JPH10143372A - Method for analyzing and evaluating knowledge information - Google Patents

Abstract

PROBLEM TO BE SOLVED: To efficiently detect error and waste of knowledge information by analyzing and evaluating conditional description constituting the knowledge information. SOLUTION: Each conditional description constituting the knowledge information 30 contains a conditional expression in which a condition to connect a condition name and a condition value together by an equality sign, a non-equality sign or an inequality sign is further connected by an AND sign or an OR sign. An knowledge analyzing part 3 generates plural equivalent conditional descriptions by dividing the conditional description when the conditional expression contains an OR expression, generates plural conditional descriptions by dividing the conditional description when the conditional expression contains the non-equality sign expression or the inequality sign expression like converting into an equivalent equality sign expression and issues an analyzing result management list 64 by arranging the conditional descriptions based on a pattern of combination of the order of the condition name and the condition value taken by the condition, An knowledge evaluating part 6 detects relations of duplication, discrepancy and inclusion based on the analyzing result management list 64.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to knowledge information processing using an information processing apparatus, and more particularly to a technique for analyzing and evaluating knowledge information.

[0002]

2. Description of the Related Art Knowledge information processing technology or artificial intelligence technology is used for advanced scheduling planning, such as creation of train schedules in transportation, creation of crew schedules, and creation of elevator maintenance plans. Conventionally, U.S. Pat. No. 4,648,044 "Basic Explorer"
As described in “t System Tool”, it is known that knowledge is described by rules and inference is performed using the knowledge. Furthermore, in addition to the description of the rules, the objectives for using these rules and their achievement policies and means are described as strategies, and the rules are executed based on the strategies selected in accordance with the set objectives, and inferences are made. As an information processing system to be advanced, there is Japanese Patent Application Laid-Open No. 1-2838, "Information Processing System". According to this technology, it is possible to separate the purpose and strategy for executing problem solving and planning from the rules necessary for execution, and to describe them in association with each other. A plurality of strategies for a goal can be set in a priority order using a goal / strategy net in which relationships are connected by AND or OR, and a plurality of sub-goals can be set for each strategy. It is possible to describe while dividing and organizing the problem, and it is also possible to change the strategy according to the situation and create a well-balanced plan that cooperates between the lower objectives. But with this technology,
Of the knowledge required for inference, knowledge of upper level knowledge can be expressed by composition of purpose and strategy, but for lower knowledge such as rules required for execution, conditions due to differences in situations and conditions necessary for inference execution The number of combinations becomes huge,
Expression by the composition of purpose and strategy becomes practically difficult in a real system.

[0003]

Among the above-mentioned techniques, USP 4,648,044 relating to the description of knowledge based on rules, includes a description of a purpose for executing inference and an execution means, and a description of another method when a problem occurs in inference execution. Since all are described as rules, they are mixed and it is difficult to understand and confirm the description, and problems are likely to occur from the viewpoint of reliability. Further, when dealing with an object having a large number of combinations and complicated processing, such as creation of a train diagram of a transportation agency or a crew diagram, the number of rules to be described is enormous, and there is a problem that the problem easily develops into a problem of processing capability. As a means for solving this, the technique of the above-mentioned Japanese Patent Application Laid-Open No. 1-2838, "Information Processing System" has been devised. According to this technology, the purpose and strategy for executing problem solving and planning can be described as high-level knowledge and the detailed rules required for execution can be separately described as low-level knowledge, and difficult problems can be divided into simple problems. Not only can it be described while being organized, but also the understanding of the described knowledge and the efficiency of changes can be improved.

[0004] However, in the above-mentioned prior art, an expression in which a knowledge description is made at the time of reference or editing of knowledge is treated as an object in its original form, and any structural and content analysis / evaluation is required. It had to be treated without being done. For this reason, there is a problem that the ease of understanding of knowledge and the possibility of finding errors and waste such as inconsistencies and omissions depend on the skill at the time of knowledge description, such as the order in which knowledge is described and how to classify the description. Was. Moreover, even if the knowledge was described in an orderly manner at first, each time a change such as addition / deletion of the knowledge was repeated based on the inference result, the order was lost due to the time margin at the time of the change or the transition of the person in charge, It also includes the problem that it is difficult to understand knowledge and find errors and waste.

For this reason, from the viewpoints of facilitating understanding of knowledge and finding errors and waste more efficiently, knowledge closely related to the description content expressing the knowledge is rearranged or specified from various viewpoints. It was necessary to process, arrange, and evaluate the description contents in various ways by extracting only the knowledge that meets the conditions described above and decomposing the complicated description contents into simpler expressions without affecting the meaning. However, there are various viewpoints when associating knowledge.
In addition, since it is necessary to use different viewpoints of association depending on the purpose of grasping the knowledge, there has been a problem that the arrangement and processing of the description contents are naturally limited.

SUMMARY OF THE INVENTION An object of the present invention is to analyze, evaluate, and analyze description contents expressing knowledge in order to solve the above-mentioned problems of the prior art.
An object of the present invention is to provide a method of extracting information, deepen understanding of knowledge, and efficiently detect errors and waste.

[0007]

The present invention is directed to knowledge information comprising a plurality of condition descriptions, each condition description comprising a conditional expression and an expression executed when the conditional expression is satisfied. . Conditional expressions generally have an equality expression, an inequality expression or an inequality expression in which a condition name and a condition value are connected by an equal sign, an inequality sign or an inequality sign, and these elements are further connected by an AND sign or an OR sign. Expressed by an expression.

According to the present invention, when a conditional expression for each condition description of knowledge information includes an OR expression that connects two expression expressions with an OR symbol, the condition description is divided so as to exclude all OR expressions. If a plurality of equivalent conditional descriptions are generated and the expression of the conditional expression in each conditional description of the knowledge information includes an inequality expression or an inequality expression, all possible descriptions corresponding to the condition name Based on the condition value list that sets the condition value of, the condition description is divided so as to be converted to an equal expression equivalent to an inequality expression or an inequality expression, and a plurality of condition descriptions are generated. Or expression,
For knowledge information including all condition descriptions generated by dividing condition descriptions so as to exclude inequality expressions and inequality expressions, the pattern of the combination of the set condition name order and all possible condition values The method is characterized by a knowledge information analysis method in which all condition descriptions in the knowledge information are arranged on the basis of the above to create an analysis result table.

The present invention also provides knowledge information for detecting duplication of two condition descriptions, inconsistency between two condition descriptions, inclusion relation between two condition descriptions, and the like based on the analysis result table thus obtained. Is characterized by the analytical evaluation method.

Further, the present invention is characterized by a method for analyzing and evaluating knowledge information in which only a condition description belonging to a combination pattern of a combination of a condition name and a condition value specified for an analysis result table is extracted and displayed on a display device. And

Further, the present invention performs the above-described condition description generation processing so as to extract only a condition description belonging to a combination pattern of a combination of a specified condition name and a condition value, and extracts the extracted condition description from the knowledge information. It is characterized by a method of analyzing and evaluating knowledge information in which descriptions are arranged to create an analysis result table.

[0012]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below in detail with reference to the drawings.

FIG. 1 is a configuration diagram of an information processing apparatus for performing knowledge information analysis and evaluation according to the present invention. The processing device 1 is a main body of the information processing device, and its main storage device stores and executes programs of a command reception unit 2, a knowledge analysis unit 3, a knowledge evaluation unit 6, a knowledge extraction unit 7, and a display unit 9. You. The storage device 8 is a part of the main storage device of the processing device 1 or an external storage device connected to the processing device 1.
A condition list such as a condition order list 61 and a result management table such as an analysis result management table 64 are stored. The knowledge information 30 is information that expresses the knowledge information by using conditional descriptions in IF to THEN format. The condition list is a list that lists conditions for analyzing the knowledge information 30 and extracting analysis / evaluation results. The result management table is a table for storing analysis / evaluation results. The input device 19 is connected to the processing device 1,
It is a device such as a keyboard or a pointing device for inputting a command for analyzing / evaluating the knowledge information 30 or for inputting conditions for analyzing / evaluating the knowledge information 30. Display device 2
Reference numeral 1 denotes a device that is connected to the processing device 1 and displays an analysis / evaluation result.

The command receiving unit 2 is a processing unit that transfers control to the knowledge analyzing unit 3, the knowledge evaluating unit 6, or the knowledge extracting unit 7 according to a command input from the input device 19. The knowledge analysis unit 3
OR of the specified knowledge information 30 in the condition description
A processing unit that simplifies expressions and simplifies expressions such as inequality signs to pattern the condition description to create the analysis result management table 64. The knowledge evaluation unit 6 is a processing unit that checks the created analysis result management table 64 for waste such as inconsistency, omission, duplication, etc. in the condition description and between the plurality of condition descriptions, and creates a management table of the evaluation result. . The knowledge extracting unit 7 is a processing unit that selectively extracts an analysis / evaluation result of a condition description corresponding to a specific extraction condition. The display unit 9 is a processing unit that displays on the display device 21 the results analyzed, evaluated, and extracted by the knowledge analysis unit 3, the knowledge evaluation unit 6, and the knowledge extraction unit 7. Note that FIG. 1 does not show general software such as an operating system (OS) or a window system that intervenes between the storage device 8, the input device 19, the display device 21, and the processing unit.

FIG. 2 shows knowledge information 3 to be analyzed and evaluated.
It is a figure which shows the specific example of the condition description of 0, and has shown the description example by the most general "IF-THEN". In the present embodiment, a condition description based on the description rules of the C language, which is one of the programming languages, is used.
The structure of each condition description 31 is composed of a conditional expression 32 described in parentheses following the IF and a THEN expression 33 following the parentheses. Each conditional expression 32 is a single condition 34 or a plurality of conditions connected by AND or OR. Each condition 34 is composed of a condition name 35 on the left side, an operator 36, and a right side expression. In particular, when the right side expression is unary and a constant, it is referred to as a condition value 37.
The condition name 35 corresponds to a specific one, for example, a traffic light, and the condition value 37 corresponds to, for example, a red, orange, or blue traffic light. The condition name 35 and the condition value 37 may be expressed by words mixed with Chinese characters and kana. In FIG. 2, a conditional expression number 38 is added as a serial number to each condition description 31 in order to correspond to an analysis / evaluation result described later. In this embodiment, since the conditional description expression based on the description rule of the C language is used, AND and OR are described by “&&” and “||” respectively, and the inequality signs (<, >, ≦, ≧) are described as “<”, “>”, “<=”, “> =”, respectively, and the equal sign (=) and the inequality sign (≠) are expressed as “==” and “! =”, Respectively. Shall be described. Examples of expressions relating to analysis, evaluation, and extraction operations for these condition descriptions will be described with reference to FIGS.

FIG. 3 is a diagram showing an example in which the result of knowledge information analysis according to the present invention is expressed in a table format, and shows the result of patterning the condition description 31 based on the possible condition values 37 of the basic condition 41. I have. The pattern number 42 is a number assigned corresponding to one combination pattern of the combinations of the basic conditions 41. Whether the basic condition 41 is a condition that appears in common in a plurality of condition descriptions 31 or a condition that frequently appears in a plurality of condition descriptions 31 among the condition names 35 on the left side of the condition 34 constituting the conditional expression 32 of each condition description 31 ,
Alternatively, it is specified by the operator. In the example of FIG. 3, a combination of a condition number 37 for a pattern number 42 and a basic condition 41 characterizing each pattern is shown for each pattern, and a conditional expression number 38 and each condition description 31 for associating the condition description 31 belonging to each pattern. Are described as a conditional expression 43 outside the basic condition and a THEN expression 33. Conditional expression number 3
8 is a condition description 31 corresponding to one combination pattern
And matches the conditional expression number 38 of the condition description 31 in the knowledge information 30. When one condition description 31 is divided into a plurality of condition descriptions 31, each of the divided condition descriptions 31 becomes a conditional expression number 38 by adding a branch number to the conditional expression number 38 of the original condition description 31. As described later, in the result of the knowledge information analysis, all the OR expressions in the conditional expression 43 outside the basic conditions are replaced by AND expressions. However, for convenience of explanation, OR expressions in the conditional expression 43 outside the basic conditions are replaced here. It is shown in the form left behind.

FIGS. 4 to 6 show a plurality of condition descriptions 3 for simplifying the expression of each condition description 31 constituting the knowledge information.
It is a figure showing the example in the case of dividing into one. 4 shows a case where an OR expression 45 is included in the condition description 31. The OR expression 45 shown in FIG. 4B while maintaining the logical equivalent of the condition description 31 of FIG. This shows a specific example in a case where a complicated condition description is simplified by dividing into a plurality of condition descriptions 31 not including 45.
In the example of FIG. 4, the parentheses in the conditional expression are nested one by one, which is a relatively simple case. However, a complicated case where parentheses are nested multiple times can be similarly simplified. FIG. 5 shows a case where an inequality expression 47 using an inequality sign (<,>, ≦, ≧) and an inequality expression 48 using an inequality sign (等) are included in the condition description. (B), while retaining the logically equivalent meaning of the condition description 31 based on the allowable condition value 49 that can be taken for each condition, only the equality expression 46 that does not use inequality or inequality as shown in FIG. Shows a specific example in a case where a complicated condition description is simplified by dividing it into a plurality of condition descriptions. FIG. 6 shows a specific example of a complicated condition description 31 in which an OR expression 45, an inequality expression 47, and an inequality expression 48 are mixed in the condition description. An example in which the condition description 31 is simplified by combining the division into the description 31 and the division into a plurality of condition descriptions 31 using only the equality expression 46 without using the inequality expression 47 or the inequality expression 48 will be described. I have. Due to the simplification of these condition descriptions, the condition 34 which becomes the basic condition 41 in the specific example of FIG.
Are combined by OR, or when the condition value 37 that can be taken as the basic condition 41 by the expression using the inequality sign or the inequality sign is not explicit, the simplification of the above condition description 31 allows the condition description 31 shown in FIG. Patterning becomes possible.

FIG. 7 is a diagram showing an example of an analysis result of knowledge information by simplification or patterning of a condition description and an evaluation result 51 for evaluating redundancy and inconsistency between the condition descriptions. FIG. 7A is a specific example of an analysis / evaluation result showing all possible combinations of allowable condition values 49 of each basic condition 41, and FIG. A specific example is shown in which only the analysis / evaluation results for the combination of the basic condition 41 and the specific condition value 37 are extracted. In these figures, among the combination patterns, those for which the conditional expression number 38 does not exist indicate that there is no condition description for the combination. They aim for suggestions.

FIG. 8A shows the result of selecting only the ones to be watched from among the evaluation results 51 in FIG. 7A. First of all, the redundancy evaluation targets "duplicate" where exactly the same condition description exists and "include" where one condition description semantically includes the other condition description, and aims at avoiding useless condition description. And The evaluation of inconsistency is intended for "inconsistency between expressions" between condition descriptions and "inconsistency in expressions" in one condition description, and aims at improving the reliability of the condition description. In FIG. 8A,
In the example of "duplicate", the description contents are exactly the same for both the condition description "9" and "13" of the conditional expression No. 38, indicating that one condition description can be removed. FIG. 8B shows an expression of both condition descriptions before analysis, and is shown in an example of “duplication” in FIG. 8A by avoiding an inequality expression or an inequality expression by analyzing knowledge information. Such a clear comparison is possible. In the example of "inclusion" in FIG. 8A, the description contents of the non-basic condition expression 43 in the condition descriptions of both "6" and "11" of the conditional expression number 38 have an inclusion relationship with each other. It is pointed out that the condition description of the other party can be removed. FIG. 8C shows the expression of both condition descriptions before analysis, and FIG. 8 is obtained by avoiding the inequality expression and the inequality expression by analyzing the knowledge information.
A clear comparison is possible as shown in the example of "include" in (a). In the example of “contradiction between expressions” in FIG. 8A, the description contents of the THEN expression 33 are different in the condition descriptions of both “7” and “12” of the conditional expression No. 38, and they are mutually contradictory. Point out. FIG. 8D shows the expression of both condition descriptions before analysis, and FIG. 8 is obtained by avoiding the inequality expression and the inequality expression by analyzing the knowledge information.
A clear comparison is possible as shown in the example of "contradiction between formulas" in (a). In the example of “inconsistency in expression” in FIG. 8A, the basic condition 4
1 point out that there is a contradiction. FIG.
8 shows an expression of this condition description before analysis, and FIG. 8 is obtained by avoiding an inequality expression and an inequality expression by analyzing the knowledge information.
A clear indication as shown in the example of “a contradiction in expression” in FIG.

FIG. 9 is a diagram showing a specific example of a table configuration required for analysis, evaluation, and extraction processing on knowledge information.
FIG. 8A shows an example of the configuration of the condition order list 61, in which the conditions that become the basic conditions 41 and their order are set by setting the condition name 35 and the condition type 55. After the basic condition 41, the condition names 35 of the conditions 34 other than the basic condition and their order may be registered. FIG. 7B shows an example of the configuration of the condition value list 62. One or a plurality of condition values 37 are provided for each condition name 35.
Is set, an allowable condition value 49 shown in FIG. 7A or the like is set for each condition. FIG. 9 (b-1) is an example of a condition value list 62-1 listing all possible condition values 37 for the condition name 35, and the extraction condition before the knowledge analysis unit 3 analyzes the knowledge information 30. Is not input, the entire condition value list 62-1 is applied to the analysis processing. FIG. 9B-2 is an example of a condition value list 62-2 listing only the condition values 37 corresponding to the extraction conditions. The knowledge analysis unit 3 inputs the extraction conditions before analyzing the knowledge information 30. In this case, the condition value list 62-2 is applied to the analysis processing. The condition value list 62-2 of FIG. 9B-2 is a condition value set when the extraction condition list 63 of FIG. 9C is applied to the condition value list 62-1 of FIG. 9B-1. The example of the list 62-2 is shown. The condition value list 62-1 is a list preset in the storage device 8 before the processing of the knowledge analysis unit 3. The condition value list 62-2 is a list created from the condition value list 62-1 when an extraction condition is input during the processing of the knowledge analysis unit 3. Figure (c)
Is a configuration example of the extraction condition list 63, and sets an extraction condition value 50 shown in FIG. 7B by setting a condition name 35 and a condition value 37 as conditions for extracting a condition description. According to the example of FIG. 9C, since the basic condition a is not set as the extraction condition, all possible condition values of the basic condition a are set as the extraction conditions. FIG. 6D shows an example of the configuration of the analysis result management table 64, which is a management table that expresses the analysis results in a table format as shown in FIGS.
2. Consists of conditional expression number 38, basic condition 41, non-basic condition expression 43, and THEN expression 33. FIG. 11E is a configuration example of the evaluation result management table 65, which is a management table expressing the evaluation result 51 in a table format as shown in FIG. 7 and FIG. Consists of conditional expression number 56. FIG. 7F shows an example of the configuration of the analysis impossible management table 66 for storing the condition description that has become analysis impossible in this analysis, and is composed of a conditional expression number 38.

FIG. 10 is a flowchart showing the flow of processing for performing analysis, evaluation and extraction and displaying the processing results on the display device 21. When a knowledge information name is input via the input device 19 (S1), the command receiving unit 2 retrieves the knowledge information 30 having the designated name from the storage device 8 (S2).
When the user inputs selection of whether to specify or automatically set an analysis command and a basic condition as a base of the analysis as processing options (S3), the knowledge analyzer 3 is started. When the automatic setting is specified, the knowledge analysis unit 3 refers to the knowledge information 30 and extracts one or a plurality of basic conditions that appear in common in each condition description 31 or appear at a frequency equal to or higher than a preset frequency. (S4), condition order list 6 in order of appearance
It is set to 1 (S5). In the case of the basic condition designation, when one or a plurality of condition names 35 are input via the input device 19 (S6), the condition names 35 are similarly set in the condition order list 61 in the order of input (S5). . After setting the condition name 35 of the basic condition in the condition order list 61, a condition name 35 other than the basic condition may be input and set in the condition order list 61. Next, the knowledge analyzer 3 waits for an extraction condition input before the analysis processing (S7). If no input of the extraction condition is specified via the input device 19, the process proceeds to S10. When the extraction condition is input, the input condition is set in the storage device 8 as the extraction condition list 63 (S8), and the condition value list 62-1 is set.
The condition value 35 corresponding to the condition name 35 satisfying the set extraction condition list 63 is extracted from the condition value list 62.
-2 is created and stored in the storage device 8 (S9). If there is no input of the extraction condition, the knowledge analysis unit 3 proceeds with the analysis of the knowledge information 30 based on the condition value list 62-1.
When there is an input of an extraction condition, the analysis process of the knowledge information 30 is advanced based on the condition value list 62-2. Hereinafter, the condition value list 62 indicates the condition value list 62-1 or the condition value list 62-2 according to this condition.
Thereafter, for each condition description 31 of the specified knowledge information 30, a condition description including an OR expression is divided (S10), a condition description order is changed (S11), and a condition description including an inequality / inequality expression is divided (S12). ) And patterning of the condition description based on the combination of the basic condition values (S13), and the analysis result management table 64 obtained as a result is stored in the storage device 8 (S14). Next, the display unit 9 edits the analysis result management table 64 into a table format as shown in FIG. 3 and displays it on the display device 21 (S15), and returns to S3. S10, S11, S1
Details of each of the analysis processes 2 and S13 will be described later with reference to the flowcharts of FIGS.

After the analysis processing is completed, if a comprehensive evaluation for performing redundancy evaluation, inconsistency evaluation, or both is specified as a processing option (S3), the knowledge evaluation unit 6 is activated. The knowledge evaluation unit 6 evaluates the created analysis result management table 64 for redundancy (S16) and inconsistency (S1) according to designation.
7) Or, a process of comprehensive evaluation (S18) is performed, and if a redundant description or inconsistent description is detected in the result condition description 31, the results are collectively stored as an evaluation result management table 65 and stored in the storage device 8 (S19). Next, the display unit 9 edits the analysis result management table 64 and the evaluation result management table 65 into a table format as shown in FIG. 7A, displays the edited result on the display device 21 (S20), and returns to S3. The details of each evaluation process in S16, S17, and S18 will be described later with reference to the flowcharts in FIGS.

Instead of displaying the entire analysis / evaluation result of the knowledge information 30, it is possible to display only the analysis / evaluation result of the specific condition description 31 by specifying the extraction condition. After the analysis processing is completed, when an extraction command is input as a processing option (S3), the knowledge extraction unit 7 is activated. When an extraction condition including a condition name 35 and a condition value 37 is input via the input device 19 (S21), the knowledge extraction unit 7 sets the input extraction condition as an extraction condition list 63 (S22). Next, only the condition description having the condition name 35 and the condition value 37 set in the extraction condition list 63 is extracted from the analysis result management table 64 obtained in S14 (S23). If the evaluation result management table 65 is obtained in S19, only the evaluation result corresponding to the extracted condition description 31 is extracted. Next, the display unit 9 edits the extracted analysis result management table or analysis / evaluation result management table into a table format as shown in FIG. 3 or FIG.
It is displayed on the top (S24), and returns to S3.

After the analysis processing is completed, when the entire display is designated as the processing option (S3), the command receiving unit 2 activates the knowledge extracting unit 7. The knowledge extracting unit 7 extracts the entire analysis result management table 64 obtained in S14 (S25).
If the evaluation result management table 65 has been obtained in S19, the evaluation result management table 65 is also extracted. Next, the display unit 9 edits the analysis result management table 64, if present, into the table format as shown in FIG. 3 or FIG. 7A, and displays it on the display device 21 (S24). Return to S3. That is, the display on the display device 21 can be returned to the display state before the extraction display.

When "return" is input as a processing option (S3), the process returns to S1. When "end" is input (S3), the analysis / evaluation / extraction process ends.

Next, the details of each of the analysis processes in S10, S11, S12 and S13 will be described with reference to the flowcharts of FIGS.

FIG. 11 is a flowchart showing the flow of the process of dividing the condition description including the OR expression in S10. First, the knowledge analysis unit 3 positions the check pointer in the condition description 31 at the head of the knowledge information 30 (S31), and extracts the condition description 31 at the position of the check pointer (S32). If there is an unnecessary parenthesis that does not affect the logical meaning even if it is not in the condition description, it is removed (S33). The parentheses indicating the range of the if sentence are unnecessary for the processing and are considered to be for the time being. Then, it is determined whether or not parentheses are further included (S34).
If the parentheses are included, it is determined whether an OR expression exists outside the parentheses (S35). If no OR expression exists outside the parentheses, the outermost parenthesis is removed and an O depending on the outermost parenthesis is removed.
The condition description 31 is divided based on the R expression (S36) and replaced with the original condition description 31 (S38). Unless further parentheses are included inside, all of the OR expressions enclosed by the outermost parentheses are OR expressions that depend on the outermost parenthesis. If parentheses are further included inside, the OR expression included between the outermost parenthesis and the inner parenthesis is subject to division. S3
If an OR expression exists outside the parentheses in step 5, the condition description 31 is divided based on the OR expression (S37) and replaced with the original condition description 31 (S38). Thereafter, the check pointer is positioned at the first condition description 31 among the plurality of replaced condition descriptions 31 (S39), and the process returns to S32. If the condition description 31 does not include parentheses in S34, the condition description 3
It is determined whether an OR expression exists in S1 (S40). If an OR expression exists, the condition description 31 is divided based on the OR expression (S37) and replaced with the original condition description 31 (S3).
8) Position the check pointer at the first condition description 31 among the plurality of replaced condition descriptions 31 (S39),
Return to 32. If the OR expression does not exist in S40, it is determined whether or not the next condition description 31 exists (S41). If so, the check pointer is positioned at the next condition description (S4).
2) Return to S32. If the next condition description 31 does not exist in S41, the knowledge analysis process for the division of the condition description 31 including the OR expression ends. By the above processing, all the parentheses in the condition description 31 depending on the OR expression are removed.

FIG. 12 is a flowchart showing the flow of the process of changing the condition description order in S11. This processing is performed for each condition description 31 of the knowledge information obtained as a result of the above processing.
Are rearranged in the order set in the condition order list 61. First, the check pointer is positioned at the first condition description 31 (S51), and the condition description 31 at the check pointer position is extracted (S52). Condition order list 6
The first condition set to 1 is extracted (S53),
It is checked whether or not the condition description exists in the condition description 31 (S54). If the condition description 31 exists, the condition description position is changed so that the condition in the condition description 31 comes first (S55). Furthermore, it is checked whether there is any other description of the same condition in the condition description (S56).
Is repeated. That is, the description position is changed so that the description under the same condition continues. If the condition is not described in S54, or if the description position of the condition is completely changed in S56, it is checked whether there is a next condition set in the condition order list 61 (S57). Is taken out (S58), and the process returns to S54. For the next condition, the description position is changed so as to follow the condition whose position has already been changed. If there is no next condition in S57, the condition description whose description position has been changed is replaced with the original condition description 31 (S5).
9) It is determined whether the next condition description 31 exists (S6).
0) If there is, the check pointer is positioned at the next condition description (S61), and the process returns to S52. If there is no next condition description in S60, the knowledge analysis process for changing the condition description order ends. By this processing, the description positions of the conditions in the condition description 31 are arranged in the order of the conditions in the condition order list 61. It is to be noted that the description positions of the conditions in the condition description 31 are rearranged according to the order of the conditions in the condition order list 61.
This is for speeding up the process of registering each condition description 31 in the analysis result management table 64 including a combination pattern of conditions described later, and is not an essential process.

FIG. 13 is a flowchart showing a flow of the division process of the condition description including the inequality / inequality use condition in S12. In this process, it is checked whether or not each condition description 31 of the knowledge information obtained as a result of the above process includes a basic condition expressed by an inequality sign or an inequality expression. First, the check pointer is positioned at the first condition description 31 (S7).
1), the condition description 31 at the check pointer position is extracted (S72). Next, it is determined whether or not the condition description includes a basic condition description in an inequality / inequality expression (S73). If the basic condition description in inequality or inequality expression is included, all possible condition values 37 of the corresponding basic condition are extracted from the condition value list 62 based on the description (S74). For example, if there is a basic condition description of a> 0 in the condition description, a condition value 37 of a = 1 is extracted according to the example of FIG. 9B.
If there is no condition value 37 to be taken out, the condition description is deleted. Next, it is checked whether there is any other description of the same basic condition in the condition description (S75). Condition name 3 on the left side
If there is another condition 34 in which 5 is the same, there is another description about the same basic condition. For example, if there is another basic condition description a = 0, a <1 or the like relating to the same basic condition in the condition description, it is determined that there is another description of the same basic condition. The condition description is divided into one or more condition descriptions converted into equivalent equal signs based on condition values that can be assumed if there is no other description of the same basic condition (S76), and replaced with the original condition description (S77). The check pointer is positioned at the first condition description among the plurality of replaced condition descriptions (S78), and the process returns to S72. When there is only one equivalent condition value, no division occurs. If there is another description of the same basic condition in S75, all possible condition values of the corresponding basic condition are fetched from the condition value list 62 based on the description as described above (S79), and the condition value fetched in S74 is Then, the shared portion is taken out (S80). For example, S7
In step S4, the condition value of a = 1 is extracted by the basic condition description of a> 0.
If the condition values of 0 and a = 1 are taken out, the shared part in S80 is a = 1, while the condition values of a = -1 and a = 0 are taken out in S79 by the basic condition description of a <1. Then, there is no share in S80. Further, it is checked whether there is any other description of the same basic condition in the condition description (S81), and if it is, S79 and S80 are repeated. In S80, the condition value extracted in S74, the condition value extracted in S79, the condition value extracted in S79,
····················································································· Extracted When all the descriptions of the same basic condition have been checked in S81, it is checked whether or not there is a shared portion as a result of extracting the shared portion of the condition value in S80 (S82). Similarly, the condition description is divided based on possible condition values (S76), replaced with the original condition description (S77), and the check pointer is positioned at the first condition description among the replaced plurality of condition descriptions (S7).
8) Return to S72. That is, the condition description is divided into one or more condition descriptions in which each basic condition portion that is the same basic condition is converted into an equivalent equal sign. Condition description 31 with basic condition description that does not satisfy condition value list 62
Is deleted. In S82, if there is no shared portion as a result of extracting the shared value of the condition value in S80, the condition description is invalidated and deleted from the knowledge information 30 (S83), and the process proceeds to S93 for the next condition description. That is, if there is no shared condition value, the condition description 31 that combines such the same basic condition description with an AND condition has no meaning and is deleted from the knowledge information 30. If there is no inequality or inequality basic condition description in the condition description in S73, it is checked whether there is a plurality of descriptions about the same basic condition in the condition description (S85). That is, it is checked whether there are a plurality of equal expression with the same condition name 35 on the left side. If there is no description of the same basic condition, it is checked whether any of the basic conditions set in the condition order list 61 has no description in the condition description (S86). Move to the operation for the next condition description. If there are multiple descriptions of the same basic condition in S85, it is checked whether the condition values are the same (S8).
8) If they are the same, duplicate descriptions of the same basic condition are removed (S89). That is, the same equal expression of the same basic condition combined by AND conditions is combined into one equal expression. Further, it is checked whether there is a plurality of descriptions about the same basic condition in the condition description (S90).
Repeat 89. When all the descriptions for the same basic condition have been checked in S90, the condition description is replaced with the original condition description (S91), and the process proceeds to the check in S86. If the condition values are different in a plurality of descriptions for the same basic condition in S88, the condition description is invalidated and deleted from the knowledge information 30 (S92), and S93.
Shifts to the processing for the next condition description. That is, if the condition values on the right side of the equal sign are different for the same basic condition, the condition description 31 is deleted from the knowledge information 30 because it is an inconsistent condition description. Such a condition description is a part of the original condition description including the inequality and inequality expressions divided into a plurality of condition descriptions, and is not inconsistent by the user. I do not care. If there is a basic condition not described in the condition description in S86, all the possible condition values 37 of the corresponding condition are extracted from the condition value list 62, and the condition description is divided based on the condition value (S87). Replace with original condition description (S7
7), the check pointer is positioned at the first condition description among the plurality of replaced condition descriptions (S78), and the process returns to S72. The condition description 31 having a basic condition description that does not satisfy the condition value list 62 is deleted. In S93, it is determined whether there is a next condition description. If there is, the check pointer is positioned at the next condition description (S94), and the process returns to S72. If the next condition description does not exist in S93, the knowledge analysis process for the division of the condition description including the inequality / inequality expression ends.

FIG. 14 is a flowchart showing the flow of the process of patterning the condition description based on the combination of the basic conditions in S13. In this process, the individual condition description 31 of the knowledge information obtained as a result of the above process is registered in the analysis result management table 64 at a position according to the combination pattern of the condition values of the basic conditions. First, all combination patterns are developed based on the types of the basic conditions set in the condition order list 61 and the possible condition values 37 of each condition set in the condition value list 62, and the analysis result management before the condition description registration is performed. The table 64 is created (S100). Next, knowledge information 3 including all the condition descriptions 31 such as the condition description 31 obtained by dividing the check pointer
0 (S101), and the condition description at the check pointer position is extracted (S10).
2). The analysis result management table 64 is searched according to the combination pattern of the condition values of the basic conditions of the condition description to find the registration position of the condition description (S103). If found, it is checked whether another condition description is already present at the position of the pattern (S104). If not, the contents of the condition description are registered at the position according to the format of the analysis result management table 64 (S106). If another condition description has already been registered,
A new empty area is inserted after the registered position (S10
5) In the same manner as described above, the contents of the condition description are registered in the free area according to the format of the analysis result management table 64 (S10).
6). Thereafter, it is determined whether or not the next condition description 31 exists in the knowledge information 30 (S108). If there is, a check pointer is positioned at the next condition description 31 (S109), and the process returns to S102. If the combination pattern is not found in S103, the condition expression number 3 of the condition description is added to the analysis impossible management table 66.
8 is registered (S107). If there is no next condition description in S108, the knowledge analysis process for patterning the condition description based on the combination of the basic conditions is ended.

Next, the details of each evaluation process of S16, S17 and S18 will be described with reference to the flowcharts of FIGS.

FIG. 15 is a flowchart showing the flow of the processing of the redundancy evaluation in S16, the inconsistency evaluation in S17, and the comprehensive evaluation in S18. Based on the analysis result management table 64 obtained as a result of the above-described knowledge analysis processing, the knowledge evaluation unit 6 compares and checks the condition description contents with each other, and registers the result in the evaluation result management table 65. First, the check pointer is positioned at the head of the analysis result management table 64 (S121), and the presence or absence of the registered condition description is checked (S122). If so, the second pointer is placed next to the check pointer in the analysis result management table 64 Positioning (S123), it is checked whether the basic condition value combination pattern of the condition description 31 of the check pointer position and the basic condition value combination pattern of the second pointer position are the same (S124).
3 (S125), and if they are the same, the THEN expression 33 also checks whether they are the same (S12).
6) If the evaluation type is the same, if the evaluation type is the redundant evaluation or the comprehensive evaluation (S127), the evaluation type 57 is added to the evaluation result management table 65.
Is registered as "duplicate" together with the conditional expression number 38 of the two condition descriptions (S128). Thereafter, it is checked whether the end of the analysis result management table 64 is after the second pointer position (S13).
4) If not the end, position the second pointer to the next condition description in analysis result management table 64 (S135), S124
Return to If the end of the table is determined in S134, it is checked whether the position following the check pointer position is the end of the analysis result management table 64 (S136). If not, the check pointer is positioned at the next condition description in the analysis result management table 64 (S13).
7) Return to S122. In S126, the THEN expression 33
Are different from each other, if the evaluation type is inconsistency evaluation or comprehensive evaluation (S129), the evaluation type 57 is registered in the evaluation result management table 65 as "contradiction between expressions" together with the conditional expression number 38 of the two condition descriptions. (S130), and then proceeds to S134. When the conditional expressions 43 outside the basic conditions are different from each other in S125, if the evaluation type is redundant evaluation or comprehensive evaluation (S1
31), it is checked whether the conditional expression 43 other than one basic condition includes the conditional expression 43 other than the other basic condition (S13).
2). If the conditional expression 43 other than one basic condition belongs to a subset of the conditional expression 43 other than the other basic condition, the conditional expressions 43 other than both basic conditions have an inclusive relation. Since all the conditions 34 in the conditional expression 43 other than the basic condition of the condition description 31 are AND-coupled, it is recognized that two condition descriptions 31 having the same condition including the AND expression have an inclusive relation. If it is included, the evaluation type 5 is added to the evaluation result management table 65.
7 is registered as "included" together with the conditional expression number 38 of the two condition descriptions (S133), and the routine goes to S134. S13
If the condition expression 1 does not include the condition expression 43 outside the basic condition, the process proceeds directly to S134. If the basic condition value combination pattern of the condition description of the check pointer position is different from the basic condition value combination pattern of the second pointer position in S124, and S122
If there is no condition description registered in step S1
Move to 36. If it is the end of the analysis result management table 64 in S136, if the evaluation type is inconsistency evaluation or comprehensive evaluation (S138), the evaluation within the condition description alone is performed (S13).
9) Next, if the evaluation type is comprehensive evaluation (S140),
The condition description that cannot be analyzed is evaluated (S1).
41), then terminates the knowledge evaluation process.

Next, the evaluation in the condition description alone in S139 and S1
Forty analysis impossible condition description evaluation will be described with reference to the flowchart of FIG.

FIG. 16A is a flowchart showing the flow of the evaluation process in the single condition description in S139. The knowledge evaluation unit 6 searches the analysis result management table 64 for a missing number of the conditional expression number 38 (S151), and if found, searches the conditional expression number 38 of the analysis impossible management table 66 (S1).
52). If not found, the conditional expression number 38 is registered in the evaluation result management table 65 with the evaluation type 57 as “contradiction in expression” (S153). That is, since the condition description 31 for which the condition value of the basic condition is not determined, such as a> 0 & a <0, is not stored in the analysis result management table 64 or the analysis impossible management table 66, the conditional expression number 38 is omitted. Then S15
The above process is repeated until no missing numbers are found in step 1.
If the conditional expression number 38 is found in the analysis impossible management table 66 in S152, the process returns to S151.

FIG. 16B is a flowchart showing the flow of the process of evaluating the analysis impossible condition description in S140. The knowledge evaluation unit 6 positions the check pointer at the head of the analysis impossible management table 66 (S161), and checks whether the conditional expression number 38 is registered at the check pointer position (S1).
62) If it is registered, the conditional expression number 38 is registered in the evaluation result management table 65 with the evaluation type 57 set to “analysis impossible” (S163), and the check pointer is positioned next to the analysis impossible management table 66 (S164). ), Returning to S162, all conditional expression numbers 38 registered in the analysis impossible management table 66.
And the above processing is repeated.

According to the present embodiment, the extraction condition list 63 can be designated at the start of the analysis processing of the knowledge information 30 or after the analysis processing. Extraction condition list 6 at the start of analysis processing
By specifying 3, the knowledge analysis unit 3 performs analysis processing only on the condition description 31 that satisfies the conditions of the set condition value list 62-2.
The analysis processing time can be reduced as compared with the case where the analysis processing is performed based on the entire device 1, and the necessary storage device capacity can be saved. On the other hand, when the extraction condition list 63 is specified after the analysis processing, the analysis processing time is longer and more storage capacity is required, but the extraction condition list 63 is variously changed with respect to the generated analysis / evaluation result. -There is an advantage that the result obtained by extracting the evaluation result can be obtained.

[0037]

According to the present invention, in a knowledge information analysis / evaluation system used for problem solving and planning, knowledge information can be patterned and classified / arranged, so that it is easy to associate condition descriptions with each other and to compare and refer to them. As a result, the efficiency of understanding the knowledge can be improved. In addition, since contradiction and redundancy between condition descriptions can be evaluated, it is easy to find errors in the knowledge description and to find waste, thereby improving the reliability of the knowledge information.

Further, since all combinations of specific conditions and specific condition values among the conditions included in the condition description constituting the knowledge information are automatically developed, it is possible to discriminate what is described as knowledge and what is not described as knowledge among the combinations. Is clear, it is possible to reconfirm the undescribed combinations, and it is also possible to find a description omission and obtain a suggestion when adding knowledge.

Even when the amount of knowledge information is large or when the number of combinations of conditions and condition values is enormous, analysis by specific conditions and combinations of specific condition values is performed from the analysis and evaluation results for the entire knowledge information.・ It is possible to extract only the evaluation results, and it is also possible to analyze and evaluate only the knowledge information extracted by specific conditions and combinations of specific condition values. There is also an effect that the evaluation result can be easily grasped and the examination can be efficiently performed from a focused viewpoint.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of an information processing apparatus that performs analysis and evaluation of knowledge information according to an embodiment.

FIG. 2 is a diagram showing a specific example of a target condition description.

FIG. 3 is a diagram showing a specific example of expressing a knowledge information analysis result in a table format.

FIG. 4 is a diagram illustrating a specific example in a case where a condition description including an OR expression is simplified.

FIG. 5 is a diagram showing a specific example in a case where a condition description including an inequality expression and the like is simplified.

FIG. 6 is a diagram showing another specific example when simplifying a complicated condition description.

FIG. 7 is a diagram showing an example of a result table obtained by analyzing and evaluating knowledge information.

FIG. 8 is a diagram showing a specific example of a result of evaluating knowledge information.

FIG. 9 is a diagram illustrating a data configuration of tables used when analyzing, evaluating, and extracting knowledge information in the embodiment.

FIG. 10 is a flowchart illustrating an overall flow of processing for analyzing, evaluating, extracting, and displaying knowledge information according to the embodiment;

FIG. 11 is a flowchart illustrating a flow of a process for simplifying a condition description including an OR expression according to the embodiment.

FIG. 12 is a flowchart illustrating a processing flow when changing the expression order in the condition description according to the embodiment;

FIG. 13 is a flowchart illustrating a flow of a process for simplifying a condition description including an inequality expression according to the embodiment;

FIG. 14 is a flowchart illustrating a flow of processing for patterning a condition description based on a combination of basic conditions according to the embodiment.

FIG. 15 is a flowchart illustrating a flow of a process for checking redundancy or inconsistency with respect to a condition description according to the embodiment.

FIG. 16 is a flowchart illustrating a flow of processing for checking a condition description in a single condition description and a condition description that cannot be analyzed according to the embodiment.

[Explanation of symbols]

3: Knowledge analysis unit, 6: Knowledge evaluation unit, 7: Knowledge extraction unit, 3
1 ... condition description, 32 ... condition expression, 33 ... THEN expression, 35
... condition name, 37 ... condition value, 41 ... basic condition, 43 ... basic condition non-conditional expression, 45 ... OR expression, 46 ... equality expression, 47
... inequality expression, 48 ... inequality expression, 61 ... condition order list, 62 ... condition value list, 63 ... extraction condition list, 64
... Analysis result management table, 65 ... Evaluation result management table

 ──────────────────────────────────────────────────続 き Continued on the front page (72) Inventor Kazuo Kurihara 1070 Ma, Hitachinaka City, Ibaraki Prefecture Inside Mito Plant of Hitachi, Ltd.

Claims (9)

[Claims] 1. Knowledge information is composed of a plurality of condition descriptions,
Each condition description is composed of a conditional expression and an expression executed when the conditional expression is satisfied, and the conditional expression defines a condition name and a condition value with both sides of an equal sign, an inequality sign or an inequality sign. A method for analyzing, by an information processing apparatus, knowledge information in a case where expressions of expressions to be performed are expressions combined by AND symbols or OR symbols, wherein each conditional expression of the knowledge information is expressed by two expressions. Is included by OR symbols, the condition description is divided to eliminate all OR expressions, and a plurality of equivalent condition descriptions are generated. For each condition description of the knowledge information, If the expression in the expression includes an inequality expression or an inequality expression, the expression is converted to an inequality expression or an inequality expression based on a condition value list that sets all possible condition values corresponding to the condition name. Convert to equivalent equals representation In this way, the conditional description is divided to generate a plurality of conditional descriptions, and all the conditional descriptions generated by dividing the conditional description so as to exclude OR expressions, inequality expressions, and inequality expressions from the conditional expressions as described above. Create an analysis result table by arranging all the condition descriptions in the knowledge information based on the set order of condition names and all possible condition value combinations for the knowledge information including the condition description. Analysis and evaluation method of knowledge information characterized by: 2. An expression of an expression related to a specific condition name among condition names appearing in the conditional expression is regarded as a basic condition, and the basic condition is converted from the conditional expression into an OR expression, an inequality expression and an inequality expression as described above. 2. The knowledge information including a condition description generated by dividing a condition description so as to exclude an expression is created based on a combination pattern of basic conditions. Analytical evaluation method. 3. A method according to claim 1, wherein, among the condition names appearing in said conditional expression, a condition name commonly appearing in each condition description and an expression of an expression relating to a condition name appearing at a predetermined frequency or more are regarded as basic conditions. The method for analyzing and evaluating knowledge information according to claim 2. 4. When the conditional expressions in the analysis result table belong to the same combination pattern and the expressions executed when the conditional expressions satisfy the conditional expressions are the same, the conditional expressions are duplicated. 2. The method for analyzing and evaluating knowledge information according to claim 1, wherein: 5. A condition description in the analysis result table, wherein when the condition expressions of the two condition descriptions are composed of a part belonging to the same combination pattern and one part including the other, one condition description is provided. 2. The method for analyzing and evaluating knowledge information according to claim 1, wherein it is determined that the condition information includes the other condition description. 6. The condition description in the analysis result table is as follows.
If the conditional expressions of two conditional descriptions belong to the same combination pattern and the expressions executed when satisfying the conditional expressions are different,
2. The method according to claim 1, wherein the two conditions are determined to be inconsistent.
Analysis and evaluation method of the described knowledge information. 7. The analysis result table according to claim 1, wherein only a condition description belonging to a combination pattern of a combination of a specified condition name and a condition value is extracted and displayed on a display device. Analysis and evaluation method of knowledge information. 8. The above-described generation processing is performed so as to extract only a condition description belonging to a combination pattern of a combination of a specified condition name and a condition value, and the extracted condition descriptions in the knowledge information are arranged and analyzed. 2. The method for analyzing and evaluating knowledge information according to claim 1, wherein a result table is created. 9. A combination having no condition description corresponding to a pattern of a combination of a specified condition name and all possible condition values, the condition description including all condition descriptions generated. 2. The method for analyzing and evaluating knowledge information according to claim 1, wherein a pattern is extracted.