 4
At 0, the input request of the knowledge element dictionary name is displayed to the user.

In step 230, the knowledge element dictionary name is input from the input device 30. For example, "Trai
Extraction, analysis and evaluation of knowledge elements are performed on the procedure / rule-type expression knowledge information of the train schedule creation plan named "nDiag-0". As a result, a knowledge element dictionary called "TrainDic-0" is created. If this knowledge element dictionary is used, there is a part that can be used in "TrainDiag-1" which is the target of extraction and analysis evaluation of this knowledge element, so this knowledge element dictionary "TrainDic-0" Enter

Next, in step 240, the command receiving unit 11 retrieves the existing knowledge element dictionary 120 stored in the storage device 100, stores it in the main storage device in the processing device 10, and extracts the knowledge element extraction unit. 13 is started.

Next, in step 400, the knowledge element extraction unit 13 of the processing device 10 extracts knowledge elements from the procedure / rule type expression knowledge information 110 extracted in step 210. Here, the details of the knowledge element extraction processing executed by the knowledge element extraction unit 13 will be described with reference to FIG.

In step 405, the knowledge element extraction unit 13 shown in FIG. 1 performs initial setting of the extraction processing. That is, the work area for extraction prepared in the main storage device of the processing device 10 is cleared and the initialization is performed.

Next, in step 410, the knowledge element extraction unit 13 positions the row pointer at the first row of the procedure / rule type expression knowledge information 110 fetched in step 210.

In step 415, the knowledge element extracting unit 13 extracts the knowledge description content for one row at the row pointer position. For example, “dtime == 0” described in the first line (line number 1) of the procedure / rule type expression knowledge information 110 as shown in FIG. 3 is extracted. Here, if there is no description in the corresponding line, the process ends.

Next, the processing of steps 420 to 450 will be described. Here, for convenience of explanation, the line number 1 of the procedure / rule type expression knowledge information 110 as shown in FIG.
The description will be specifically made taking the description of 0 to 14 as an example. In step 450, the knowledge element extraction unit 13 updates the row pointer.
It is assumed that the line pointer is located at line number 10. Therefore, in step 415, the knowledge element extraction unit 13 sets “if (wti)” described in the row number 10.
me == 10) x = y + 10 ".

In step 420, the knowledge element extracting unit 13 extracts the description contents of the condition description and the process description and the description position from the extracted knowledge description contents. In the example of the knowledge description in the C language shown in FIG. 3, the condition description is a portion enclosed in parentheses following a reserved word in the C language such as IF, ELSE, or FOR, and the processing description ends with a semicolon “;”. The individual parts are separated. The knowledge element extraction unit 13 is configured to automatically set a tentative name based on the description position in the procedure / rule type expression knowledge information 110. For example, when the extracted description content starts from the m-th character in n lines If the knowledge element name is "condition nL
m ”and“ processing nLm ”. For example, for the description of line number 10 in FIG.
For time = 10, the line number is “10” and “w” is the fifth character (because there is a blank space for one character before “if” in line number 10). Therefore, “condition 10L5” is set as the knowledge name, and the description content for this is extracted as “wtime == 10”. Similarly, "condition 10L18" is set as the knowledge name for line number 10, and the description content for this is "x
== y + 10 ”.

Next, in step 420, the knowledge element extraction unit 13 determines whether or not there is a continuation to the next line. If there is a continuation, the process proceeds to step 425, and if not, the process proceeds to step 500. . This is because the description content may be described over a plurality of lines.
For example, the description content of the line number 10 shown in FIG. 3 is a case where there is no straddle, and the line numbers 13 and 14 are an example of a case where there is a straddle.

Here, assuming that the description of line number 10 is to be continued and there is no straddle, the process proceeds to step 600 and will be described. The case where there is straddling will be described later. In step 600, the knowledge element extraction unit 13 registers the knowledge element name and the description content extracted in step 420 in the extraction work area in the main storage device of the processing device 10. Step 60
In the processing at 0, the knowledge element registration processing in the case of the judgment condition description is different from the other knowledge element registration processing, and details thereof will be described later with reference to FIG.
The registration of the knowledge element with respect to 0 is performed as follows.

Here, the knowledge element registration process for the extraction work area will be described with reference to FIG. The extraction work area 500 includes a knowledge element name 510 and a description content 5
20. As described above, in the step 420 of FIG. 4, in the extraction process for the row number 10 of FIG. 3, “condition 10L5” is set as the knowledge name, and “wttime” is described as the description content thereof
== 10 ”are extracted and stored in the knowledge element name 510 and the description content 520 as shown in FIG. Similarly, “Condition 10L18” is set as the knowledge name for the row number 10, and when “x == y + 10” is extracted as the description content for this, as shown in FIG. Element name 510 and description content 52
0 is stored.

Next, in step 445 of FIG. 4, the knowledge element extracting unit 13 determines whether or not there is another condition description or processing description in the rest of the description extracted in step 420. Returning to step 420, if not, proceed to step 450.

Here, since there is no remaining description for the line number 10, there is no remaining description in the determination in step 445, and the process proceeds to step 450.

At step 450, the knowledge element extracting unit 13 updates the row pointer, and returns to step 420.
That is, the row pointer is updated from “10” to “11”, and in step 420, the knowledge element with the row number 11 is extracted.

Next, from step 425 to step 430
The case of proceeding to will be described. For example, step 420
In FIG. 3, “if ((wti
extraction of me> = 30 && z >> = 10) is performed based on the knowledge element name “condition 13L5” and the description content “wtime> =
30 &&z> = 10 ”, then, in step 425, the line number 13 is changed to the line number 14
Therefore, it is determined that there is a “straddle”, and the process proceeds to step 430.

At step 430, the knowledge element extracting unit 13 updates the row pointer. That is, the row pointer is set to “11”. Next, in step 435, the knowledge element extracting unit 13 extracts the knowledge description content of the next one-line sentence. That is, the knowledge element extraction unit 13
Is "|| wtime" described in line number 14.
== dtime + 10 &&z> = 20) x =
10 + y "is taken out.

Further, at step 440, the knowledge element extracting unit 13 extracts a continuation part of the condition description and the process description, adds the extracted part to the description content extracted from the previous line, and returns to step 425.

Next, referring to FIG. 6 and FIG.
Details of the knowledge element registration process at 00 will be described. In step 610, the knowledge element extraction unit 13
Regarding the description contents extracted in step 420,
A description type such as a condition description and a process description is determined. The description type is determined based on reserved words and grammar used in languages and rules. For example, in the knowledge description example in the C language shown in FIG. 3, the description type is determined in accordance with the grammar of the C language. For example, a description enclosed in parentheses following “IF” or “ELSE” is a branch. The description is a determination condition description for processing, and the description enclosed in parentheses following “FOR” is a condition description for loop processing. A description delimited by a semicolon ";" is a processing description for setting a value or calling a function. In the present embodiment,
Declaration descriptions, such as variable and function declarations, are not considered as knowledge elements, and are therefore excluded from extraction.

After determining the description type in step 610, except for the condition description and the declaration description for determination,
In step 620, the knowledge element extracting unit 13
510 of the extraction work area 500 shown in FIG.
And the description content 520 are registered. For example, for the description content “x = y + 20” for the line number 12 shown in FIG. 3, “process 12L3” is registered as the knowledge element name 510 and the description content 520 is “x = Y + 20 ”is registered. Step 610
If the description type is determined to be a declaration description, no processing is performed on the declaration description.

Next, if it is determined in step 610 that the condition is a condition description for determination, the process proceeds to step 630. When the description type is a condition description for determination, generally, a single condition that does not include “AND” or “OR”
Since there may be a complex condition in which single conditions are complicatedly linked by “AND”, “OR”, and “brackets”, in the present embodiment, registration of the entire complex condition as registration in the extraction work area 500 includes: The multi-stage registration of registration of the individual conditions constituting this, and the registration of the condition part enclosed in parentheses is employed.

Therefore, in step 630, the knowledge element extraction unit 13 sets the extraction work area 500 shown in FIG.
The knowledge element name 510 and the description content 520 are registered. For example, the description content “wt” for line number 13 shown in FIG.
image> = 30 &&z> = 10 ”, as shown in FIG.
3L5 ”is registered, and“ wtim ”is described as the description content 520.
e> = 30 &&z> = 10 "is registered.

Next, at step 640, the knowledge element extraction unit 13 decomposes the judgment condition description. Here, with reference to FIG. 7, the details of the decomposition processing of the judgment condition description will be described.

In step 641, the knowledge element extraction unit 13 determines whether there is “AND” or “OR” in the condition description. If there is no condition description, the condition description is not decomposed and no processing is performed. If so, the process proceeds to step 642.

Next, in step 642, the knowledge element extraction unit 13 determines the presence or absence of parentheses in the condition description, and if not, proceeds to step 643;
Proceed to.

In step 643, if there is no parenthesis, the knowledge element extraction unit 13 extracts a single condition from the condition description, and in step 644, executes step 620.
Similarly to the above, the information is registered in the extraction work area 500 shown in FIG. 5, and the processing of steps 643 and 644 is repeated for all the single conditions in the condition description. For example,
The description content “c + d == e” of the line number 23 shown in FIG.
&& f! = G ”, the description content“ c + d =
= E "and the description content"&& f! = G "are extracted and registered in the extraction work area 500 shown in FIG. 5 together with the knowledge element names.

Next, the processing in the case where there is a parenthesis in the determination in step 642 will be described. In step 645, the knowledge element extraction unit 13 cuts out the conditional description in parentheses. Next, in step 646, the knowledge element extraction unit 13 registers the entire condition description in parentheses in the extraction work area 500 as in step 620. Further, in step 647, the knowledge element extracting unit 13 decomposes the cut-out condition description in the same manner as in step 640 shown in FIG. Then step 6
At 48, the knowledge element extraction unit 13 removes the step-cut decision condition description and the parentheses surrounding it from the original decision condition description before clipping, and returns to step 642.
The presence or absence of another parenthesis in the determination condition description is determined. As described above, in the processing of step 640 and step 647, the processing of step 641 to step 648 described above is performed recursively.

For example, the description contents of line numbers 13 and 14 shown in FIG. 3 ((wtime> = 30 &&z> = 10)
|| wtime == dtime + 10 &&
For z> = 20, the description content “wtime> = 3
0 &&z> = 10 ”, and the description contents“ wtime> = 30 ”and the description contents“ z> = 10 ”, which are the sole conditions, are extracted from the extracted work area 5.
00 is registered. Similarly, the description content “wtime
== dtime + 10 &&z> = 20 ”, and the description content“ wttime ”which is a single condition
> = 30 ”and the description content“ z> = 30 ”are registered in the extraction work area 500.

With the above, the process of extracting the knowledge element name and the description content from the description content by the knowledge element extraction unit 13 described with reference to FIGS. 4, 6 and 7 and registering them in the extraction work area 500 is completed. FIG. 5 shows the row number 10 shown in FIG.
Extraction work area 5 for description content of ~ line number 14
00 shows the registered content of the knowledge element name 510 and the description content 520.

Next, returning to FIG. 2, in step 900, the equivalence analysis unit 14 of the processing apparatus 10 shown in FIG. 1 is activated, and the equivalence analysis evaluation and integration of the knowledge elements are performed. Here, an example in which the description content of the knowledge element can be evaluated to be semantically equivalent will be described with reference to FIG. In the example shown in FIG. 8, a description will be given taking a C language expression as an example. FIG. 8 (a)
Is an example of the description content of the knowledge element when the meaning is the same regardless of the presence or absence of white space characters. Specific examples are shown. In this example, "&
A case is shown in which a rule is provided only before and after “&” and “||”, and the knowledge element of the process description is only before and after the equal sign. For example, the description content “a = b + c +
d ", description content" a = b + c + d ", and description content" a
= B + c + d "has the same meaning regardless of the presence or absence of a blank character. Therefore, these descriptions are integrated into “a = b + c + d”.

FIG. 8B shows an example of the description contents of a knowledge element including parentheses which have no effect on the meaning even if removed, and the direction in which the parentheses are not required may be omitted. Shows an example of integrating knowledge elements. For example, the description content “a = b + (c + d)” and the description content “a = (b + c) + d” are descriptions of knowledge elements that have no effect on the meaning even if parentheses are removed. The contents are integrated into “a = b + c + d”.

FIG. 8 (c) shows a specific example in which differences in description contents are absorbed by changing the order of items or the order of items, and knowledge elements are integrated. In this example, an example is shown in which the left-hand side is shifted from the left-hand side to the right-hand side so that the description on the left-hand side is unary, and the left-hand side is changed so that the constant term comes to the end of the expression. For example, the description content “a = 10 +
b "and the description content" a-10 = b "have the same description content even if the order of the transitions and terms is changed, and these description contents are integrated into" a = b + 10 ". .

FIG. 8 (d) shows an example of a grammatically equivalent expression of a language used in the description of the description contents, and shows a specific example in which a plurality of expression methods are integrated into a specific expression. An example is shown. For example, the description content “a = b =
0, the description content “a = 0, b = 0”, and the description content “a = 0; b = 0” are grammatically equivalent expressions, and these description contents are “a = 0”. ;
b = 0 ".

Here, referring to FIG.
A description will be given of the equivalence analysis evaluation and integration processing of the knowledge elements in the above. First, the equivalence analysis unit 14 determines in step 90
Preprocessing for equivalence analysis evaluation from 5 to step 935 is performed, and then, equivalence analysis evaluation processing from step 1200 and steps 940 to 995 is performed.

First, for pre-processing for equivalence analysis evaluation, in step 905, the equivalence analysis unit 14
Perform initial settings for the work area. In this initial setting, the work area for extraction 500 shown in FIG. 5 is changed to a work area for equivalence analysis.

Next, in step 910, the equivalence analysis unit 14 sets the read pointer to the work area 5 for equivalence analysis.
00 and the step 9
At 15, the equivalence analysis unit 14 extracts the knowledge element at the position of the read pointer.

At step 920, the equivalence analysis unit 1
4 performs white space correction on the extracted knowledge elements according to the rule shown in FIG. 8A, and then, in step 925, as shown in FIG. Remove parentheses. Further, step 930
In FIG. 8 (c), if there is a constant term on the left side as shown in FIG. 8 (c), the transposition is performed. Further, in step 1000, as shown in FIG. Unification of

Here, the unifying process of the description of the equivalent expression in the language will be described with reference to FIG. In this example, FIG.
A case of unification for an equivalent expression example in a language as shown in (d) will be described.

In step 1010, the equivalence analysis unit 14 determines the expression pattern of the description content extracted in step 915 in FIG. "A = b = n;" type or "a
= N, b = n; ", in step 1020, replace with" a = n; b = n; "
b; type, in step 1030, “a;
± = b; ”or“ a = a ± 1; ”or“ ±
± a; ”, in step 1040,“ a
±± b; ”. Note that this example shows an example in which the above-described replacement rule is provided, and another rule can be adopted.

Thereafter, in step 935 of FIG.
The equivalence analyzer 14 updates the read pointer, returns to step 915, and extracts the next knowledge element. Step 9
In 15, if it is the last knowledge element, the preprocessing for equivalence analysis evaluation ends, and the process proceeds to step 1200.

Here, referring to FIG.
The contents of the equivalence analysis work area that has been pre-processed for equivalence analysis evaluation in step 935 will be described. The equivalence analysis work area 1100 includes a knowledge element name 1110 and a description content 1120. FIG.
The knowledge element name 1110 and the description content 11 shown in FIG.
Reference numeral 20 is the same as the knowledge element name 1110 and the description content 1120 of the extraction work area 500 shown in FIG. 5, but the description content is integrated.

For example, “wtime == 10” in the description content 520 for “condition 10L5” in FIG.
At 0, “wtime == 10”. In other description contents, blank character correction is similarly performed. Note that, in the example shown in FIG.
Although the examples of 25, 930, and 1000 are not shown, these pre-processes are performed according to the content of the description.

In the example shown in FIG. 11A, the description content 1120 of “knowledge == 10” for the knowledge element name 1110 “waiting time is 10 minutes” and the knowledge element name 1110 is “green light” "Singou == blue" in the description 1120 for "a"
"e" indicates an existing element extracted from the knowledge element dictionary 120 shown in FIG. 1 in step 240 shown in FIG. These existing elements are also registered in the equivalence analysis work area 1100, but are executed in the knowledge element equivalence analysis evaluation processing, which will be described later. Note that FIG.
The (b) equivalence evaluation flag 1130 will be described later.

Next, returning to FIG. 9, a description will be given of the knowledge element equivalency analysis evaluation processing of steps 1200 and 940 and thereafter. In step 1200 of FIG. 9, initial settings are made for the work area for equivalence analysis. Here, the initial setting process for the equivalence analysis work area will be described with reference to FIG.

In step 1210, the equivalence analysis unit 14 shown in FIG. 1 makes the existing knowledge element dictionary 120 (FIG. 1) by selecting the input in step 220 shown in FIG.
Is used or not.

If the existing knowledge element dictionary 120 is not used, in step 1220, the equivalence analysis unit 14
The equivalency evaluation flag 1130 corresponding to the number of knowledge elements registered in the equivalence analysis work area 1100 is set to “unevaluated”. Note that the equivalency evaluation flag 1130 shown in FIG. 11B includes “evaluated” and “same element”, but these are set to “unevaluated” in the initial state.

If the existing knowledge element dictionary 120 is in use,
In step 1230, the equivalence analysis unit 14 inserts the knowledge element in the existing knowledge element dictionary before the knowledge element registered in the equivalence analysis work area 1100. The equivalence evaluation flag 1130 is set to “existing element” in accordance with the number of knowledge elements.
Set to. For the extracted extracted knowledge elements, the equivalence evaluation flag 1130 is set to “unevaluated” in step 1220 corresponding to the number of knowledge elements.

That is, as shown in FIG. 11, the knowledge element name 1110 in the existing element dictionary is changed to “the waiting time is 10 minutes”.
Of description 1120, “wtime == 1
“0” and the description element 1120 “singoug == b” for the knowledge element name 1110 “green light”
“lue” is inserted. In addition, the equivalence evaluation flag 1
In “130”, “existing element” is set.

Next, in step 940 of FIG. 9, the equivalence analysis unit 14 positions the read pointer at the head knowledge element of the equivalence analysis work area 1100, and
At 45, the knowledge element at the read pointer position is retrieved.

In step 950, the equivalence analysis unit 1
4 judges the equivalence evaluation flag 1130 corresponding to the extracted knowledge element, and proceeds to step 955 if “evaluated” or “existing element”, and proceeds to step 995 if “evaluated”.

At the time when the initial setting in step 1200 is completed, since it is “unevaluated” or “existing element”, in step 955, the equivalence analysis unit 14 analyzes the knowledge element name 1110 and the description 1120 Registered in the work area 1100.

Next, in step 1400 of FIG.
The equivalence analysis unit 14 uses the same knowledge information 130 shown in FIG.
Performs registration processing for. Here, the registration process for the same knowledge information will be described with reference to FIGS.

First, the internal configuration of the same knowledge information 130 shown in FIG. 1 will be described with reference to FIG. The identical knowledge information 130 includes an integrated knowledge element table 132 and an identical knowledge element table 134. The integrated knowledge element table 132 is a table in which knowledge elements obtained by integrating knowledge elements having the same description content are registered. The knowledge element name 132A, the number 132B of the same knowledge elements corresponding to the knowledge element name 132A, and the same description content Of the same knowledge ring pointer 132C, which is the link information with the knowledge element. Same knowledge element table 13
4 is a knowledge element name 3 of a knowledge element having the same description content
1 is a table storing the knowledge element names 134A. The integrated knowledge element table 132
, The knowledge element of the description content in which the knowledge element of the same description content does not exist is also registered.
2B becomes "0" and the same knowledge link pointer 132C
Becomes "none". The specific contents of the same knowledge information 130 will be described with reference to the flowchart of FIG.

In step 1410 of FIG. 14, the equivalence analysis unit 14 registers the knowledge element name 1110 shown in FIG. 11 at the position of the read pointer in the integrated knowledge element table 132 shown in FIG.

Next, in step 1420, the equivalence analysis unit 14 determines the equivalence evaluation flag 1130 (FIG. 11) corresponding to the knowledge element.
Proceed to step 1430, and if it is not “existing element”, proceed to step 1450.

Here, the equivalence evaluation flag is set to “existing element”
In step 1430, the equivalence analysis unit 14 sets “none” to the same knowledge link pointer 132C of the integrated knowledge element table 132, and furthermore,
In step 1440, “0” is set to the number of identical knowledge elements 132B. The example illustrated in FIG. 13 illustrates a state in which all the processes of the equivalence analysis illustrated in FIG. 9 are completed. For example, “the green signal” of the knowledge element name 132A is used.
With regard to, “none” is set in the same knowledge link pointer 132C, and “0” is set in the same knowledge element number 132B, and the initial state processed in steps 1430 and 1440 is shown.

If the equivalence evaluation flag is other than “existing element”, in step 1450, the equivalence analyzer 14 sets the knowledge element name 132
A is registered in the same knowledge element table 134,
At 460, the pointer position is set to the same knowledge link pointer 132C of the integrated knowledge element table 132, and at step 1470, the same knowledge element number 132 is set.
B is set to “1”. For example, as shown in FIG. 13, with regard to “processing 12L3” of the knowledge element name 132A, “1” is set to the number of identical knowledge elements 132B, and “$ 01104” is set to the identical knowledge link pointer 132C. Here, the same knowledge link pointer 132C
“$ 01104” indicates the position where “process 12L3” is registered in the same knowledge element table 134.

Next, returning to step 960 of FIG. 9, the equivalence analysis unit 14 sets the equivalence evaluation flag 11 shown in FIG.
In 30, “evaluated” is set.

Next, in step 960, the equivalence analysis section 14 positions the search pointer at the knowledge element next to the read pointer position in the equivalence analysis work area 1100, and in step 970, the knowledge element at the search pointer position Take out. That is, steps 960 to 990 described later are processes for searching for the same knowledge element as the knowledge element extracted in step 945. For example, in the example shown in FIG. A knowledge element having the same description content as the description content “wtime == 10” for “time is 10 minutes”, for example, “wtim” of “condition 10L5”
e == 10 ".

In step 975, the equivalence analysis unit 1
4 determines the equivalency evaluation flag 1130 corresponding to the extracted knowledge element.
At 80, the contents of the knowledge element at the read pointer position and the search pointer position are compared.

If they are the same, in step 1500, the equivalence analyzer 14 registers the knowledge element at the search pointer position in the same knowledge information 130 corresponding to the knowledge element at the step read pointer position.

Here, referring to FIG.
The same knowledge element registration processing at 0 will be described. In step 1510, the equivalence analysis unit 14 registers the knowledge element name 1110 of the search pointer position in the same knowledge element table 134, and sets the pointer position in the same knowledge link pointer 132C of the integrated knowledge element table 132 in step 1520. Then step 153
At 0, the number of identical knowledge elements 132C is updated by "+1" (step 153). For example, in the example shown in FIG. 11, the description content “wttime == 10” for “waiting time 10 minutes” at the beginning of the knowledge element name 1110
As a knowledge element having the same description content as “Condition 10L”
When "wtime == 10" of "5" is found,
“Condition 10L5” is registered in the knowledge element name 134A of the same knowledge element table 134 shown in FIG.
In C, the knowledge element name 1 in which “condition 10L5” is registered
The address “$ 01000” of 34A is set as the pointer position. Further, since the number of identical knowledge elements 132C is “0” in the initial state as described in step 1440 of FIG. 14, this is updated to “+1” and becomes “1”. In the example shown in FIG. 13, three identical knowledge elements have been found, and the identical knowledge element number 132C has been updated to "3".

Next, returning to step 985 in FIG. 9, the equivalence analysis unit 14 sets the equivalence evaluation flag 1130 corresponding to the knowledge element of the search pointer position in the equivalence analysis work area 1100 shown in FIG. In step 990, the search pointer is updated, and the process returns to step 970 to fetch the next knowledge element.

Also, in the discrimination processing in step 980, when the description contents of both knowledge elements are different, in the discrimination processing in step 975, the equivalence evaluation flag 1130 corresponding to the knowledge element at the search pointer position is set to “identical element”. If it is “existing element”, the search pointer is similarly updated in step 990, and
To retrieve the next knowledge element. If there is no next knowledge element in the discrimination processing in step 970, step 9
At 95, the equivalence analyzer 14 updates the read pointer and returns to step 945 to retrieve the next knowledge element.

In step 950, the equivalence evaluation flag 11 corresponding to the knowledge element at the position of the read pointer is set.
If 30 is the "same element", the read pointer is similarly updated, and the process returns to step 950 to fetch the next knowledge element.

By repeating the above processing, FIG.
The same elements are searched for in order from the top in the description contents 1120 shown in FIG. 12, and the result is registered and set in the integrated knowledge table 132 and the same knowledge element table of the same knowledge information 130 shown in FIG.

In step 945, if there is no next knowledge element, in step 997, among the knowledge elements in the equivalence analysis work area 1100, the knowledge element whose equivalence evaluation flag 1130 is “identical element” is removed. Then, the equivalence analysis evaluation process of the knowledge elements is ended. Note that FIG.
Shows the state before the processing of step 997 is executed. When step 997 is executed, among the knowledge elements in the equivalence analysis work area 1100, the equivalence evaluation flag 1130 is set to “identical element”. Is a knowledge element (for example,
If the knowledge element name 1110 is "condition 10L5" or "processing 14
L34 ”).

Next, returning to FIG. 2, in step 1700, the similarity analysis unit 15 of the processing device 10 shown in FIG. 1 is activated, and the similarity analysis evaluation and rearrangement of the knowledge elements are performed. Here, an example in which the description content of the knowledge element can be evaluated as similar will be described with reference to FIG. FIG. 16 is a specific example in which the following various cases are classified according to the similarity of the left side, right side, and operator constituting the description content of each knowledge element extracted from the procedure / rule type expression knowledge information 110. Is shown. Note that, also in this example, a description will be given using a C language expression as an example.

FIG. 16 (a) shows a combination of knowledge elements having description contents having similarities that the variables and operators used in the description contents are the same and only the value of the constant term included in the right side is different. An example is shown. What is shown in this example is referred to as similarity level 1 in the following description. The similarity level 1 further includes similarity degrees 1.0, 1.1, 1.2, and 1.3.
This will be described later.

FIG. 16 (b) shows an example of a combination of knowledge elements having description contents having similarities that variables and array elements used in the description contents are the same and operators are different. What is shown in this example is referred to as similarity level 2 in the following description. The similarity level 2 further includes a similarity score of 2.0, 2.1, 2.2, 2.3, 2.4, 2,.
Analysis will be made as 5, 2.6, which will be described later.

FIG. 16C shows an example of a combination of knowledge elements having description contents having similarities that the operator is the same as the left side in the description contents and only the right side contents are different. What is shown in this example is referred to as similarity level 3 in the following description. The similarity level 3 is analyzed as a similarity degree of 3.0, which will be described later.

FIG. 16D shows an example of a combination of knowledge elements having description contents having similarities such that only the left side of the description contents is the same and the contents of the operator and the right side are different. In the following description, it is assumed that the similarity level is 4. The similarity level 4 is analyzed as the similarity degree 4.0, which will be described later.

In the present embodiment, similarity levels (1 to 4) indicating the degree of similarity are set, and the knowledge elements in the knowledge element dictionary 120 are rearranged according to the similarity level. It is also possible to adopt an evaluation method other than the above for the order of similar patterns and similarity degrees.

Next, the similarity analysis evaluation and rearrangement processing of the knowledge elements executed by the similarity analysis unit 15 of the processing device 10 will be described with reference to FIG.

In step 1800, the similarity analysis unit 15 performs an initial setting for the similarity analysis work area. Here, the contents of the initial setting will be described with reference to FIG.

In step 1810, the similarity analysis unit 15 uses the contents of the work area for equivalence analysis 1100 created in the equivalence evaluation process (FIG. 2, step 900) in the similarity analysis process used in the knowledge element similarity evaluation process. Transfer to the work area for analysis. Here, the configuration of the work area for similarity analysis will be described with reference to FIG. The similarity analysis work area 1900 includes a knowledge element name 1910 and a description content 1
920. Further, a similarity evaluation flag 1930, a basic knowledge element number 1940, a dictionary creation work area 1950, a similarity management table 1960
It has. The dictionary creation work area 1950 includes a knowledge element name 1952 and description contents 1954. The similarity management table 1960 includes a basic knowledge element number 1962 and a similarity 1964. The details of each unit will be described later.

Since the knowledge element name 1910 and the description content 1920 have the same data structure as the knowledge element name 1110 and the description content 1120 shown in FIG. 11, the equivalence analysis work is performed instead of transferring the content. The name of the area 1100 may be simply changed to the similarity analysis work area 1900.

Next, in step 1820 of FIG. 18, the similarity analysis unit 15
00, the similarity evaluation flag 1930 is set to “not yet evaluated” corresponding to the number of knowledge elements registered in
Clear the basic knowledge element number 1940. Further, in step 1830, the dictionary creation work area 1950
Is cleared, and in step 1840, the similarity management table 1960 is cleared.

Next, returning to step 1702 of FIG.
The similarity analysis unit 15 sets a head similarity level to be evaluated and a head basic knowledge element number.

Next, in step 1704, the similarity analysis unit 15 positions the read pointer at the head knowledge element of the similarity analysis work area 1900, and
At 6, the knowledge element at the read pointer position is extracted.

In step 1708, the similarity analysis unit 15 determines the similarity evaluation flag 1930 corresponding to the extracted knowledge element. If the evaluation flag is “unevaluated” or “basic element”, the process proceeds to step 1710, where If “completed”, the process proceeds to step 1736.

The case where the similarity evaluation flag 1930 is “unevaluated” or “basic element” will be described below. In step 1710, the similarity analysis unit 15 positions the search pointer at the knowledge element next to the read pointer position in the similarity analysis work area 1900. In step 1712, the similarity analysis unit 15 extracts the knowledge element at the search pointer position. Through the processing of to 1736, a knowledge element similar to the knowledge element extracted in step 1706 is extracted, and a similarity is set. In step 1736, in order to update the search pointer, similarity analysis is sequentially performed on all knowledge elements registered in the similarity analysis work area 1900. For example, in the example shown in FIG. 19, the knowledge element name 191 which is the first knowledge element
As a knowledge element similar to “wttime == 10” where 0 is the description content 1920 for “waiting time of 10 minutes”, “wttime == 20” where the knowledge element name 1910 is the description content 1920 for “condition 11L5”
Or “wtime == dtim” in which the knowledge element name 1910 is the description content 1920 for “condition 14L7”.
e + 10 ”is extracted, and“ wtim ”of these knowledge elements is extracted.
e == 10 "is analyzed.

In step 1710, the similarity analysis unit 15 determines the similarity evaluation flag 1930 corresponding to the extracted knowledge element.
Proceed to step 1716, where "evaluated" or "basic element"
If so, the process proceeds to step 1736 to update the search pointer.

If the evaluation is not completed, step 1716
In step (2), the similarity analysis unit 15 determines the description contents of the knowledge elements at the read pointer position and the search pointer position according to the similarity level in steps 2000, 2100, and 22.
At 00 and 2300, a similarity analysis evaluation is performed. Since the similarity levels of the knowledge elements are sequentially set in step 1740 from similarity level 1 to similarity level 4, steps 2000, 2100, 2200, and 2300 are sequentially selected. In step 2000, it is determined whether or not the similarity level 1 is satisfied, and if the similarity level 1 is satisfied, the degree of similarity is analyzed. Step 210
Similarly, 0, 2200, and 2300 analyze whether or not they correspond to the respective similarity levels, and if they correspond to the similarity levels, the degree of similarity is analyzed.

First, referring to FIG.
The similarity analysis processing of level 1 at 0 will be described. In the similarity analysis process of level 1 in the present embodiment, when the similarity determination rule as shown in FIG.

In step 2010, the similarity analysis unit 15 compares the left sides of the description contents of the knowledge element at the knowledge element pointer position and the search element at the search pointer position, and if they are the same, proceeds to step 2020;
Proceed to 80 and return as "No similarity". For example, in the example shown in FIG. 19, the knowledge element name 1910 at the knowledge element pointer position has a "waiting time of 10 minutes" and the description content 1920
Is the knowledge element of “wtime == 10”, the knowledge element name 1910 at the search pointer position is “green light”, and the description content 1920 is “singou == b”.
If the knowledge element is “le”, the left side of the description content is different, and the result is “no similarity”.

Next, in step 2020, the similarity analysis unit 15 compares the operators. If the operators are the same, the process proceeds to step 2030. If the operators are different, the process proceeds to step 2080 and returns "no similarity". For example, in the example shown in FIG. 19, the knowledge element name 1910 at the knowledge element pointer position
Is "Waiting time is 10 minutes" and the description 1920 is "wti
Assuming that the knowledge element of "me == 10", the knowledge element name 1910 at the search pointer position is "condition 13L6", and the description content 1920 is a knowledge element of "wtime >> = 30", the operator of the description content is Because they are different
"No similarity".

Further, in step 2030, the similarity analysis unit 15 examines the contents on the right side, and proceeds to step 204.
At 0, if both right sides are only constants, 1.0 is returned as the similarity, and at step 2050, both right sides are a pattern of “a ± n”, “a” is the same variable for both, and “n” Is a constant or not, if the similarity is “1.
Returns "1". In step 2060, if both right sides are patterns of “a [m] ± n” and the same array variable “a [m]” has the same subscript m and “n” is a constant or not, it is similar. Return “1.2” as the degree and step 20
70, both right sides are patterns of “a [m] ± n”, and the same array variable “a [m]” has a different subscript m,
If “n” is a constant or not, the similarity is “1.3”
In step 2080, if the right side is another combination or the left side or the operator is different, “No similarity” is returned.

For example, in the example shown in FIG. 19, the knowledge element name 1910 at the knowledge element pointer
Minutes, the description 1920 is “wtime == 1
0, and the knowledge element name 1910 at the search pointer position is “condition 11L5” and the description content 1920
Is a knowledge element of “wtime == 20”, “similarity 1.0” is obtained. In addition, the knowledge element name 1910 at the search pointer position is “condition 14L7” and the description content 1920 is “wttime == dtime + 1”.
If it is a knowledge element of "0", it will be "no similarity".

In the example shown in FIG. 19, the knowledge element name 1910 at the knowledge element pointer position is changed to "Process 10L18".
The description content 1920 is the knowledge element of “x == y + 10”, and the knowledge element name 191 at the search pointer position is described.
0 is “Process 12L3” and the description 1920 is “x =
= Y + 20 "," similarity 1.1 "is obtained.

This example shows an example in which the above-described similarity determination rule is provided, and another rule can be adopted. Also, a case is shown in which the similarity is set by further subdividing the similarity level, but another setting method is also possible. These can be similarly applied to the following levels of processing.

Next, referring to FIG. 21, step 2100 will be described.
Will be described below.
In the similarity analysis process of level 2 in the present embodiment, when the similarity determination rule as shown in FIG.

In step 2105, the similarity analysis unit 15 compares the left sides of the description contents of the knowledge element at the knowledge element pointer position and the search element at the search pointer position, and if they are the same, proceeds to step 2110;
Proceed to 65 and return as "No similarity". For example, in the example shown in FIG. 19, the knowledge element name 1910 at the knowledge element pointer position has a "waiting time of 10 minutes" and the description content 1920
Is the knowledge element of “wtime == 10”, the knowledge element name 1910 at the search pointer position is “green light”, and the description content 1920 is “singou == b”.
If the knowledge element is “le”, the left side of the description content is different, and the result is “no similarity”.

Next, in step 2110, the similarity analysis unit 15 determines the type of the operator. If the operator is a comparison operator, the process proceeds to step 2115; otherwise, the process proceeds to step 2165. , "No similarity".

Further, in step 2030, the similarity analysis unit 15 examines the contents on the right side, and proceeds to step 213.
At 0, if both right-hand sides are only constants, "2.0" is returned as the similarity. For example, in the example shown in FIG.
The knowledge element name 1910 at the knowledge element pointer position is “waiting time is 10 minutes” and the description content 1920 is “wtime ==
10 ", the knowledge element name 1910 at the search pointer position is" condition 13L6 ", and the description content 19
Assuming that 20 is a knowledge element of “wtime> = 30”, “similarity 2.0” is obtained.

In step 2135, if both right sides have only the same variable, “2.1” is returned as the similarity, and in step 2140, both right sides are set to “a ±
In the pattern "n", "a" is the same variable for both, and "n"
If is a constant or not, "2.2" is returned as the similarity.

If both right-hand sides have only the same array element “a [m]”, the subscript m is compared in step 2120, and if they are the same, “2.3” is returned as the similarity in step 2145. If not, step 2150
Returns “2.5” as the similarity.

Further, both right sides are patterns of “a [m] ± n”, and “n” is a constant or not, and the same array variable “a”
In the case of [m] ", the subscript m is compared in step 2125, and if they are the same," 2.4 "is returned as the similarity in step 2155; 2.6 is returned.

Otherwise, in step 2165, “No similarity” is returned.

Next, referring to FIG.
Will be described below.
In the similarity analysis process of level 3 in the present embodiment, when the similarity determination rule as shown in FIG.

In step 2210, the similarity analysis unit 15 compares the left sides of the description contents of the knowledge element at the knowledge element pointer position and the search element at the search pointer position, and if they are the same, proceeds to step 2220;
Proceed to 40 and return as "No similarity".

Next, in step 2220, the similarity analysis unit 15 compares the operators. If the operators are the same, the process proceeds to step 2230, and if they are different, the process proceeds to step 2240 and returns “no similarity”.

If the operators are the same, step 2230
, The similarity analysis unit 15 determines the similarity as “3.
Returns "0". For example, in the example shown in FIG. 19, the knowledge element name 1910 at the knowledge element pointer
Minutes, the description 1920 is “wtime == 1
0, and the knowledge element name 1910 at the search pointer position is “condition 14L7” and the description content 1920
Is a knowledge element of “wttime == dtime + 10”, “similarity 3.0” is obtained.

Next, referring to FIG.
, The level 4 similarity analysis process will be described.
In the similarity analysis processing of level 4 in the present embodiment, when the similarity determination rule as shown in FIG.

At step 2310, the similarity analysis unit 15 compares the left sides of the description contents of the knowledge element at the knowledge element pointer position and the search pointer position, and proceeds to step 2320 if they are the same, and proceeds to step 2320 if they are different.
Proceed to 40 and return as "No similarity".

Next, in step 2320, the similarity analysis unit 15 compares the operators. If the operators are the same, the process proceeds to step 2330. If the operators are different, the process proceeds to step 2340 to return “no similarity”. If the operators are the same, in step 2330, the similarity analysis unit 15
“4.0” is returned as the similarity.

Next, returning to step 1718 in FIG.
The similarity analysis unit 15 determines in steps 2000, 2100, and 2
The analysis and evaluation results at 200 and 2300 are determined to be “not similar” or “similar”, and if “similar” (similarity level 1.0 to 4.0) ”, the process proceeds to step 1720 and“ no similar ” If so, the flow advances to step 1736 to update the search pointer and perform a similar similarity analysis on the next knowledge element.

If the result of the analysis evaluation is “similar”, in step 1720, the similarity analysis unit 15 sets the similarity evaluation flag 19 corresponding to the knowledge element at the read pointer position.
30 is determined, and if “evaluation is not completed”, step 172 is executed.
Go to step 2 if it is "basic element", go to step 1728.

If the evaluation has not been completed, step 1722
19, the similarity analysis unit 15 describes the knowledge element name 1910 and the description content 1920 of the knowledge element at the read pointer position shown in FIG. 19C as the knowledge element name 1952 of the knowledge element in the dictionary creation work area 1950. The content is registered in 1954.

In step 1724, the similarity analysis unit 15 converts the current basic knowledge element number and the analysis evaluation into the basic knowledge element number 1962 and similarity 1964 of the position in the similarity management table 1960 corresponding to the position. The resulting similarity is set, and the next basic knowledge element number is updated. In step 1726, the similarity evaluation flag 1930 corresponding to the knowledge element at the position of the read pointer in the similarity analysis work area 1900 is set to " Basic elements "
Set the basic knowledge element number.

In the example shown in FIG. 19, for example, the description content “wti
me == 10 ”, the similarity management table 19
In the basic knowledge element number 1962 of the position in 60, for example,
“1” is set, “1.0” is set in the similarity 1964, and the similarity evaluation flag 1930 corresponding to the knowledge element at the read pointer position in the similarity analysis work area 1900 is set to “basic element”. Similarly, the knowledge element name 191
Description content “x == y + 10” of “Process 10L8” of 0
For example, “10” is set to the basic knowledge element number 1962 of the position in the similarity management table 1960, “1.1” is set to the similarity 1964, and the similarity analysis work area 1900 The similarity evaluation flag 1930 corresponding to the knowledge element at the position of the read pointer is set to “basic element”.

In step 1720, if the similarity evaluation flag is “basic element”, step 1728
, The basic knowledge element number 1962 is extracted, and the flow advances to step 1730.

Next, in step 1730, the similarity analysis unit 15 replaces the knowledge element name 1910 and the description content 1920 of the knowledge element at the search pointer position with the knowledge element name 1952 and the description content 19 of the dictionary creation work area 1950.
In step 1732, the current basic knowledge element number and the similarity obtained as an analysis evaluation result are stored in the basic knowledge element number 1962 and the similarity 1964 of the position in the similarity management table 1960 corresponding to the position. In step 1734, the similarity evaluation flag 1930 corresponding to the knowledge element at the search pointer position in the similarity analysis work area 1900 is set to “evaluated”.

Thereafter, in step 1736, the search pointer is updated, and the process returns to step 1712 to fetch the next knowledge element and repeat the above-described analysis evaluation processing.

Further, when there is no next knowledge element in step 1706, the similarity level is set to the next similarity level in step 1740, and the above processing is repeated up to similarity level 4.

At step 1740, if there is no similar level to be processed next, at step 1742,
The similarity analysis unit 15 rearranges the order of the knowledge elements in the dictionary creation work area 1950 using the basic knowledge element number 1962 and the similarity 1964 as keys. The contents of the dictionary creation work area 1950 and the similarity management table 1960 shown in FIG. 19 show the state before the rearrangement is performed. When sorting is complete,
The basic knowledge element number 1962 is "1", ..., "10",
Sorted in ascending order of “11”, and within the same basic knowledge element number, the similarity 1964 is “1.0”, “1.
, "2.0", ..., "4.0" in ascending order.

Finally, in step 1744, the similarity analysis unit 15 stores the knowledge element whose similarity evaluation flag 1930 in the similarity analysis work area 1900 is “unevaluated” in the dictionary creation work area 1950. A certain knowledge element is added after the last knowledge element, and the similarity analysis evaluation processing of the knowledge element ends.

Returning to step 250 in FIG. 2, the similarity analysis unit 15 stores the knowledge element dictionary 120 obtained as a result of the similarity analysis evaluation and rearrangement of the knowledge elements in step 1700 in the storage device 100. Store.

Next, at step 260, the display unit 18 of FIG.
Is edited into a table format as shown in FIG.
Display on 0.

Here, the knowledge element dictionary 120 displayed on the display device 40 will be described with reference to FIG. The knowledge element dictionary 120 displayed on the display device 40 includes a knowledge element name 122 and description contents 124. Here, as understood from comparison with the examples shown in FIGS. 19C and 19D, in the display example shown in FIG. 24, the basic knowledge element numbers 1962 are assigned to the knowledge elements in the dictionary creation work area 1950. And the similarity 1964 as a key, the order of the knowledge elements is rearranged. That is, the description name 1 for the knowledge element name 122 corresponding to “the waiting time is 10 minutes”
24, “wtime == 10”, as shown in FIG. 19D, since the basic knowledge element number 1962 is “1”, the basic knowledge element number 1962 has the same “1” below it. Belonging to “wtime == 20”, “wti
me> = 30 ”,“ wtime == dtime +
10 "is displayed. Moreover, these display orders are in the order of similarity.

Next, at step 270, the display unit 3
From 0, the display of the next processing option is displayed on the display device 40,
The command receiving unit 30 is in a state of waiting for input. Input device 3
When the designation of “return” is input from 0, the process returns to step 220.

[0158] Here, in step 270, when the designation of "dictionary editing" is input, in step 2500, the knowledge element editing unit 16 is started. Further, when an instruction of “dictionary editing” is input as a processing option in step 220, in step 280, the display unit 30
Requests input of a knowledge element dictionary name from the display device 40,
When the knowledge element dictionary name is input from the input device 30, in step 290, the knowledge element dictionary 120 is stored in the storage device 1
00, and in step 2500, the knowledge element editing unit 16 is activated.

Next, the knowledge element dictionary editing process performed by the knowledge element editing unit 16 will be described with reference to FIGS. 25 and 26. In step 2510, the knowledge element editing unit 16
26, the knowledge element name input screen 2 is displayed on the display screen 2600 of the display device 40 via the display unit 18 as shown in FIG.
610 is displayed. The knowledge element name input screen 2610 is
The knowledge element dictionary 120 is edited and displayed in a table format.

In Step 2520, the input device 30
, A new knowledge element name 2614 is input for each of the original knowledge element names 2612. In the example shown in FIG. 26, in the initial state, the new knowledge element name 2614 is blank. Original knowledge element name 2 with "10 minutes waiting time"
612 is extracted from the existing dictionary and has already been edited, so that the new knowledge element name 2614 also includes
"Waiting time is 10 minutes" is input. The original knowledge element name 2612 of “condition 11L5” is the description content 26 on the right side thereof.
By looking at 16, the content of the knowledge element can be confirmed.
In addition, as described in FIG. 24, since the images are rearranged in a similar order, the description content
By comparing “me == 10” and “wtime == 20”, the similarity between the two can be easily grasped. Therefore, the new knowledge element name 2614
It is possible to easily input “20 minutes of waiting time” by referring to “10 minutes of waiting time”.

When the input operation is completed in step 2530, the knowledge element dictionary 120 obtained as a result of the input is stored in the storage device 100 in step 2540. If "stop" of the input process is input in step 2530, the process ends without storing.

Further, if the "description location search" of the knowledge element is designated during the input operation of step 2520, search processing of the description location is performed from step 2550. That is, in step 2550, the knowledge element editing unit 16 searches the same knowledge information 130 for the knowledge element where the cursor is located on the knowledge element name input screen 2610, and describes the knowledge element based on the knowledge element name in the same knowledge information 130. The position is calculated, and in step 2560, the description location in the knowledge description of the procedure / rule type expression knowledge information 110 is displayed on the description location display screen 2620. For example, on the knowledge element name input screen 2610, “wti
When the cursor 2630 exists at “me == 10”,
A description location display 2622 in the knowledge description of the procedure / rule type expression knowledge information 110 is performed on the description location display screen 2620. This makes it possible to easily input a correct new knowledge element name by confirming the relationship before and after the specified description content.

At step 2570, input of the next search is awaited, and if "search continuation" is instructed, the next description position for the same knowledge element is calculated and displayed as described above. When "end search" is instructed, the process returns to step 2510 and returns to the display state of the knowledge element name input screen 2610.

According to the present embodiment, in a knowledge information analysis / evaluation system used for problem solving and planning, a knowledge element constituting the description content is extracted from knowledge information expressed by procedures and rules, and Unify expressions for parts where expressions are repeatedly described, or for parts that have originally the same meaning but different expressions, and integrate knowledge elements. Since the knowledge elements can be classified and arranged according to the degree, it is possible to avoid contradictions between the knowledge elements and confusion and misunderstanding when understanding the knowledge by different expressions for the same content,
Since the contents of the knowledge description can be easily understood and errors such as grasping of a relevant part at the time of correction and improvement of the knowledge and omission of change can be prevented, efficiency can be improved and reliability of the knowledge information is improved.

Also, the name setting for each piece of knowledge for facilitating the understanding of the knowledge information can be performed while comparing the similarities and differences between the knowledge elements based on the order in which the knowledge elements are arranged in the knowledge element dictionary. And a name expression suitable for the semantic content of each knowledge element.

[0166]

According to the present invention, a knowledge description is analyzed and evaluated in consideration of the equivalence and similarity of its semantic contents, and it is easy to set a knowledge expression that clearly expresses the semantic contents of the knowledge. It becomes.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an overall configuration of a knowledge information processing system using a method for analyzing knowledge information according to an embodiment of the present invention.

FIG. 2 is a flowchart showing an overall processing flow of a knowledge information analysis method according to an embodiment of the present invention.

FIG. 3 is an explanatory diagram of an example of procedure / rule type expression knowledge information to be analyzed and evaluated in the knowledge information analysis method according to an embodiment of the present invention.

FIG. 4 is a flowchart showing a flow of a knowledge element extraction process executed by a knowledge element extraction unit in the method for analyzing knowledge information according to one embodiment of the present invention.

FIG. 5 is an explanatory diagram of an extraction work area used for a knowledge element registration process executed by a knowledge element extraction unit in the knowledge information analysis method according to one embodiment of the present invention.

FIG. 6 is a flowchart showing a flow of a knowledge element registration process executed by a knowledge element extraction unit in the knowledge information analysis method according to one embodiment of the present invention.

FIG. 7 is a flowchart showing a processing flow of a sub-step of FIG. 6;

FIG. 8 is an explanatory diagram of an example that can be evaluated to be semantically equivalent based on the equivalence analysis performed by the equivalence analysis unit in the method for analyzing knowledge information according to the embodiment of the present invention.

FIG. 9 is a flowchart showing a flow of an equivalence analysis evaluation and integration process of knowledge elements in an equivalence analyzer in the method for analyzing knowledge information according to an embodiment of the present invention.

FIG. 10 is a flowchart showing a flow of a process of unifying descriptions of language equivalent expressions in the equivalence analysis unit in the knowledge information analysis method according to the embodiment of the present invention.

FIG. 11 is an explanatory diagram of contents of a pre-processed equivalence analysis work area in the equivalence analyzer in the knowledge information analysis method according to the embodiment of the present invention.

FIG. 12 is a flowchart showing a flow of an initial setting process in an equivalence analysis unit in the method for analyzing knowledge information according to an embodiment of the present invention.

FIG. 13 is an explanatory diagram of the same knowledge information stored in the storage device in the knowledge information analysis method according to one embodiment of the present invention.

FIG. 14 is a flowchart showing a flow of registration processing for the same knowledge information in the equivalence analysis unit in the knowledge information analysis method according to the embodiment of the present invention.

FIG. 15 is a flowchart showing a flow of registration processing for the same knowledge information in the equivalence analysis unit in the knowledge information analysis method according to the embodiment of the present invention.

FIG. 16 is an explanatory diagram of an example in which similarity can be evaluated as a basis of a similarity analysis performed by a similarity analysis unit in the knowledge information analysis method according to an embodiment of the present invention.

FIG. 17 is a flowchart illustrating a flow of a similarity analysis evaluation and a rearrangement process performed by the similarity analysis unit in the method for analyzing knowledge information according to the embodiment of the present invention;

FIG. 18 is a flowchart showing a flow of a similarity analysis work area initial setting process executed by a similarity analysis unit in the knowledge information analysis method according to one embodiment of the present invention.

FIG. 19 is an explanatory diagram showing a configuration of a work area for similarity analysis used by a similarity analysis unit in the method for analyzing knowledge information according to one embodiment of the present invention.

FIG. 20 is a flowchart showing a flow of a level 1 similarity analysis process of the similarity analysis unit in the knowledge information analysis method according to one embodiment of the present invention.

FIG. 21 is a flowchart illustrating a flow of a level-2 similarity analysis process performed by the similarity analyzer in the method for analyzing knowledge information according to the embodiment of the present invention;

FIG. 22 is a flowchart illustrating a flow of a level-3 similarity analysis process performed by the similarity analyzer in the method for analyzing knowledge information according to the embodiment of the present invention;

FIG. 23 is a flowchart illustrating a flow of a level 4 similarity analysis process of the similarity analysis unit in the knowledge information analysis method according to the embodiment of the present invention;

FIG. 24 is an explanatory diagram of a display example of a knowledge element dictionary displayed on the display device in the method for analyzing knowledge information according to one embodiment of the present invention.

FIG. 25 is a flowchart showing the flow of a knowledge element dictionary editing process by the knowledge element editing unit 16 in the knowledge information analysis method according to one embodiment of the present invention.

FIG. 26 is an explanatory diagram of a display example on a display device in a method for analyzing knowledge information according to an embodiment of the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 10 ... Processing apparatus 11 ... Command reception part 13 ... Knowledge element extraction part 14 ... Equivalence analysis part 15 ... Similarity analysis part 16 ... Knowledge element editing part 18 ... Display part 20 ... Medium drive 22 ... Medium 30 ... Input device 40 ... Display device 100: Storage device 110: Procedure / rule type expression knowledge information 120: Knowledge element dictionary 130: Same knowledge information

Claims (10)

[Claims] 1. A knowledge information analysis method for analyzing knowledge information composed of knowledge elements described in a rule-type expression or a procedural expression, wherein the knowledge elements constituting the knowledge information are extracted and extracted. The knowledge elements are arranged and integrated based on the analysis and evaluation of the equivalence and similarity to the described contents of the knowledge elements, and a knowledge element dictionary in which the arranged and integrated knowledge elements are arranged is generated. How to analyze knowledge information. 2. The knowledge information analysis method according to claim 1, wherein the integration of the knowledge elements is performed by unifying expressions of knowledge elements having different expression methods but having the same semantic contents. Analysis method of knowledge information. 3. The method for analyzing knowledge information according to claim 1, wherein the organization of the knowledge elements is performed by organizing the knowledge elements based on the similarity of the expression method. Analysis method. 4. A method for analyzing knowledge information according to claim 1, wherein a plurality of knowledge elements having a connection of semantic contents and a unity are treated as one knowledge element, and a knowledge element dictionary in which said knowledge elements are arranged is generated. A method for analyzing knowledge information characterized by: 5. The method for analyzing knowledge information according to claim 1, wherein the extraction of the knowledge element is performed while referring to the already created knowledge element dictionary, and a new knowledge element is added to the already created knowledge element dictionary. A knowledge information analysis method characterized by additionally generating a knowledge element dictionary. 6. The knowledge information analysis method according to claim 1, wherein a name of a knowledge element constituting said generated knowledge element dictionary is changed to generate a new knowledge element dictionary. Analysis method. 7. A knowledge information processing system having a processing device for analyzing and processing knowledge information constituted by knowledge elements described in a rule type expression or a procedural expression, wherein the processing device converts the knowledge information into the knowledge element A knowledge element extraction unit for extracting the knowledge elements, and an equality analysis for generating a knowledge element dictionary by integrating and arranging the knowledge elements based on the equivalence to the description contents of the knowledge elements extracted by the knowledge element extraction unit And a similarity analysis unit that arranges and arranges the knowledge elements based on the similarity to the description contents of the knowledge elements extracted by the knowledge element extraction unit and generates a knowledge element dictionary. A knowledge information processing system characterized by the following. 8. The knowledge information processing system according to claim 7, wherein said equivalence analysis unit registers a knowledge element obtained by integrating knowledge elements having the same description content in an integrated knowledge element table, and outputs the same description content. A knowledge element name of a knowledge element having the same knowledge element table is stored in the same knowledge element table, and the same knowledge element table is linked to the integrated element table so that the two are associated with each other. 9. The knowledge information processing system according to claim 7, wherein the processing device further changes a name of a knowledge element constituting the generated knowledge element dictionary and edits the knowledge element into a new knowledge element dictionary. A knowledge information processing system comprising an element editing unit. 10. A knowledge element extraction unit for extracting a knowledge element from knowledge information described in a rule type expression or a procedural expression, and a knowledge element extracted by the knowledge element extraction unit based on equivalence to the description content of the knowledge element. An equivalence analysis unit that integrates and arranges the knowledge elements to generate a knowledge element dictionary; and stores the knowledge elements on the basis of the similarity to the description content of the knowledge elements extracted by the knowledge element extraction unit. A storage medium, comprising: a similarity analysis unit that arranges and arranges to generate a knowledge element dictionary.