JP2821261B2 - Case search and inference processing method - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a method of searching a case database and a method of inference processing for searching a case and using the searched case to solve a problem.

[Conventional technology]

The case information is data composed of data representing a problem and data representing an answer or solution to the problem. One purpose of using the case information is that the system searches for a solution case (similar example) of a problem similar to the problem currently desired to be solved, and the user refers to the answer or solution there. Another purpose of use is for the system to search for similar examples, make inferences using similar examples, and find a solution to the current problem. Therefore, the technical problem in the use of case information is to develop a method of searching for similar examples that can be used as references for the current problem from among the stored multiple cases, and a method of solving problems using similar examples. That is.

Conventional methods for searching and using similar examples include, for example, case-based training workshops.
Proceedings (1988) pp. 21-30 (Case-Based Reas)
oning Workshop (1988) pp21-30). Here, the problem is represented using a group of objectives to be achieved in the problem and a group of constraints to be observed in solving the problem. First, for the similar example search, the following processing method is proposed.

(1) The purpose group of the current problem and the problem group of the case are compared, and a case containing many common purposes is selected.

(2) When a plurality of cases are obtained in the processing of (1), the current problem constraint group is compared with the case problem constraint group, and a case in which more important constraints are shared more. Select (However, information on the importance of the constraint is given to the system in advance.) This method is basically based on the following idea.

(1) The more the current problem and the case problem share more features, the more similar the case.

(2) In order to reflect not only the number of shared features but also their importance in processing, weighting of features is performed.

For the use of similar examples, the following processing methods have been proposed.

(1) A partial problem represented by a feature group common to the current problem and the case problem is regarded as a common partial problem between the current problem and the case problem, and a part corresponding to this partial problem is determined from the solution of the case. Take out.

(2) Modify the extracted partial solution so as to conform to the current problem and use it as a part of the solution to the current problem.

[Problems to be solved by the invention]

In many fields, sub-problems contained in one problem are not independent of each other. That is, the features that represent the problem are not independent of each other. In the above conventional method, since the interference relation between the features is not taken into consideration, a case that is less useful for solving the current problem may be searched.

For example, in a computer room layout problem, the arrangement of individual devices can be considered as a partial problem. Here, the arrangement of the card reader and the arrangement of the katakana printing keyboard card punch are dependent on the arrangement of the console display. That is, there is an interference relationship between the three subproblems.

Assume that a set of names of devices to be arranged is used as a feature group representing a problem. Here, it is assumed that the current problem is represented by {card reader, katakana printing keyboard card punch, console display}, that is, the layout problem of the card reader, katakana printing keyboard card punch, and console display. Also, the case includes {card reader, katakana printing keyboard card punch) but does not include a console display (for example, it may be placed in another room). In this case, the current problem and the case problem are:
Sharing Katakana printing keyboard card punch 孔
These arrangements in the case cannot be applied to the current problem. This is because the arrangement of the card reader and the katakana printing keyboard card punch differs depending on the presence or absence of the console display.

On the other hand, suppose the current problem involves a card reader and a katakana printing keyboard card punch and not a console display. Assume also that the case problems include a card reader, a katakana printing keyboard card punch, and a console display. Also in this case, for the same reason as described above, the case cannot be used for the arrangement of the card reader and the katakana printing keyboard card punch.

In this way, case solutions are not always available for common features. Therefore, a case with many common features is not always a useful case.

In the above example, the availability of the console display determines the usefulness of the case. However, for some problems, the console display is not important.
For example, the arrangement of the communication control processing device and the power supply does not depend on the arrangement of the console display. Thus, when searching for cases in these arrangements, it is not important whether the cases include a console display. Thus, the importance of the feature depends on the combination of the problem and the case, and is not constant. Thus, even if the feature weighting is used, a useful case cannot always be retrieved.

As described above, according to the conventional case search method, a case with little usefulness may be searched for the current problem solving. Therefore, there is a problem that time is wasted due to redoing the search, and a determination error based on an inappropriate case occurs.

Further, in the conventional case utilization method, all partial solutions of cases corresponding to common characteristics of the current problem and the problem of the case are used as partial solutions of the current problem. Therefore, in using the case, it was necessary to correct the influence of interference between features. For example, in the above example, the arrangement in the case of the card reader and the katakana printing keyboard card punch is not available as it is due to the influence of the console display, so that it is necessary to correct these arrangements. for that reason,
In the conventional case utilization method, complicated processing for correcting interference between features is required, and there have been problems that a development period and a development cost of the inference system are increased, and performance and reliability are reduced.

Further, in the conventional case search and use method, only the case where all the solutions of the current problem are obtained from the case has been assumed.
However, in reality, there is a case where the user wants to specify a part of the solution of the current problem or a part of the solution by another inference method that does not use the case. The conventional case search and use method could not meet such a demand.

A first object of the present invention is to provide a case search method for preferentially searching for cases having great utility for solving a current problem.

A second object of the present invention is to provide a case utilization method which does not need to perform interference correction between features for a case solution.

A third object of the present invention is to provide a case search method in which a part of the solution to the current problem is predetermined and the remaining partial solution is obtained.

Further, a fourth object of the present invention is that a part of the solution of the current problem is predetermined, and when the remaining partial solution is obtained, it is not necessary to perform the interference correction between features on the solution of the case. The purpose is to provide a case use method.

[Means for Solving the Problems] In order to achieve the first and second objects, first,
Clarify how the interfering relationship between problem features affects the usefulness of the case. Regarding the usefulness of the case,
The following properties hold:

(Theorem 1) When the subproblem P of the case problem satisfies the following condition, the solution or solution S of the corresponding case is obtained as a part of the solution or solution of the current problem without correction of interference. Available.

(1) P is a partial problem corresponding to a feature set A common to the current problem and the case problem.

(2) A is not interfered by features that are only included in one of the current problem and the case problem.

 We briefly prove the above theorem.

(Proof) Assuming that a partial problem corresponding to A in the current problem is P ', P and P' are equal according to the condition (1). From condition (2), P and P 'are independent of the case problem and other parts of the current problem. Therefore, the solution or solution S of P in the case is also the solution or solution of P 'in the current problem solving. Therefore, S
Is a partial or partial solution of the current problem,
Available without interference correction.

FIG. 16 shows the problem of arranging a console display, a card reader, a katakana printing keyboard card punch, a central processing release device, and a disk drive. The following shows a portion that can be used without correction when using an arrangement example of a drive device and a communication control processing device.
Block 1601 is a feature set representing the current problem, block 16
02 is a feature set representing the problem of the case, and block 1603 is a feature set (set A in Theorem 1) corresponding to the used part. A block 1603 is a portion that is not interfered by a feature (console display) included in only one of the common portions of the blocks 1601 and 1602.

The partial problem represented by block 1603, that is, the problem of the arrangement of the central processing unit and the disk drive, is
(= P ′), and the solution of the corresponding case, that is, the arrangement of the central processing unit and the disk drive unit is represented by Theorem 1.
(A part that can be used without interference correction).

It is considered that a case having a large usefulness for the current problem solving is a case where the partial problem P ′ that can be solved using the above case is a large case. The size of P 'can be approximated by the number of elements of the corresponding feature set A.

Therefore, in order to achieve the first object, the present invention provides a case base for storing at least one case information including data representing a problem and data representing the answer or solution thereof, a means for inputting the problem, In an information processing system having a case search unit for searching case information including a problem similar to a specified problem or a partial problem from a case base, a dependency storage unit is provided which indicates a relationship between features of a group of assumed problems. , When a question is input from the input means, the case search means refers to the dependency information,
The feature is to preferentially search for cases that are common to the input problem and the case problem, and that have a large number of non-interfering features from the features contained only in one of the current problem and the case problem. {Case search method (1)}.

The part of the solution or solution of the case that can be used for the current problem without correcting the interference is the partial solution or solution S in Theorem 1.

Therefore, in order to achieve the second object, the present invention provides a case base for storing at least one case information including data representing a problem and data representing its answer or solution, a means for inputting the problem, In an information processing system having a case searching means for searching case information including a problem similar to a searched problem or a partial problem from a case base and an inference means for finding a solution to the input problem using the searched case, When the dependency storage unit that represents the relationship between the features of the problem group and the case search unit searches for a similar example of the input problem, the inference unit refers to the relationship information to determine the current problem. And a case solution that corresponds to a set of features that are common to the case problem and that are not interfered by features that are only included in either the current problem or the case problem. Rui is characterized by utilizing a part of the solution {inference processing method (1)}.

Next, in order to achieve the third and fourth objects,
The above Theorem 1 is extended. (Theorem 2) In the current problem, it is assumed that no solution or solution is determined, or a part of the solution or solution is determined. On the other hand, it is assumed that a part or the whole of the solution or solution is found in the case problem.

At this time, when the subproblem P of the case problem satisfies the following condition, a part S of the solution or solution of the case corresponding to the subproblem P is a solution or solution of the current problem without correction of interference. Available as a department.

(1) P is a partial problem corresponding to a feature set A common to the current problem and the case problem.

(2) P has no solution in the current problem,
In the case, a solution has been found.

(3) A is not interfered by features that are only included in one of the current problem and the case problem.

(4) A is not interfered by a feature corresponding to a subproblem that is a common subproblem of the current problem and the problem of the case, and a solution is obtained in both, and the solutions of both are different.

(5) A is not interfered with by a feature corresponding to a subproblem that is a common subproblem of the current problem and the problem of the case, and for which no solution is found in the case.

 We briefly prove the above theorem.

(Proof) Assuming that a partial problem corresponding to A in the current problem is P ', P and P' are equal according to the condition (1). From the condition (2), a solution S is obtained for P, and a solution is not obtained for P ′. From the conditions (3) to (5), P and P ′
Is only interfered with by the subproblems for which the solution is already equal in the current problem and the case, and independent of the other subproblems. Thus, P and P 'are either independent of other subproblems or suffer equivalent interference. Therefore, the solution or solution S of P in the case is also the solution or solution of P 'in the current problem solving. Thus, S can be used as a partial or partial solution of the current problem without interference correction.

As described above, in order to achieve the third object, according to the present invention, a case base storing one or more case information including data representing a problem and data representing a solution or a solution thereof, a means for inputting the problem, In an information processing system having a case search means for searching case information including a problem similar to a problem to be input or a partial problem thereof from a case base, a dependency storage means for representing a relationship between features of a supposed problem group is provided. When the problem is input from the input means, the case search means refers to the dependency information and finds the solution of the partial problem corresponding to the characteristic common to the input problem and the problem of the case only in the case. Are not affected by features that are only included in one of the current problem and the case problem, and the partial solution differs between the current problem and the case. Priority is given to cases where the number of features that do not receive interference from features corresponding to subproblems and that do not interfere with features corresponding to subproblems whose partial solutions are found only in the current problem {Case search method (2)}.

In view of the above, the fourth object is to provide a case base storing one or more case information including data representing a problem and data representing the answer or solution thereof, a means for inputting the problem, and an input problem or a partial problem thereof. In an information processing system that has case search means for searching case information containing similar problems from a case base and inference means for finding a solution to the input problem using the searched cases, A dependency relationship storage unit that represents a relationship between the input problem and the case problem by referring to the dependency information when the case search unit searches for a similar example of the input problem. And the corresponding subproblem is found only in the case, and is included in only one of the current problem and the case problem. It seems that there is no interference from the features that are not interfered with, and that there is no interference from features corresponding to subproblems whose partial solutions differ from the current problem in the case, and that partial solutions are found only in the current problem. The inference processing method (2) is characterized by using a solution or solution part of a case corresponding to a set of features that are not interfered by features corresponding to various partial problems.

Finally, theorem 1 is modified to find another way to achieve the first and second objectives.

(Theorem 3) When a similar example of the current problem is searched using the following search conditions (1) and (2), a case corresponding to the feature set A of the condition (1) among the solutions of the similar example Solution or solution S
Can be used as part of a solution or solution to a current problem without interference correction. However, an arbitrary set of features common to the current problem and the case problem is denoted by A '.

(1) The feature set representing the problem in the case must include the set A of all features affecting A 'that are included in the current problem.

(2) The feature set representing the problem of the case must not intersect with the set B of the features affecting A 'that are not included in the current problem. (That is, the intersection with B must be empty.) The above theorem is simply proved.

(Proof) Since the case is searched using the condition (1), A is a set of features common to the current problem and the case problem. A
Is a closure of an interference relationship based on A '(a feature set affecting A', a feature set affecting the feature set, a feature set affecting the feature set ... This is the final feature set obtained as a result obtained until it does not become larger. Thus, A is equal to the set of features that interfere with A. Therefore, from (1), all of the features that interfere with A and are included in the current problem are also included in the problem of the case. As a result, features included only in the current problem do not interfere with A. Similarly, from (2), A
Features that interfere with, and are not included in the current problem, are not included in the case problem. Thus, features included only in the case problem do not interfere with A. From the above, A (the subproblem represented by) satisfies the two conditions of Theorem 1,
The part corresponding to A in the solution or solution of the case can be used without correcting the interference.

According to Theorem 3, in order to achieve the first object, in the present invention, a case base for storing one or more case information including data representing a problem and data representing the answer or solution thereof, and a means for inputting the problem In an information processing system having a case search unit for searching case information including a problem similar to an input problem or a partial problem from a case base, a dependency storage unit representing a relationship between features of a group of assumed problems And, when a question is input from the input means, referencing the dependency information to determine a feature that the search target case should have, and a step of obtaining a feature that the search target case should not have, When a question is input from the input means, the case search means searches the case information using the two types of characteristics obtained in the two steps. It features that {case retrieval system (3)}.

Further, in order to achieve the second object, the present invention provides:
In the above information processing system, there is provided inference means for finding a solution to the input problem using the searched case, and the inference means uses a solution or a solution part of the case corresponding to the feature that the search target case should have. {Inference processing method (3)}.

[Action]

According to the case search method (1), the case search means refers to the dependency information and has a feature common to the input problem and the problem of the case, and further includes one of the current problem and the problem of the case. A case in which the number of elements in a set of features that are not subject to interference from features included only in the large number of elements is preferentially searched.

According to Theorem 1, the part corresponding to the feature set in the case is a part that can use a solution or a solution without correcting interference. The size of this part can be approximated by the number of elements in the feature set. Therefore, according to the case search method (1), the case where the portion that can be used without correction of interference is large, that is, the case with high usefulness is preferentially searched, so that the first object can be achieved. it can.

According to the inference processing method (1), the inference means refers to the relation information and has features common to the current problem and the case problem, and is included in only one of the current problem and the case problem. The solution or solution part of the case corresponding to the set of features that do not interfere with the feature to be used is used.

From the theorem 1, according to the inference processing method (1), a portion that can be used without correcting interference among cases is selectively used, so that processing for interference correction becomes unnecessary. Thus, the second object can be achieved.

According to the case search method (2), the case search means refers to the dependency information, and is a feature common to the input problem and the problem of the case, and the solution of the corresponding partial problem is only in the case. Is determined and does not interfere with features that are contained only in one of the current problem and the case problem, and interferes with features that correspond to subproblems whose partial solution differs from the current problem to the case A case in which the number of elements of a set of features that do not receive interference and are not affected by features corresponding to a partial problem whose partial solution is obtained only in the current problem is preferentially searched.

According to Theorem 2, the part of the case corresponding to the feature set is a part where the solution or solution can be used without correcting the interference when a part of the solution of the input problem is already determined. The size of this part can be approximated by the number of elements in the feature set. Therefore, according to the case search method (2), the case where the portion that can be used without correction of interference is large, that is, the case with high usefulness is preferentially searched, so that the third object can be achieved. it can.

According to the inference processing method (2), the inference means:
With reference to the dependency information, a feature common to the input problem and the problem of the case, and a solution of the corresponding partial problem is obtained only in the case, and only one of the current problem and the case problem is solved. Without interference from the included features, and without interference from features corresponding to subproblems whose partial solutions differ from the current problem to the case,
In addition, a solution or solution part of a case corresponding to a set of features not affected by a feature corresponding to a partial problem whose partial solution is obtained only in the current problem is used.

From Theorem 2, according to the inference processing method (2), when a part of the solution of the input problem is already determined, a part that can be used without correcting the interference is selectively used.
The process for the interference correction becomes unnecessary. Therefore, the fourth
Can achieve the purpose.

According to the case search method (3), the feature set A in the condition (1) of Theorem 3 should be a feature that the search target case should have, and the feature set B in the condition (2) must not have the feature set B. As a feature, cases can be searched. According to Theorem 3, the solution or solution part of the case corresponding to the feature set A can be used without correcting the interference. Therefore, the features that the search target case should have are initialized to the entire feature set of the input problem, and if the search is repeated while gradually reducing this set, a case in which the feature set A is maximized is found first. Will be. as a result,
The first object can be achieved because the case where the portion that can be used without interference correction is the largest can be searched.

Further, according to the inference processing method (3), a solution or solution of a case corresponding to a feature that the search target case should have is used.

According to the theorem 3, according to the inference processing method (3), a portion that can be used without interference correction among the cases is selectively used, so that processing for interference correction is not required. Thus, the second object can be achieved.

〔Example〕

Hereinafter, a first embodiment of the present invention will be described with reference to FIGS.

FIG. 2 shows a hardware configuration of the present invention. The input / output device 201 inputs case information, interference-related information, a case use rule, a problem, and a request for explaining a case search process, and passes them to the central processing unit 202. Further, it receives and outputs the answer to the question and the description of the case search process from the central processing unit 202.

The central processing unit 202 receives the case information, the interference-related information, and the case use rule from the input / output device 201, and
Store in 03. Further, when a problem is received from the input / output device 201, a similar example of the problem is searched from the case information group similarly stored in the storage device 203 with reference to the interference relation information stored in the storage device 203. Using the case use rules also stored in the storage device 203, the answer of the current problem is obtained based on the answer in the similar example, and the answer is output via the input / output device 201. At this time, an intermediate result obtained in the similar example search process is stored in the storage device 203 as search process information. Further, upon receiving a request for explanation of the case search process from the input / output device 201, referring to the inference process information and the interference relationship information stored in the storage device 203, the description of the similar example search process is provided via the input / output device 201. Output.

The storage device 203 receives case information, interference-related information, case use rules, and search process information from the central processing unit 202 and stores them. In addition, in response to the reading by the central processing unit 202, the central processing unit 202 transfers the case information, the interference-related information, the case use rule, and the search process information to the central processing unit 202.

FIG. 3 shows a functional configuration of the present invention. Block 301 is processing, which includes input / output processing 302, control processing 303, and case search processing.
304, a case use process 305, and an explanation process 306. Block 3
Reference numeral 07 denotes a memory, an interference-related information section 308, a case base 309,
It comprises a search process information section 310 and a case use rule group section 311.

The input / output processing 302 inputs the interference relation information, the case information, and the case use rule, and stores them in the interference relation information unit 308, the case base 309, and the case use rule group unit 311 respectively. Also,
The problem is input and passed to the control process 303. Here, the problem is represented by a set of features. Further, an explanation request of the case search process is input and passed to the explanation processing 306. On the other hand, the input / output processing 302 receives the answer to the question from the control processing 303 and outputs it. Further, a description of the case search process is received from the description process 306 and output.

The control process 303 receives a feature group representing a problem from the input / output process 302. Next, this is passed to the case search processing 304, and the similar example and the feature set of the above problem are received from the case search processing 304. Finally, the similar example and the feature set are passed to the case use process 305, the answer to the problem is received from the case use process 305, and the answer is passed to the input / output process 302.

The case search process 304 receives a feature set representing a problem from the control process 303. Next, by referring to the interference relationship between the features stored in the interference relationship information unit 308, among the features common to the input problem and the case problem, the features that are included in only one of the input problem and the case problem are not affected. A set of those that are not performed (a reference feature set) is obtained, and a case in which the number of elements is maximized is searched from the case base 309, and is set as a similar example.

In the process of searching for similar examples, the case search processing 304 stores a plurality of cases that are candidates for similar examples and the searched cases in the search process information unit 310. Further, for each of the candidate cases, a set of features common to the input problem and the case problem, a set of features included in only one of them, and a set of reference features are stored in the search process information section 310. These pieces of information stored in the search process information section 310 will be referred to as search process information.

Finally, the case search process 304 passes the similar example and the feature set serving as the reference to the control process 303.

The case use process 305 receives a similar example and a reference feature set from the control process 303. Next, a solution to the current problem is obtained using a portion corresponding to a reference feature set among the solutions in the similar example. This process is performed by executing the case use rule group stored in the case use rule group unit 311. Specifically, each rule in the case use rule group is executed in the order of description, that is, in order from the top. Finally, the obtained solution is passed to the control process 303.

The explanation processing 306 receives an explanation request for the case search process from the input / output processing 302. On the other hand, search process information and interference relationship information are received from the search process information unit 310 and the interference relationship information unit 308, respectively, and by using these, information describing the process of similar example search is generated, and input / output processing is performed. Output this via 302.

The interference relationship information unit 308 receives from the input / output unit 302 interference relationship information indicating an interference relationship between features, and stores the information. Further, in response to the reading of the case search processing 304 and the explanation processing 306, the interference-related information is passed to these processings.

The case base 309 receives case information including data representing a problem and data representing a solution to the problem from the input / output unit 302, and stores it. In addition, in response to the reading of the case search process 304, this is passed to the same process.

The search process information section 310 converts the search process information into the case search process 3
Receive from 04 and memorize this. Explanation processing 306
This is passed to the same process in response to the reading of.

The case use rule group unit 311 defines a rule for finding a solution to the current problem based on the solution of the similar example,
And store it. In addition, case use processing 30
In response to the reading of 5, this is passed to the same processing.

FIG. 4 shows a configuration of the interference relation information stored in the interference relation information section 308 of FIG. The interference relation information indicates an interference relation between features and is stored in a table format. This table is realized by a list structure or an array as a data structure.

The left and right sides of the table each contain a set of features. Block 401 is one item of the table, and corresponds to one interference relationship. This indicates that the feature set on the left side of the block 401 interferes with the feature set on the right side. Strictly speaking, when the left-hand side of the <feature 11> problem represented by <feature 12> ~ <feature 1k _1> changes <feature 11 '> the right side <
Features 12 '> ~ <Features 1 m _1' indicates that the change is a problem represented by>. Here, changing the problem means changing the solution or side effect when the problem is solved.

For example, consider an interference relationship in which the features on the left side are <arrangement of console display> and the features on the right side are <arrangement of card reader> and arrangement of katakana printing keyboard card punch. . This relationship is represented by the <card layout> and the <card layout> when the problem of the console display changes. The problem to be addressed, namely the placement of the card reader and the placement of the katakana printing keyboard card punch, will vary. Specifically, it is shown that the arrangement result of the console display affects the arrangement result of the card reader and the arrangement result of the katakana printing keyboard card punch.

FIG. 5 shows a configuration of a case information group stored in the case base 309 of FIG. The case information group is stored in a table format. The left side of the table is data representing a problem, and is a set of features. The right side is data representing the answer to the question. Block 501 corresponds to one piece of case information, and includes <feature 11 ″><feature 12 ″> to <feature 1p.
₁ ”indicates that the answer to the question represented by“> ”is <answer 1>.

FIG. 6 shows a configuration of the case use rules stored in the case use rule group unit 311 in FIG. The left side of the case use rule represents a set of features. The right side represents a process for solving the problem represented by the feature set on the left side. Block 60
1 corresponds to one case use rule, and <
Processing for solving the characteristic 11><feature12> ~ problem represented by <feature 1r _1> indicates that a <process 1>.

FIG. 1 shows the operation of the case search process 304 of FIG.
As described above, the case search process 304 determines a set of features common to the input problem and the case problem that are not interfered with by features included in only one of the input problem and the case problem (reference feature set). Then, select the case that maximizes the number of elements. However, calculating a reference feature set for all cases in the case base 309 may require an enormous amount of processing time when the number of cases is large. Therefore, in this process, first, after searching for examples that are candidates for selection, a reference feature set is obtained for each candidate, and the candidate having the largest number of elements is selected.

Block 101 is a process for searching for a candidate case. Let N be the number of candidates to be searched. If there are many features that are common to the current problem and the case problem, it is expected that there will be many features that will serve as references. Therefore, here, N cases are searched in the order of the most common features between the current problem and the case problem. Substitute the searched case data into variables C1 to Cn.

In this process, first, for each case, a product set of a feature set representing the current problem and a feature set representing the problem of the case is obtained, and then N cases are selected in descending order of the number of elements of the product set. This can be achieved by performing

Block 102 obtains a reference feature set for each case Ci in C1 to Cn. This process will be discussed later.
This will be described in detail with reference to FIG.

The block 103 selects a case in which the number of elements of a feature set serving as a reference is the largest from C1 to Cn, and substitutes the selected case into a variable Cbest. The block 104 returns Cbest and the reference feature set, and ends the processing. The information searched or calculated in each block (candidate case, reference feature set, selected case) is stored in the search process information section 310 in FIG.

FIG. 7 shows a method of calculating a reference feature set in block 102 of FIG. Here, a feature set representing the current problem is FP, and a feature set representing the problem of case C is FC.

Block 701 finds the intersection of FP and FC, that is, the set of features common to the current problem and the case problem, and
Substitute for om. Block 702 calculates Fco from the union of FP and FC.
A set excluding the elements of m, that is, a set of features included only in one of the current problem and the case is obtained and assigned to a variable Fonly.

A block 703 obtains a set of features of Fcom that interfere with features in Fonly, and substitutes the set for a variable Finf. here,
When considering a directed graph in which a feature is a vertex and an edge is an interference relationship between features, this processing can be reduced to a process of obtaining a transitive closure of a relationship in an effective graph. For example, algorithm design and analysis I, Science Inc. 1977) pages 180 to 182
It can be realized by a processing method as shown on the page.

Block 704 obtains a set from Fcom excluding the elements of Finf. This is a feature set serving as a reference for C. This feature set is returned, and the process ends.

A copy of the feature set obtained in each block is stored in the search process information section 310 shown in FIG.

Hereinafter, referring to FIGS. 8 to 10, it will be shown that an appropriate similar example can be searched by this embodiment. Here, it is assumed that various devices are arranged in the computer room. In this problem, the arrangement of each device such as arranging a console display and arranging a central processing unit is a feature of the problem, and the arrangement result of the devices is the answer to the problem. In a real problem, it is necessary to consider factors such as the shape of the room and the position of the columns, but these factors are not considered here in order to make the example easier to understand.

FIG. 8 is an example of an interference relation information table stored in the interference relation information section 308 of FIG. Block 801 is
This shows that the arrangement of the console display affects the arrangement of the card reader.

FIG. 9 is an example of a case group stored in the case base 309 of FIG. Block 901 indicates that the solution to the problem of arranging the console display, card reader, and floppy disk is <arrangement 1>.

FIG. 10 is an example of a case use rule group stored in the case use rule group unit 311 in FIG. Each block is
The rule indicates that "if the device on the left side needs to be arranged, perform the process on the right side." For example, block 1001 represents a rule that if the console display needs to be arranged, execute <processing 1 ′>.

Here, the execution order of the rules has an important meaning. The arrangement of each device has an interference relationship as shown in FIG.
If the arrangement of device A interferes with the arrangement of device B, B
Must be arranged in consideration of the arrangement result. Therefore, it is necessary to execute the arrangement rule of the device A prior to the arrangement rule of the device B. For example, as shown in block 801 of FIG. 8, the placement of the console display interferes with the placement of the card reader. Therefore, it is necessary to execute the block 1001 which is the arrangement rule of the console display before the block 1003 which is the arrangement rule of the card reader.

As described above, the case use process 305 in FIG. 3 executes rules in the order of description, that is, in the order of blocks 1001 to 1006. Therefore, the description order of the blocks 1001 to 1006 reflects the interference relationship shown in FIG.

Below, "Console display, card reader,
In this case, the present embodiment selects an appropriate similar example from the group of cases shown in FIG. 9 using the interference-related information group shown in FIG. 8 for the problem of "place a katakana printing keyboard card punch and a tape reader." To do so. Further, it shows that a solution to the current problem is obtained based on the selected similar example using the case use rule group of FIG.

The above problem is input by the input / output processing 302 in FIG. 3, and is passed to the case search processing 304 via the control processing 303.
Here, the case search process performs the above-described process. First, the problem features are assigned to a variable FP. Here, {arrangement of console display, arrangement of card reader, arrangement of katakana printing keyboard card punch, and tape reader} are substituted. In the following, “disposition of A” will be simply described as “A” unless it is particularly necessary to distinguish between them.

Next, the case search processing 304 and the processing of FIG. 1 are performed. First
The block 101 in the figure narrows the number of candidate cases to N.
Here, it is assumed that N = 3. The number of common features of each case of the blocks 901 to 904 and the current problem are 2, 3, 3, and 4, respectively. Therefore, three cases of blocks 902 to 904 are selected and assigned to variables C1 to C3, respectively.

A block 102 obtains a reference feature set for each of the cases C1 to C3 by the processing of FIG. First, processing is performed for C1. Substitute the feature set of block 902 (console display, katakana printing keyboard card punch, tape reader) into FC. Block 701 finds a common subset of feature set FP of the current problem and feature set FC of the case. , Substitute for Fcom. Here, {console display, katakana printing keyboard card punch, tape reader} is substituted.

A block 702 obtains a feature set excluding the Fcom element from the union of FP and FC, and substitutes it for FPonly. here,
{Card reader} is substituted. Block 703 is the Fco
m, ie, a set of features affected by the card reader among the {console display, katakana printing keyboard card punch, tape reader} is obtained and substituted into Finf.
Here, {Katakana printing keyboard card punch, tape reader} is substituted.

Block 704 removes the Finf element from Fcom and returns the result as a feature set that is the basis of C1. Where,
Returns {console display}.

Next, the block 102 obtains a reference feature set for C2, that is, the case of the block 903, by performing the same processing as described above. Here, a {console display, card reader, katakana printing keyboard card punch} is obtained. In addition, block 102 includes C3 or block 904.
For the case (1), a reference feature set is obtained by the same processing. Here, a {console display, card reader} is obtained.

Next, the block 103 selects a case in which the number of elements of the reference feature set is the largest from C1 to C3. Here, C1-C3
Since the feature sets that serve as the reference for are {console display}, console display, card reader, katakana printing keyboard card punch, console display, card reader, respectively, C2, that is, the case of block 903 is selected. . Block 104 is a block
The case of 903 and the reference feature set {console display, card reader, katakana printing keyboard card puncher} are returned to the control processing 303 in FIG.

The control process 303 uses the partial solution corresponding to the reference feature set {console display, card reader, katakana printing keyboard card punching machine} among the solutions in the case of the block 903, that is, <arrangement 3>, using the tenth solution. Find the solution to the current problem with reference to the rules in the figure. Specifically, <
In the arrangement 3>, the arrangement of a console display, a card reader, and a katakana printing keyboard card punch is used. Other devices, that is, {tape readers}, are arranged using the rule of block 1005 in FIG.

The above is the operation of the embodiment in this example. According to the conventional method, the more the given problem and the case share more characteristics, the more similar the case is to the problem. Therefore, in the case of the above problem, as a similar example, the block 904 in FIG. 9 is used. Is selected. However, block 904 includes a line printer. Line printers interfere with the placement of katakana printing keyboard card punches. Thus, in the answer in block 904, <Arrangement 4>, the arrangement of the katakana printing keyboard card punch is influenced by the liprinter. On the other hand, the current problem does not include a line printer. Therefore, in the current problem, the arrangement of the katakana printing keyboard card punch in <Arrangement 4> cannot be used as it is. Also, the line printer indirectly interferes with the placement of the tape reader via the katakana printing keyboard card punch. Therefore, the arrangement of the tape reader cannot be used as it is in <Arrangement 4>. As described above, even if the case of the block 904 is selected by the conventional case search method, only the arrangement of the console display and the card reader can be used as it is.

On the other hand, according to the present embodiment, the case of the block 903 is searched. In this case, the console display, the card reader, and the katakana printing keyboard card punch are shared with the current problem, but the devices that affect these arrangements are included in this case and the current problem. Not. Therefore, the arrangement of all of these devices in block 903 can be used as is in the current problem. Thus, according to the present embodiment, compared to the conventional case search method,
Search cases with great utility.

As described above, according to the present embodiment, in a case search system that searches a case base that is useful for problem solving from a case base, it is possible to preferentially search for a case that includes a large amount of information that can be directly applied to the current problem. It is possible to prevent wasting time due to retried search and erroneous determination based on inappropriate cases.

Further, according to the conventional case utilization method, the solution of the case is used for all the features common to the current problem and the problem of the case. Therefore, processing for correcting interference between features is required. For example, let us consider a case where the case of the block 903 is not present in the case base in FIG. 9 in the solution example of the computer room layout problem. In this case, the case of block 904 is selected. As a result, in the case utilization process, the arrangement of the katakana printing keyboard card punch and the tape reader which is affected by the arrangement of the line printer is used. In order to use these arrangements, processing for eliminating the influence of the arrangement of the line printer is required. As described above, according to the conventional case utilization method, a process for correcting the influence of interference between features is required.

On the other hand, according to the present embodiment, that is, only the features in the reference feature set, that is, only the features free from the influence of interference are used. For example, even in the case of using the case of the block 904, the arrangement of the katakana printing keyboard card punch and the tape reader is not used, and only the arrangement of the console display and the card reader which are not affected by interference is used. Therefore, processing for correcting interference is not required.

As described above, according to the present embodiment, in the inference system that searches the case base to be a reference for problem solving from the case base and obtains a solution to the problem using the case, the information contained in the case includes the current problem. Information that can be directly applied to the current problem can be selected and used, which simplifies the process of applying case information to the current problem, reduces the cost and time required for system development, Performance and reliability can be improved.

Next, a second embodiment of the present invention will be described with reference to FIGS.

Hardware configuration of this embodiment, configuration of interference-related information,
The configuration of the rule group is the same as that of the first embodiment described with reference to FIGS. The case in this embodiment is similar to the first embodiment described with reference to FIG. 5, but differs in the following points. In the first embodiment, the entire solution of the problem represented by the feature set on the left side is described on the right side. For this,
In the present embodiment, the whole or a part of the solution to the problem is described on the right side. That is, in this embodiment, a case in which only part of the problem is solved is used.

In this embodiment, a case where a part of the solution of the input problem is designated in advance is also included. Thus, the input problem consists of a feature set representing the problem and a solution to a subset thereof.

The functional configuration of the present embodiment is similar to that of the first embodiment described with reference to FIG. 3, but the case retrieval process and the case use process are performed by reflecting the input problem and the difference between the cases described above. different.

In the case search process, a case in which the number of elements of a feature set (a reference feature set) that satisfies the following conditions (1) to (5) is maximized is searched from the case base as a similar example.

(1) A set of features common to the current problem and the case problem.

(2) The solution of the corresponding partial problem is not found in the current problem, but is found in the case.

(3) It is not interfered by features included only in one of the current problem and the case problem.

(4) It is a common subproblem of the current problem and the problem of the case, and a solution is found in both of them, and both solutions are not interfered by a feature corresponding to a subproblem having a different solution.

(5) There is no interference from features corresponding to sub-problems for which no solution has been found in the case.

Further, in the case search process, the searched case and a feature set serving as a reference thereof are passed to the case use process.

In the case use process, a solution to the current problem is obtained based on a partial solution corresponding to a reference feature set among solutions of similar examples. This process is performed by executing the case use rule group stored in the case use rule group unit 311.

Hereinafter, the operation of the case search process will be described with reference to FIGS. 11 and 12.

In this process, similar to the case search process in the first embodiment, first, a case that is a candidate for selection is searched,
For each candidate, a reference feature set is obtained, and the candidate having the largest number of elements is selected.

Block 1101 in FIG. 11 is a process for searching for a candidate case. Let N be the number of candidates to be searched. Here, of the above conditions (1) to (5), (1),
In consideration of only (2), N cases are searched in the descending order of the number of features satisfying these, and variables C1 to Cn are substituted.

A block 1102 obtains a reference feature set for each of C1 to Cn. This processing will be described later in detail with reference to FIG.

Blocks 103 and 104 perform the same processing as in the first embodiment described with reference to FIG. 1, return the case where the number of reference features is the maximum, and return the reference feature set, and end the processing. I do.

Next, the operation of the block 1102 in FIG. 11 will be described with reference to FIG. Here, among the features of the current problem, the set of features for which the solution of the corresponding partial problem is predetermined is FPD, the feature set for which the solution is undetermined is FPN, and the feature of the case problem A set of features whose solution is predetermined is FC
D, Let the feature set whose solution is undecided be FCN.

Blocks 701 and 702 obtain a feature set common to the current problem and the problem of the case, a feature set included in only one of them, by the same processing as in the first embodiment described in FIG.
Substitute the variables Fcom and Fonly, respectively.

A block 1203 obtains a feature set corresponding to a sub-problem whose solution differs between the current problem and the case problem from the common set of the current problem and the case problem, and substitutes it into a variable Fdif.

Block 1204 includes a feature set that is included in only one of the current problem and the case problem, a feature set corresponding to a subproblem that has a different solution in the current problem and the case problem, and a solution in the case problem. The union of the feature sets corresponding to the uncompleted subproblems is obtained and substituted into the variable Fdisag.

A block 1205 obtains a set of Fcoms that are interfered by the features in Fdisasg, and substitutes the set for the variable Finf '. This processing can be realized by a method similar to that of the block 703 in FIG.

A block 1206 returns a feature set obtained by removing the elements of Finf 'from Fcom, and ends the process.

According to the above processing, the condition (1) is satisfied by the block 701, and the condition (2) is satisfied by the blocks 1202 and 1206. In addition, blocks 1203-1207
Thereby, (3) to (5) are satisfied. Therefore, the reference feature set can be obtained by the processing of FIG. 12, and the case where the number of elements of the reference feature set is the largest can be searched by the processing of FIG.

According to the above-mentioned theorem (2), the solution of the partial problem corresponding to the reference feature set among the cases can be used without correcting the interference. According to the present embodiment, a case in which such a feature set is maximized can be searched.

As described above, according to the present embodiment, in the case search system for searching a case to be a reference for problem solving from the case base, when a part of the solution of the current problem is specified in advance, Even if some of them are not found, you can preferentially search for cases that include a lot of information that can be applied to the current problem as it is, preventing time-consuming re-searching and incorrect decisions based on inappropriate cases can do.

Furthermore, the case use process of the present embodiment is performed by
The solution of the partial problem corresponding to the reference feature set is used.

Therefore, according to the present embodiment, a part of the current solution of a problem is specified in advance in an inference system that searches a case that is a reference for solving a problem from a case base and obtains a solution to the problem using the case. Even if the solution of the case is not found, the information included in the case can be selected and used as it is for the current problem. Process can be simplified, system development costs and time are reduced, system performance,
Reliability can be improved.

Hereinafter, a third embodiment of the present invention will be described with reference to FIGS. 13 and 14.

Hardware configuration of this embodiment, configuration of interference-related information,
The configuration of the case group and the configuration of the rule group are the same as those in the first embodiment described with reference to FIGS. 2, 4, 5, and 6. The functional configuration is
Similar to the first embodiment described with reference to FIG.
Case search processing and case use processing are different.

The case search process refers to the dependency relationship information to obtain features that the search target case should have and features that the search target case should not have, and uses these to search the case from the case base. Further, the searched case and the features that the search target case should have are passed to the case use process.

The case use process receives the case and the features that the search target case should have from the case search process, and uses the part of the case solution corresponding to the partial problem represented by this feature to find the solution to the current problem. Ask.

FIG. 13 shows the operation of the case search process in this embodiment. In this process, the input argument is the variable T to which the interference-related information is substituted and the variable FI to which the feature set to be held in the similar example is substituted. As an initial value, a part related to the input problem in the interference relation information in the interference relation information section is substituted for T. More specifically, a table including interference-related information included in the problem in which all the features on both sides are input is substituted. Further, the entire feature set of the input question is substituted into the FI.

When a similar example is searched using this FI, the feature set shared by the question and the similar example is FI, that is, the entire feature set of the input question. Therefore, all the feature sets that affect the problem corresponding to the feature set shared by the problem and the similar example and are included in the input feature set of the problem are also included in the feature set of the similar example. From this, (1) of the above two conditions of Theorem 3 is satisfied.

Hereinafter, the processing content will be described. A block 1301 obtains a feature set that the similar example should not have, and substitutes the variable FE. Here, Proc is a procedure for obtaining a feature set that a similar example should not have, using a feature set FI that the similar example should have. Specifically, a feature set that affects the problem corresponding to the FI and is not included in the feature set of the input problem is obtained. When a similar example is searched using such an FE, a feature set that affects the problem corresponding to the feature set shared by the problem and the similar example, and is not included in the feature set of the input problem, is a similar example. Will not be included. From this, (2) of the two conditions of Theorem 3 is satisfied.

The processing contents of the proc will be described later in detail with reference to FIG.

Block 1302 searches the case base for cases having FI and no FE. As described above, an appropriate case can be retrieved by using FI and FE. Block 1303 is
It is determined whether there is a corresponding case. If there is a case, the block 1304 passes the case to the control processing 303 in FIG. 3, and then stops the processing. If there was no case,
Proceed to block 1305.

A block 1305 generates a list of the features in question that do not appear on the left side of T, and substitutes this for a variable L. L
Is a list of features to be excluded from FI.

Block 1306 determines whether L is empty; if L is empty, there are no features to remove from the FI, ie, a new FI
Cannot be obtained, the process is at a dead end, and the process returns. If L is not empty, block 1307
Proceed to.

A block 1307 extracts one feature from L and substitutes it for a variable F. Block 1308 removes the retrieved features from L. Block 1309 removes F from FI and returns the result to FI '
Substitute for FI ′ is a new candidate for a feature set that a similar example should have. When a similar example is searched using FI ′, the feature set shared by the problem and the similar example is FI ′. here,
The feature F excluded from FI does not appear on the left side of the interference relation information T. Therefore, F is not a feature affecting the subproblem corresponding to FI '. Therefore, all the features that affect the partial problem corresponding to FI ′ and are included in the feature set of the input problem are included in FI ′, that is, the feature set of the searched similar example. . From this, two conditions (1) for the above similar example to be appropriate are satisfied.

A block 1310 substitutes a table obtained by removing F from T for a variable T '. A block 1311 executes this process recursively with FI' and T 'as arguments instead of FI and T. . If a similar example is found in the course of this recursive execution, block 1304 outputs the similar example and stops the operation. If a similar example is not found, the process returns via the block 1306 and returns to the block 1311. As a result, the process proceeds to block 1306. In this case, block
From 1306 to 1310, except for the features different from the previous one from FI,
Again in block 1311 the case will be searched. When all the possibilities are exhausted, in the top-level processing, the result is Yes in block 1306, and the entire processing returns to the control processing 303 in FIG.

FIG. 14 shows the operation of the procedure Proc in the block 1301 in FIG. The Proc obtains a feature set that the similar example does not have, using the feature set FI that the similar example should have as an argument. Block 1401 substitutes FI for the variable S.

A block 1402 obtains an interference relation information group in the interference relation information section 307 shown in FIG. 3 which is not related to the current problem. Specifically, it finds interference-related information that includes a feature that is not the current problem on the left side or the right side, generates a table including these, and substitutes it for the variable T1.

A block 1403 extracts interference relation information whose right side is included in S of T1, generates a table including these, and substitutes it into a variable T1 ′. A block 1404 calculates a set of features included in the left side of T1 ′. Is obtained and substituted into the variable S '.
A feature set that directly affects the partial problem corresponding to the FI is substituted for S ′.

It should be noted that even if there is no interference-related information including the feature and the problem, the feature may indirectly affect the answer to the problem. For example, if there is interference relationship information “A affects B” and interference relationship information “B affects C”, an interference relationship “A affects C” is established.

Therefore, it is necessary to find a feature set that affects the partial problem corresponding to the FI, including such indirect effects. For that purpose, the processes of the blocks 1403 and 1404 may be repeated until S converges.

Block 1405 determines convergence. If the convergence is achieved, that is, if there are no features not included in S in S ′, block 1406 determines that the features in FI that are excluded from S are set as a feature set that similar examples should not have in FIG. Returns to the main routine. If the convergence does not occur, that is, if there is a feature not included in S in S ′, block 1407 sets the union of S and S ′ to S again, and returns to block 1403.

As described above, according to this embodiment, a case that satisfies the conditions (1) and (2) of Theorem 3 is searched. Here, the feature set assigned to the variable FI at the end of the processing is
This corresponds to the feature set A of the condition (1). As mentioned earlier, FI
Is initialized to the entire feature set of the current problem. A is
The initial value of FI is the largest possible set of A because it is part of a feature set common to the current problem and the case problem. The set assigned to FI is reduced one element at a time by block 1309, and each time an attempt is made to retrieve a case. As a result, a case in which A is maximized is searched. According to Theorem 3, the part corresponding to the partial problem represented by A in the solution of the case is the part that can be used without correcting the interference.

As described above, according to the present embodiment, in the case search system that searches for cases that are useful for problem solving from the case base, it is possible to preferentially search for cases that include a lot of information that can be applied to the current problem without any change. It is possible to prevent wasting time due to redoing and erroneous determination based on inappropriate cases.

Furthermore, the case use process of the present embodiment is performed by
The solution of the subproblem represented by the feature set to which the variable FI is assigned is used.

Therefore, according to this embodiment, in the inference system that searches the case base that is helpful for problem solving from the case base and uses that case to solve the problem, the information contained in the case is applied to the current problem as it is. The ability to select and use available information can simplify the process of applying case information to current problems, reduce system development costs and time, improve system performance and reliability. Performance can be improved.

Next, a fourth embodiment of the present invention will be described with reference to FIG.

The hardware configuration of this embodiment, the configuration of the interference-related information,
The configuration of the case group and the configuration of the rule group are the same as those in the first embodiment described with reference to FIGS. 2, 4, 5, and 6. The functional configuration is similar to that of the third embodiment, but the operation of the case search process is different.

FIG. 15 shows the operation of the case search process of this embodiment. In the case search process of this embodiment, first, a part related to the current problem is extracted from the dependency information table and assigned to a variable T, as in the case search process of the first embodiment. Further, the total number of the feature in question is substituted into a variable N.
Next, the processing in FIG. 15 is performed.

The block 1501 in FIG. 15 initializes the variable I. This variable is a counter for selecting one feature from the feature set in question. Block 1502 determines whether all features have been selected. In the case of Yes, there is no similar example, and the routine returns. If no, proceed to block 1503.

Block 1503 increments I by one. Block 1504
Selects the I-th feature from the feature set in question, and substitutes a set having the I-th feature into a variable S1. Block 1505
Substitutes a table in which the right side of T includes the dependency information included in S1 into T2. A block 1506 assigns a set of features included in the left side of T2 to a variable S2. Through the above processing, a feature set that directly affects the partial problem corresponding to S1 is substituted into S2.

A block 1507 assigns the union of S1 and S2 to S1 '.
Here, to find the feature set that indirectly affects S1,
Blocks 1505-1507 are repeated until S1, S1 'converge. Block 1508 determines convergence. If converged,
That is, if S1 = S1 ', the flow proceeds to block 1510. If not converged, block 1509 replaces S1 with S1 'and then returns to block 1505.

As described above, S1 includes all the features included in the current problem among the features affecting the partial problem corresponding to S1. This S1 is a feature set that a similar example should have. When a similar example is searched using S1, the feature set shared by the problem with the similar example is S1. Therefore, features that affect the partial problem corresponding to the feature set shared by the problem and the similar example and are included in the problem are all included in the similar example. Then, among the conditions of Theorem 3, (1)
Is satisfied.

A block 1510 calls the procedure Proc in FIG. 14 using the S1 as an argument, obtains a feature set that the similar example should not have, and substitutes it for a variable FE. As described in the third embodiment, when a similar example is searched using this feature set,
(2) among the conditions of Theorem 3 is satisfied.

Block 1511 searches for a case having S1 and no FE from the case base 309 in FIG. 3.
It is determined whether a similar example has been found. If found, block 1513 uses a similar example in FIG.
Pass it to 303 and end the search processing. If not found, the process returns to block 1502, and the above processing is repeated using another subset.

As described above, according to the present embodiment, a case that satisfies the condition of Theorem 3 is searched. Here, the feature set assigned to the variable S1 at the end of the process corresponds to the feature set A of the condition (1). The case use process uses a part corresponding to the partial problem represented by the feature set in the solution of the case. According to Theorem 3, this part can be used without interference correction.

As described above, according to the present embodiment, in the inference system that searches the case base to be a reference for problem solving from the case base and obtains a solution to the problem using the case, the information included in the case includes the current problem as it is. Since applicable information can be selected and used, the process of applying case information to the current problem can be simplified, system development costs and time can be reduced, system performance, Reliability can be improved.

〔The invention's effect〕

In a case search system that searches for cases that serve as references for problem solving from a case base, cases that contain a lot of information that can be applied to the current problem can be preferentially searched. It is possible to prevent an erroneous decision based on a case.

In addition, in the inference system that searches cases from the case base to be a reference for problem solving and uses the cases to find the solution to the problem, from the information contained in the cases, select information that can be applied directly to the current problem. Because it can be used, it is possible to simplify the process of applying case information to the current problem, reduce the cost and time of system development, and improve the performance and reliability of the system. it can.

[Brief description of the drawings]

FIG. 1 is a flowchart showing an outline operation of a case search process in the first embodiment of the present invention, FIG. 2 is a hardware configuration diagram of the first embodiment of the present invention, and FIG. FIG. 4 is a functional configuration diagram of the first embodiment of the present invention, FIG. 4 is a configuration diagram of an interference relation information table in the first embodiment of the present invention, and FIG. 5 is case information in the first embodiment of the present invention. FIG. 6 is a configuration diagram of a table, FIG. 6 is a configuration diagram of a case use rule of the first embodiment of the present invention, and FIG. 7 is a flowchart showing a detailed operation of a case search process in the first embodiment of the present invention. , FIG. 8 is a diagram showing an example of a dependency relationship information table,
FIG. 9 is a diagram showing an example of a case information table, FIG. 10 is a diagram showing an example of a case use rule group, and FIG. 11 is a schematic operation of a case search process in the second embodiment of the present invention. FIG. 12 is a flowchart showing the detailed operation of the case search process in the second embodiment of the present invention, and FIG. 13 is a schematic operation of the case search process in the third embodiment of the present invention. FIG. 14 is a flowchart showing a detailed operation of the case search process in the third embodiment of the present invention, and FIG. 15 is a flowchart showing a schematic operation of the case search process in the fourth embodiment of the present invention. The flowchart shown in FIG. 16 is an explanatory diagram of Theorem 1 used in the method of the present invention.

──────────────────────────────────────────────────続 き Continuation of the front page (56) References “CoquityScievece”, Vol. 13, No. 4 (1989-4) p. 507-549 (58) Field surveyed (Int. Cl. ⁶ , DB name) G06F 9/44

Claims (9)

(57) [Claims] 1. A case similar to an inputted problem or a sub-problem of the problem, using a case base storing data representing a group of problems defined in advance and data representing a solution or solution of the problem. In a case search method in an information processing system that searches for a case including a problem from the case base, when a problem is input, referring to interference relationship information representing an interference relationship between features of the predefined problem group, A case search method, wherein a feature to be included in the case to be searched is obtained based on the referred interference-related information, and the case is searched using the obtained feature. 2. A case similar to an inputted problem or a sub-problem of the problem, using a case base storing data representing a group of problems defined in advance and data representing a solution or solution of the problem. In a case search method in an information processing system that searches for a case including a problem from the case base, when a problem is input, referring to interference relationship information representing an interference relationship between features of the predefined problem group, A case search method, wherein a feature that the case to be searched for must not have is obtained based on the referred interference relation information, and the case is searched using the obtained feature. 3. A case similar to an input problem or a sub-problem of the problem, using a case base storing data representing a group of problems defined in advance and sequence information including data representing an answer or solution of the problem. In a case search method in an information processing system that searches for a case including a problem from the case base, when a problem is input, referring to interference relationship information representing an interference relationship between features of the predefined problem group, The first feature that the search target case should have based on the referenced interference relationship information is obtained, and the second feature that the search target case must not have based on the referenced interference relationship information is determined. A case search method, wherein the case is searched using the obtained first feature and the second feature. 4. A case similar to an input problem or a sub-problem of the problem, using a case base storing data representing a group of problems defined in advance and data representing a solution or solution of the problem. In a case search method in an information processing system for searching for a case including a problem from the case base, interference relation information representing an interference relation of a feature of the predefined problem group is stored in an interference relation information unit, Is input, the interference-related information is referred to, and is a feature common to the input problem and the problem in the case, and is included in only one of the input problem and the problem in the case. A case search method for searching for the case based on a set of features that do not receive interference from the feature to be searched. 5. A case similar to an inputted problem or a sub-problem of the problem, using a case base storing data representing a group of problems defined in advance and data representing a solution or solution of the problem. In a case search method in an information processing system for searching for a case including a problem from the case base, interference relation information representing an interference relation of a feature of the predefined problem group is stored in an interference relation information unit, Is input, the interference-related information is referred to, (1) a feature common to the input problem and the problem in the case, and (2) a solution of a partial problem corresponding to the feature is a case. (3) the feature is not subject to interference from a feature included in only one of the input problem and the problem in the case; (4) the feature is Case retrieval method characterized in that solutions in does not receive interference from the feature corresponding to subproblems unsolicited, based on a set of conditions are satisfied features, retrieves the case. 6. A case similar to an input problem or a sub-problem of the problem, using a case base storing data representing a group of problems defined in advance and data representing a solution or solution of the problem. In a case search method in an information processing system for searching for a case including a problem from the case base, interference relationship information indicating an interference relationship between variables and constraints of the predefined problem group is stored in an interference relationship information unit. When a problem is input, the interference-related information is referred to, (1) a variable common to the input problem and the problem in the case, and (2) the value of the variable is a problem in the case. (3) the value of the variable is not interfered with by variables and constraints contained only in one of the input problem and the problem in the case; (4) the value of the variable But case retrieval method characterized by not interfered from the variable unsolicited in case, based on a set of conditions are satisfied variables, retrieves the case. 7. An input problem or a problem similar to a sub-problem of a problem is stored in a case base storing a case including data representing a group of problems defined in advance and data representing an answer or solution of the problem. An inference processing method in an information processing system for searching for a case including the case from the case base and inferring a solution to the input problem using the searched case, wherein an interference relationship of a feature of the predefined problem group is provided. Is stored in the interference relation information section, and when a case including a problem similar to the input problem is searched, the interference relation information is referred to, and (1) the input problem And (2) the feature is not interfered by a feature included only in one of the input problem and the problem in the case. Using the answer or solution in case corresponding to the set of features that satisfy the inference processing method characterized by inferring the solution of the problem. 8. A case similar to an input problem or a sub-problem of the input problem, using a case base that stores a case including data representing a problem group defined in advance and data representing an answer or solution of the problem. An inference processing method in an information processing system for searching for a case including the case from the case base and inferring a solution to the input problem using the searched case, wherein an interference relationship of a feature of the predefined problem group is provided. Is stored in the interference relation information section, and when a case including a problem similar to the input problem is searched, the interference relation information is referred to, and (1) the input problem And (2) a solution of a partial problem corresponding to the feature is obtained for the problem in the case, and (3) the feature is the input (4) The feature satisfies the following conditions: (4) the feature does not interfere with the feature of the subproblem in which no solution is found in the case. An inference processing method comprising inferring a solution to the problem using a solution or solution in a case corresponding to a set of features. 9. A method for accessing an input problem or partial information of the problem, using a case base storing data representing a group of problems defined in advance and data representing a solution or solution to the problem, Or an information processing system that searches the case base for a case in which the own station includes a problem similar to the problem for a response packet indicating the information itself, and infers a solution to the input problem using the searched case. In the inference processing method described above, interference relation information indicating interference relation between variables and constraints of the predefined problem group is stored in an interference relation information section, and a case including a problem similar to the input problem is stored. When searched by the case search means, the interference-related information is referred to: (1) a variable common to the input problem and the problem in the case; ) The value of the variable is determined for the problem in the case, (3) the variable is not interfered with by variables and constraints contained only in one of the input problem and the problem in the case, (4) the variable The value of is not interfered with by the undetermined variables in the case, using the values of the variables in the case corresponding to the set of variables that satisfy the condition of inferring the solution of the problem. Inference processing method.