Information processing method, information processing device, and information processing program

Abstract

This information processing device: creates an optimal solution of an objective function having a parameter; presents the created solution to a user; acquires first correction information for correcting the solution instructed by the user in natural language; inputs the acquired first correction information to generative artificial intelligence (AI); acquires, from the generative AI, second correction information indicating a correction policy for the parameter; determines the correction amount for the parameter on the basis of the acquired second correction information; and changes the current parameter on the basis of the determined correction amount for the parameter.

Description

Information processing method, information processing device, and information processing program

[0001] This disclosure relates to a technique for creating an optimal solution to an objective function having parameters.

[0002] For example, the technology described in Patent Document 1 discloses determining parameters for modifying knowledge in production scheduling based on the user's intended planning strategy, formulating a production schedule based on the production scheduling procedure and parameters, and displaying the formulated result as a diagram.

[0003] However, with the conventional techniques described above, it was difficult to easily change the parameters of the objective function, and further improvements were needed.

[0004] Japanese Patent Application Publication No. 6-290165

[0005] This disclosure was made to solve the above-mentioned problems and aims to provide a technique that allows for easy modification of the parameters of the objective function.

[0006] An information processing method according to one aspect of the present disclosure is an information processing method performed by a computer, which includes: creating an optimal solution to an objective function having parameters; presenting the created solution to a user; obtaining first modification information for modifying the solution, instructed by the user in natural language; inputting the obtained first modification information to a generating artificial intelligence (AI) and obtaining second modification information from the generating AI indicating a policy for modifying the parameters; determining the amount of modification of the parameters based on the obtained second modification information; and changing the current parameters based on the determined amount of modification of the parameters.

[0007] According to this disclosure, the parameters of the objective function can be easily changed.

[0008] This figure shows the configuration of the information processing system according to this embodiment. This is a flowchart for explaining information processing by the information processing device in the embodiment of this disclosure. This figure shows an example of a prompt initially acquired by the prompt acquisition unit in this embodiment. This figure shows an example of answer information initially acquired by the answer acquisition unit in this embodiment. This is a first flowchart for explaining the parameter correction amount determination process in step S7 of Figure 2. This is a second flowchart for explaining the parameter correction amount determination process in step S7 of Figure 2. This is a schematic diagram for explaining the first correction condition in this embodiment. This is a schematic diagram for explaining the second correction condition in this embodiment. This figure shows an example of a prompt acquired a second time by the prompt acquisition unit in this embodiment. This figure shows an example of answer information acquired a second time by the answer acquisition unit in this embodiment. This is a schematic diagram for explaining the third correction condition in modified example 1 of this embodiment. This is a schematic diagram for explaining the fourth correction condition in modified example 2 of this embodiment.

[0009] (Knowledge forming the basis of this disclosure) Conventionally, the conditions that a manufacturing plan for producing a product must satisfy are given as constraints in the objective function. The optimization algorithm then updates the solution to minimize the value of the objective function, ultimately generating a solution that satisfies the conditions as much as possible. In this case, the objective function has parameters that weight the constraints, but it is difficult for non-expert users to determine the values ​​of these parameters.

[0010] In the technology described in Patent Document 1 above, it is difficult to easily change the parameters of the objective function, and in particular, it does not disclose how to create a solution by repeatedly giving user instructions to modify the solution and changing the parameters.

[0011] To address the above challenges, the following technologies are disclosed.

[0012] (1) An information processing method relating to one aspect of the present disclosure is an information processing method performed by a computer, which includes: creating an optimal solution to an objective function having parameters; presenting the created solution to a user; obtaining first modification information for modifying the solution, instructed by the user in natural language; inputting the obtained first modification information to a generating artificial intelligence (AI) and obtaining second modification information from the generating AI indicating a policy for modifying the parameters; determining the amount of modification of the parameters based on the obtained second modification information; and changing the current parameters based on the determined amount of modification of the parameters.

[0013] In this configuration, first correction information for modifying the solution, instructed by the user in natural language, is input to the generating artificial intelligence (AI). Second correction information indicating the parameter modification policy is obtained from the generating AI, and the amount of parameter modification is determined based on the obtained second correction information. This makes it easy to change the parameters of the objective function. Furthermore, by repeatedly providing correction instructions for the solution and changing the parameters, even users without knowledge of the parameters of the objective function can derive a more optimal solution.

[0014] (2) In the information processing method described in (1) above, the solution may be a manufacturing plan for producing a product. With this configuration, a manufacturing plan for producing a product can be created.

[0015] (3) In the information processing method described in (1) or (2) above, the second modification information may include state information indicating that the parameter is increased, decreased, or not changed, and degree information indicating the degree to which the parameter is increased or decreased.

[0016] In this configuration, state information indicating whether a parameter is increased, decreased, or left unchanged, and degree information indicating the degree to which the parameter is increased or decreased are generated by the generating AI. Based on the qualitative state information and degree information, the quantitative amount of parameter modification can be determined.

[0017] (4) The information processing method described in (3) above further includes storing the second modification information and the modification amount in memory each time the parameter is changed, and the determination of the modification amount may include obtaining a history of the second modification information and the modification amount from the memory, and determining the modification amount of the parameter in the current case such that the absolute value of the modification amount of the parameter in the current case is smaller than the absolute value of the modification amount of the parameter in the previous case, if the state information included in the previous second modification information is an increase and the state information included in the current second modification information is a decrease, or if the state information included in the previous second modification information is a decrease and the state information included in the current second modification information is an increase.

[0018] With this configuration, the amount of parameter modification for the current parameter is determined so that the current parameter modification does not undo the previous parameter modification, thus allowing for the determination of a parameter modification amount that aligns with the user's modification intentions.

[0019] (5) The information processing method described in (3) above further includes storing the second modification information and the modification amount in memory each time the parameter is changed, and the determination of the modification amount may include obtaining a history of the second modification information and the modification amount from the memory, and determining the modification amount of the parameter in the current case such that the absolute value of the modification amount of the parameter in the current case is the same as the absolute value of the modification amount of the parameter in the previous case.

[0020] With this configuration, the amount of parameter modification for the current parameter is determined based on the amount of modification for the previous parameter, so it is possible to determine the amount of parameter modification that aligns with the user's modification intentions.

[0021] (6) The information processing method described in (3) above further includes storing the second modification information and the modification amount in memory each time the parameter is changed, and the determination of the modification amount may include obtaining the history of the second modification information and the modification amount from the memory, and determining the modification amount of the parameter for the current time such that the absolute value of the modification amount of the parameter for the current time is smaller than the absolute value of the sum of the modification amounts from the first time to the previous time, if the sign of the sum of the modification amounts from the first time to the previous time is positive and the state information included in the current second modification information is decreasing, or if the sign of the sum of the modification amounts from the first time to the previous time is negative and the state information included in the current second modification information is increasing.

[0022] With this configuration, the amount of parameter modification for the current modification is determined in such a way that it does not undo all previous parameter modifications from the initial modification onward. This allows for determining the amount of parameter modification that aligns with the user's modification intentions.

[0023] (7) In the information processing method described in (3) above, the second modification information further includes modification instruction information indicating whether or not the modification of the parameter is explicitly indicated in the first modification information, and further includes storing the second modification information and the modification amount in memory each time the parameter is changed, and the determination of the modification amount is obtained from the memory a history of the second modification information and the modification amount, and the state information included in the previous second modification information is an increase, the state information included in the current second modification information is a decrease, the degree information included in the previous second modification information is the same as the degree information included in the current second modification information, the modification instruction information included in the previous second modification information indicates that the modification of the parameter is not explicit, and the current 2. If the correction instruction information included in the correction information indicates that the correction of the parameter is explicit, or if the state information included in the previous second correction information is a decrease, the state information included in the current second correction information is an increase, the degree information included in the previous second correction information is the same as the degree information included in the current second correction information, the correction instruction information included in the previous second correction information indicates that the correction of the parameter is not explicit, and the correction instruction information included in the current second correction information indicates that the correction of the parameter is explicit, then the correction amount of the parameter in the current case may be determined such that the absolute value of the correction amount of the parameter in the current case is the same as the absolute value of the correction amount of the parameter in the previous case.

[0024] With this configuration, if a parameter is changed without explicit user instruction, and then an explicit instruction is given to revert that change, the amount of the current parameter modification will be determined in such a way that it cancels out the previous modification, thus negating modifications made against the user's intentions.

[0025] In addition, the present disclosure can be realized not only as an information processing method for executing the characteristic processing as described above, but also as an information processing apparatus including a characteristic configuration corresponding to the characteristic processing executed by the information processing method. Further, it can be realized as a computer program for causing a computer to execute the characteristic processing included in such an information processing method. Therefore, in the following other aspects, the same effects as those of the above information processing method can be achieved.

[0026] (8) An information processing apparatus according to another aspect of the present disclosure is an information processing apparatus including a processor, wherein the processor creates an optimal solution of an objective function having parameters, presents the created solution to a user, obtains first correction information for correcting the solution, which is instructed by the user in natural language, inputs the obtained first correction information to a generative artificial intelligence (AI), obtains second correction information indicating a correction policy of the parameters from the generative AI, determines a correction amount of the parameters based on the obtained second correction information, and changes the current parameters based on the determined correction amount of the parameters.

[0027] (9) An information processing program according to another aspect of the present disclosure causes a computer to function so as to create an optimal solution of an objective function having parameters, present the created solution to a user, obtain first correction information for correcting the solution, which is instructed by the user in natural language, input the obtained first correction information to a generative artificial intelligence (AI), obtain second correction information indicating a correction policy of the parameters from the generative AI, determine a correction amount of the parameters based on the obtained second correction information, and change the current parameters based on the determined correction amount of the parameters.

[0028] (10) A non-temporary computer-readable recording medium according to another aspect of the present disclosure records the information processing program described in (9) above.

[0029] The embodiments of the present disclosure will be described below with reference to the accompanying drawings. Note that the embodiments described below are all specific examples of the present disclosure. The numerical values, shapes, components, steps, order of steps, etc. shown in the following embodiments are merely examples and are not intended to limit the present disclosure. In addition, among the components in the following embodiments, the components not described in the independent claims indicating the highest-level concept are described as optional components. Also, in all embodiments, the respective contents can be combined with each other.

[0030] (Embodiment) FIG. 1 is a diagram showing the configuration of an information processing system 1 according to the present embodiment.

[0031] The information processing system 1 includes an information processing apparatus 10, a display device 20, and an input device 30.

[0032] The display device 20 is, for example, a liquid crystal display device and displays information presented to the user. The display device 20 is connected to the information processing apparatus 10 so as to be communicable with each other by wire or wirelessly. Note that the display device 20 may be connected to the information processing apparatus 10 so as to be communicable with each other via a network. The network is a local area network or a wide area network.

[0033] The input device 30 is, for example, a keyboard, a mouse, or a touch panel, and receives input of information by the user. The input device 30 is connected to the information processing apparatus 10 so as to be communicable with each other by wire or wirelessly. Also, the input device 30 may be connected to the information processing apparatus 10 so as to be communicable with each other via a network. The network is a local area network or a wide area network.

[0034] The information processing apparatus 10 includes a parameter storage unit 101, a production plan creation unit 102, a production plan presentation unit 103, a prompt acquisition unit 104, a prompt input unit 105, a generation AI (Artificial Intelligence) 106, an answer acquisition unit 107, a correction amount determination unit 108, a correction history storage unit 109, and a parameter change unit 110.

[0035] The information processing device 10 includes, for example, a computer system comprising a control program, a processing circuit such as a processor or logic circuit that executes the control program, and a recording device such as an internal memory or an accessible external memory that stores the control program. The information processing device 10 may be implemented, for example, by a hardware implementation using a processing circuit, by the execution of a software program held in memory by the processing circuit or distributed from an external server, or by a combination of these hardware and software implementations.

[0036] Furthermore, the information processing device 10 may be a server or an edge computer. If the information processing device 10 is a server, the input device 30 and the display device 20 may be terminals such as a personal computer, smartphone, or tablet computer. Also, the configuration of the information processing device 10 may be implemented in the input device 30 or the display device 20. Furthermore, the configuration of the information processing device 10 may be distributed among the input device 30, the display device 20, and the server.

[0037] The parameter storage unit 101 stores the parameters of the objective function. Parameters are also called hyperparameters. For example, the objective function has a first parameter p1 to a sixth parameter p6. The first parameter p1 is a parameter that emphasizes minimizing total power consumption. The second parameter p2 is a parameter that emphasizes minimizing operating time. The third parameter p3 is a parameter that emphasizes minimizing the workload of employees. The fourth parameter p4 is a parameter that emphasizes demand response (DR). The fifth parameter p5 is a parameter that emphasizes meeting deadlines. The sixth parameter p6 is a parameter that emphasizes adhering to the sequence of manufacturing tasks with dependencies.

[0038] The manufacturing plan creation unit 102 creates the optimal solution to an objective function with parameters. The solution is a manufacturing plan for producing the product. The manufacturing plan creation unit 102 creates the manufacturing plan. The manufacturing plan creation unit 102 creates the manufacturing plan that minimizes the value of the objective function through optimization processing. For example, the manufacturing plan represents the start time and completion time of manufacturing for each of several products.

[0039] The manufacturing plan presentation unit 103 presents the manufacturing plan (solution) created by the manufacturing plan creation unit 102 to the user. The manufacturing plan presentation unit 103 displays the manufacturing plan created by the manufacturing plan creation unit 102 on the display device 20. The display device 20 displays the manufacturing plan.

[0040] The prompt acquisition unit 104 acquires prompts (first modification information) for modifying the manufacturing plan (solution) instructed by the user in natural language. The prompts are text data and include tasks described for the generation AI 106 to execute. The input device 30 accepts prompt input from the user. The user checks the manufacturing plan displayed on the display device 20 and inputs prompts for modifying the manufacturing plan into the input device 30. The prompt acquisition unit 104 acquires the prompts entered by the user from the input device 30.

[0041] The input device 30 accepts text data, but this disclosure is not limited to this. The input device 30 may be a microphone and may accept voice data. In this case, the prompt acquisition unit 104 converts the voice data input by the input device 30 into text data.

[0042] The prompt input unit 105 inputs the prompt acquired by the prompt acquisition unit 104 to the generation AI 106.

[0043] The generating AI 106 generates a text that is a response to a given prompt. The generating AI 106 is, for example, a large-scale language model such as Chat-GPT, but is not particularly limited thereto. The generating AI 106 executes the task described in the input prompt. In this embodiment, the task is to present a policy for modifying parameters. The generating AI 106 generates response information (second modification information) that indicates a policy for modifying parameters.

[0044] As a preprocessing step, the generating AI 106 may acquire knowledge about the parameters through search-enhanced generation (RAG). Furthermore, the prompt may contain knowledge about the parameters, and the generating AI 106 may acquire this parameter knowledge from the prompt.

[0045] Furthermore, although the information processing device 10 is equipped with a generation AI 106 in this embodiment, the disclosure is not particularly limited thereto, and the information processing device 10 may be connected to a server having the function of the generation AI 106 via a network so as to be able to communicate. In this case, the information processing device 10 may send a prompt to the server and receive response information from the server. The server may be an on-premise server or a cloud server.

[0046] The response acquisition unit 107 acquires response information (second modification information) indicating the parameter modification policy from the generating AI 106. The response information includes state information indicating either increase, decrease, or leave the parameter unchanged, and degree information indicating the degree to which the parameter is increased or decreased. The degree information includes "large" as the first degree, "medium" as the second degree which is lower than the first degree, and "small" as the third degree which is lower than the second degree. The degree information is expressed in three stages.

[0047] The correction amount determination unit 108 determines the amount of parameter correction based on the response information (second correction information) obtained by the response acquisition unit 107. If it is the first time the parameter has been changed, the correction amount determination unit 108 determines whether to increase or decrease the parameter based on the state information and randomly determines the correction amount. Alternatively, if it is the first time the parameter has been changed, the correction amount determination unit 108 may determine the correction amount to a predetermined amount.

[0048] The correction amount determination unit 108 stores the response information (second correction information) and the correction amount in the correction history storage unit 109 each time the parameter is changed.

[0049] The revision history storage unit 109 stores the response information (second revision information) and the history of the amount of revision.

[0050] If the parameter has been changed for the second time or more, the correction amount determination unit 108 obtains the response information (second correction information) and the history of the correction amount from the correction history storage unit 109. Then, if the state information included in the previous response information is an increase and the state information included in the current response information is a decrease, or if the state information included in the previous response information is a decrease and the state information included in the current response information is an increase, the correction amount determination unit 108 determines the correction amount for the current parameter so that the absolute value of the correction amount for the current parameter is smaller than the absolute value of the correction amount for the previous parameter.

[0051] For example, if the degree of correction in the current correction is "large", the correction amount determination unit 108 will determine the current correction amount to be 4 / 5 of the previous correction amount. Also, for example, if the degree of correction in the current correction is "medium", the correction amount determination unit 108 will determine the current correction amount to be 3 / 5 of the previous correction amount. Also, for example, if the degree of correction in the current correction is "small", the correction amount determination unit 108 will determine the current correction amount to be 2 / 5 of the previous correction amount.

[0052] Furthermore, the correction amount determination unit 108 determines the correction amount for the current parameter if the state information included in the previous response information is an increase, the state information included in the current response information is an increase, and the degree information included in the previous response information is the same as the degree information included in the current response information, or if the state information included in the previous response information is a decrease, the state information included in the current response information is a decrease, and the degree information included in the previous response information is the same as the degree information included in the current response information, so that the absolute value of the correction amount for the current parameter is the same as the absolute value of the correction amount for the previous parameter.

[0053] For example, if the degree of the previous correction was "large" and the degree of the current correction is "medium", the correction amount determination unit 108 will determine the current correction amount to be 4 / 5 of the previous correction amount. Also, for example, if the degree of the previous correction was "large" and the degree of the current correction is "small", the correction amount determination unit 108 will determine the current correction amount to be 2 / 5 of the previous correction amount. Also, for example, if the degree of the previous correction was "medium" and the degree of the current correction is "large", the correction amount determination unit 108 will determine the current correction amount to be 1.4 times the previous correction amount. Also, for example, if the degree of the previous correction was "medium" and the degree of the current correction is "small", the correction amount determination unit 108 will determine the current correction amount to be 3 / 5 of the previous correction amount. Furthermore, for example, if the degree of the previous correction was "small" and the degree of the current correction is "large", the correction amount determination unit 108 will determine the current correction amount to be 1.8 times the previous correction amount. Also, for example, if the degree of the previous correction was "small" and the degree of the current correction is "medium", the correction amount determination unit 108 will determine the current correction amount to be 1.4 times the previous correction amount.

[0054] The parameter modification unit 110 modifies the current parameters based on the parameter modification amount determined by the modification amount determination unit 108. The parameter modification unit 110 modifies the parameters stored in the parameter storage unit 101.

[0055] After the parameters are changed by the parameter change unit 110, the manufacturing plan creation unit 102 creates a new manufacturing plan using the changed parameters. The manufacturing plan presentation unit 103 presents the new manufacturing plan created by the manufacturing plan creation unit 102 to the user. The user reviews the presented manufacturing plan and creates a prompt if they wish to further modify it. The entire process, from obtaining a prompt to presenting the manufacturing plan, is repeated until the user finishes creating the manufacturing plan.

[0056] Next, the information processing by the information processing device 10 in the embodiment of this disclosure will be described.

[0057] Figure 2 is a flowchart illustrating the information processing performed by the information processing device 10 in the embodiment of this disclosure.

[0058] First, in step S1, the manufacturing plan creation unit 102 creates a manufacturing plan using the parameters stored in the parameter storage unit 101. The parameter storage unit 101 has pre-stored initial values ​​for the parameters.

[0059] Next, in step S2, the manufacturing plan presentation unit 103 presents the manufacturing plan created by the manufacturing plan creation unit 102 to the user. The manufacturing plan presentation unit 103 outputs the manufacturing plan created by the manufacturing plan creation unit 102 to the display device 20, and the display device 20 displays the manufacturing plan.

[0060] Next, in step S3, the manufacturing plan presentation unit 103 determines whether or not to terminate the creation of the manufacturing plan. The display device 20 displays an end button for terminating the creation of the manufacturing plan and a prompt input button for entering a prompt on the display screen. If the end button is pressed by the user, the manufacturing plan presentation unit 103 determines that the creation of the manufacturing plan is to be terminated. On the other hand, if the prompt input button is pressed by the user, the manufacturing plan presentation unit 103 determines that the creation of the manufacturing plan is not to be terminated.

[0061] If it is determined at this point that the creation of the manufacturing plan is complete (YES in step S3), then the information processing ends. In other words, if the user determines that the presented manufacturing plan is reasonable, the information processing ends.

[0062] On the other hand, if it is determined that the creation of the manufacturing plan is not to be completed (NO in step S3), in step S4, the prompt acquisition unit 104 acquires a prompt to modify the manufacturing plan, which is instructed by the user in natural language. The input device 30 accepts the prompt input from the user.

[0063] Figure 3 shows an example of the prompt 201 initially acquired by the prompt acquisition unit 104 in this embodiment.

[0064] The prompt 201 shown in Figure 3 includes a sentence 211 that instructs the role of the generating AI 106, a sentence 212 that indicates the user's requests for the manufacturing plan, i.e., modifications to the manufacturing plan, a sentence 213 that indicates knowledge about parameters, and a sentence 214 that indicates the task to be performed by the generating AI 106.

[0065] Document 211 describes that the generating AI 106 is a "specialist in manufacturing plan creation" and is given the role of "adjusting multiple hyperparameters of the manufacturing planning system to meet the user's requests regarding the manufacturing plan." Document 214 describes that the task of the generating AI 106 is to "suggest one of the following modifications for each of the values ​​of the hyperparameters of the manufacturing planning system: 'increase the value,' 'do not change the value,' or 'decrease the value,'" and, "if the value is changed, specify the extent of the change using 'large,' 'medium,' or 'small.'"

[0066] Furthermore, document 212 describes a user's request regarding the manufacturing plan: "Since the time spent on DR can be used for employee breaks, we would like to create a plan with a greater DR effect." Document 213 also describes knowledge regarding the first parameter p1 to the sixth parameter p6.

[0067] Returning to Figure 2, in step S5, the prompt input unit 105 inputs the prompt acquired by the prompt acquisition unit 104 to the generation AI 106.

[0068] Next, in step S6, the response acquisition unit 107 acquires response information indicating the parameter modification policy from the generating AI 106.

[0069] Figure 4 shows an example of the response information 202 initially acquired by the response acquisition unit 107 in this embodiment.

[0070] When the generating AI 106 receives the prompt 201 shown in Figure 3, it generates and outputs the response information 202 shown in Figure 4.

[0071] As shown in Figure 4, the response information 202 includes state information 221 for each of the first to sixth parameters p1 to p6, and degree information 222 for each of the first to sixth parameters p1 to p6. The response information 202 also includes the reason for modifying or not modifying each of the first to sixth parameters p1 to p6.

[0072] For example, the state information for the first parameter p1, the second parameter p2, the third parameter p3, and the sixth parameter p6 is "do not change the value", the state information for the fourth parameter p4 is "increase", the degree information for the fourth parameter p4 is "large", the state information for the fifth parameter p5 is "decrease", and the degree information for the fifth parameter p5 is "medium".

[0073] Returning to Figure 2, in step S7, the correction amount determination unit 108 determines the parameter correction amount based on the response information obtained by the response acquisition unit 107.

[0074] Figure 5 is a first flowchart illustrating the parameter correction amount determination process in step S7 of Figure 2, and Figure 6 is a second flowchart illustrating the parameter correction amount determination process in step S7 of Figure 2.

[0075] First, in step S21, the correction amount determination unit 108 obtains the number of parameter changes. The number of parameter changes is stored in a memory (not shown) and is incremented each time a parameter is changed. The number of parameter changes is also reset when the creation of the manufacturing plan is completed.

[0076] Next, in step S22, the correction amount determination unit 108 obtains state information and degree information of the nth parameter pn (where n is a natural number) from the response information.

[0077] Next, in step S23, the correction amount determination unit 108 determines whether or not it is the first time the parameter has been changed. If it is determined that it is the first time the parameter has been changed (YES in step S23), then in step S24, the correction amount determination unit 108 determines the correction amount of the parameter according to the state information. If the state information indicates an increase or decrease, the correction amount determination unit 108 determines the correction amount to a random value. On the other hand, if the state information indicates no change, the correction amount determination unit 108 determines the correction amount to zero.

[0078] On the other hand, if it is determined that this is not the first time the parameter has been changed (NO in step S23), in step S25, the correction amount determination unit 108 determines whether the current state information is "do not change the value". If it is determined that the current state information is "do not change the value" (YES in step S25), in step S26, the correction amount determination unit 108 determines the correction amount to be zero.

[0079] On the other hand, if the current status information is determined to be not "do not change the value" (NO in step S25), in step S27, the correction amount determination unit 108 obtains the response information and the history of the correction amount from the correction history storage unit 109.

[0080] Next, in step S28, the correction amount determination unit 108 determines whether the previous state information was "do not change the value". If it is determined that the previous state information was "do not change the value" (YES in step S28), then in step S29, the correction amount determination unit 108 determines the correction amount according to the degree information. For example, an upper limit for the correction amount may be predetermined. If the degree of correction this time is "large", the correction amount determination unit 108 may determine the correction amount to the upper limit. If the degree of correction this time is "medium", the correction amount determination unit 108 may determine the correction amount to 2 / 3 of the upper limit. If the degree of correction this time is "small", the correction amount determination unit 108 may determine the correction amount to 1 / 3 of the upper limit.

[0081] On the other hand, if it is determined that the previous state information is not "do not change the value" (NO in step S28), in step S30, the correction amount determination unit 108 determines whether the current state information is different from the previous state information. The conditions under which the current state information is different from the previous state information are that the current state information is "increase" and the previous state information is "decrease", and that the current state information is "decrease" and the previous state information is "increase". If it is determined that the current state information is different from the previous state information (YES in step S30), in step S31, the correction amount determination unit 108 determines the correction amount so as to satisfy the first correction condition.

[0082] Figure 7 is a schematic diagram illustrating the first modification condition in this embodiment.

[0083] In Figure 7, the vertical axis shows the value of the nth parameter pn, and the horizontal axis shows the number of times the parameter has been changed. Arrow 41 indicates the modification of the nth parameter pn in the previous (Nth) step, and arrow 51 indicates the modification of the nth parameter pn in the current (N+1) step. The state information for the previous (N) step is "increase", and the degree information for the previous (N) step is "large". The state information for the current (N+1) step is "decrease", and the degree information for the current (N+1) step is "large".

[0084] When the state information of the previous time (the Nth time) is "increase" and the state information of this time (the (N + 1)th time) is "decrease", the correction amount determination unit 108 determines the correction amount of this time (the (N + 1)th time) so as to satisfy the first correction condition.

[0085] The first correction condition is represented by the following formula (1).

[0086] |Δd ｐ Ｎ |>|Δd ｐ Ｎ＋１ |... (1)

[0087] In the above formula (1), Δd ｐ Ｎ is the correction amount of the previous time (the Nth time), and Δd ｐ Ｎ＋１ is the correction amount of this time (the (N + 1)th time). That is, the correction amount determination unit 108 determines the correction amount Δd ｐ Ｎ＋１ of this time (the (N + 1)th time) such that the absolute value of the correction amount of this time (the (N + 1)th time) is smaller than the absolute value of the correction amount of the previous time (the Nth time).

[0088] Returning to FIG. 5, when it is determined that the state information of this time is the same as the state information of the previous time (NO in step S30), in step S32, the correction amount determination unit 108 determines whether the degree information of this time is the same as the degree information of the previous time. Here, when it is determined that the degree information of this time is the same as the degree information of the previous time (YES in step S32), in step S33, the correction amount determination unit 108 determines the correction amount so as to satisfy the second correction condition.

[0089] FIG. 8 is a schematic diagram for explaining the second correction condition in the present embodiment.

[0090] In FIG. 8, the vertical axis represents the value of the nth parameter pn, and the horizontal axis represents the number of times of parameter change. Arrow 41 indicates the correction of the nth parameter pn of the previous time (the Nth time), and arrow 51 indicates the correction of the nth parameter pn of this time (the (N + 1)th time). The state information of the previous time (the Nth time) is "increase", the degree information of the previous time (the Nth time) is "medium", the state information of this time (the (N + 1)th time) is "increase", and the degree information of this time (the (N + 1)th time) is "medium".

[0091] If the state information for the previous (Nth) time was "increase", the degree information for the previous (Nth) time was "medium", and the state information for the current (N+1) time is "increase", and the degree information for the current (N+1) time is "medium", then the correction amount determination unit 108 determines the correction amount for the current (N+1) time so as to satisfy the second correction condition.

[0092] The second modification condition is expressed by the following formula (2).

[0093] |Δd ｐ Ｎ |=|Δd ｐ Ｎ＋１ | ... (2)

[0094] In equation (2) above, Δd ｐ Ｎ Δd is the amount of correction from the previous (Nth) time. ｐ Ｎ＋１ This is the correction amount for the current (N+1)th time. In other words, the correction amount determination unit 108 determines the correction amount Δd for the current (N+1)th time so that the absolute value of the correction amount for the current (N+1)th time is the same as the absolute value of the correction amount for the previous (N)th time. ｐ Ｎ＋１ To decide.

[0095] Returning to Figure 6, if it is determined that the current degree information is different from the previous degree information (NO in step S32), in step S34, the correction amount determination unit 108 determines the correction amount according to the degree information.

[0096] For example, if the previous degree information was "large" and the current degree information is "medium," the correction amount determination unit 108 may determine the current correction amount to be 2 / 3 of the previous correction amount. Also, for example, if the previous degree information was "large" and the current degree information is "small," the correction amount determination unit 108 may determine the current correction amount to be 1 / 3 of the previous correction amount.

[0097] Furthermore, for example, if the previous degree information was "medium" and the current degree information is "large," the correction amount determination unit 108 may determine the current correction amount to be 1.4 times the previous correction amount. Also, for example, if the previous degree information was "medium" and the current degree information is "small," the correction amount determination unit 108 may determine the current correction amount to be 1 / 3 of the previous correction amount.

[0098] Furthermore, for example, if the previous degree information was "small" and the current degree information is "large," the correction amount determination unit 108 may determine the current correction amount to be 1.4 times the previous correction amount. Also, for example, if the previous degree information was "small" and the current degree information is "medium," the correction amount determination unit 108 may determine the current correction amount to be 1.2 times the previous correction amount.

[0099] Next, in step S35, the correction amount determination unit 108 stores the current response information and the correction amount in the correction history storage unit 109.

[0100] Next, in step S36, the correction amount determination unit 108 determines whether or not the correction amounts for all parameters have been determined. If it is determined that the correction amounts for all parameters have not been determined (NO in step S36), the process returns to step S22, and the correction amount determination unit 108 obtains the state information and degree information of the (n+1)th parameter pn+1 (where n is a natural number) from the response information. The process from steps S22 to S36 is repeated until the correction amounts for all parameters have been determined.

[0101] If it is determined that the amount of correction for all parameters has been determined (YES in step S36), the parameter correction amount determination process ends.

[0102] Returning to Figure 2, in step S8, the parameter modification unit 110 modifies the current parameter based on the parameter modification amount determined by the modification amount determination unit 108. For example, if the state information is "increase", the parameter modification unit 110 adds the parameter modification amount determined by the modification amount determination unit 108 to the current parameter value. Also, for example, if the state information is "decrease", the parameter modification unit 110 subtracts the parameter modification amount determined by the modification amount determination unit 108 from the current parameter value.

[0103] When the processing in step S8 is completed, the process returns to step S1, and the manufacturing plan creation unit 102 creates a manufacturing plan again using the modified parameters stored in the parameter storage unit 101. The processing in steps S1 to S8 is repeated until it is determined in step S3 that the creation of the manufacturing plan is complete.

[0104] Figure 9 shows an example of a prompt 203 acquired a second time by the prompt acquisition unit 104 in this embodiment.

[0105] Based on the response information 202 shown in Figure 4, the amount of parameter modification is determined, and the parameters are changed based on the determined modification amount. Then, a manufacturing plan is created using the modified parameters, and the created manufacturing plan is presented to the user. The user reviews the manufacturing plan and creates a prompt if there are any points that need to be modified in the manufacturing plan. In this embodiment, there are points that need to be modified in the manufacturing plan, so the prompt 203 shown in Figure 9 is created.

[0106] The prompt 203 shown in Figure 9 includes a sentence 211 that instructs the role of the generating AI 106, a sentence 212 that indicates the user's requests for the manufacturing plan, i.e., the modifications to the manufacturing plan, and a sentence 214 that indicates the task to be performed by the generating AI 106.

[0107] Text 211 states that the generating AI 106 is given the role of "adjusting the hyperparameters to meet further user requests based on the manufacturing plan creation results derived from the hyperparameter modifications."

[0108] Furthermore, text 214 states that the task of the generating AI 106 is the same as the task of the previously entered prompt.

[0109] Furthermore, document 212 describes an additional request from the user regarding the manufacturing plan: "We would like to improve the DR effect a little more. However, we would like to avoid an excessive increase in employee tasks, such as additional work that occurs during the switchover."

[0110] Figure 10 shows an example of the response information 204 acquired a second time by the response acquisition unit 107 in this embodiment.

[0111] When the generating AI 106 receives the prompt 203 shown in Figure 9, it generates and outputs the response information 204 shown in Figure 10.

[0112] As shown in Figure 10, the response information 204 includes state information 221 for each of the first to sixth parameters p1 to p6, and degree information 222 for each of the first to sixth parameters p6. The response information 204 also includes the reason for modifying or not modifying each of the first to sixth parameters p1 to p6.

[0113] For example, the state information for the first parameter p1, the second parameter p2, the fifth parameter p5, and the sixth parameter p6 is "do not change the value", the state information for the third parameter p3 is "increase", the degree information for the third parameter p3 is "small", the state information for the fourth parameter p4 is "increase", and the degree information for the fourth parameter p4 is "small".

[0114] Based on the response information 204 shown in Figure 10, the amount of parameter modification is determined, and the parameters are changed based on the determined modification amount. Then, a manufacturing plan is created using the modified parameters, and the created manufacturing plan is presented to the user. In this way, a more optimal manufacturing plan is created by repeating the series of processes of creating a manufacturing plan, obtaining prompts, obtaining response information, determining the amount of parameter modification, changing parameters, and creating a manufacturing plan.

[0115] In this way, prompts (first correction information) for modifying the manufacturing plan (solution), instructed by the user in natural language, are input to the generating artificial intelligence (AI), and response information (second correction information) indicating the parameter modification policy is obtained from the generating AI. Based on the obtained response information (second correction information), the amount of parameter modification is determined, so that the parameters of the objective function can be easily changed. Furthermore, by repeatedly giving instructions to modify the manufacturing plan (solution) by the user and changing the parameters, even a user without knowledge of the parameters of the objective function can derive a more optimal manufacturing plan (solution).

[0116] (Modification 1) In Modification 1 of this embodiment, the correction amount determination unit 108 may determine the correction amount for the current parameter so as to satisfy the third correction condition when the sign of the sum of the correction amounts from the first to the previous is positive and the state information included in the current response information (second correction information) is decreasing, or when the sign of the sum of the correction amounts from the first to the previous is negative and the state information included in the current response information (second correction information) is increasing. The third correction condition is that the absolute value of the correction amount for the current parameter is smaller than the absolute value of the sum of the correction amounts from the first to the previous.

[0117] Figure 11 is a schematic diagram illustrating the third modification condition in Modification 1 of this embodiment.

[0118] In Figure 11, the vertical axis represents the value of the nth parameter pn, and the horizontal axis represents the number of parameter changes. Arrow 61 indicates the first modification of the nth parameter pn, arrow 62 indicates the second modification of the nth parameter pn, arrow 63 indicates the previous (Nth) modification of the nth parameter pn, and arrow 64 indicates the current (N+1) modification of the nth parameter pn. The sign of the sum of the modification amounts from the first to the previous (Nth) modification is positive. The state information for the current (N+1) modification is "decrease," and the degree information for the current (N+1) modification is "medium."

[0119] If the sign of the sum of the correction amounts from the first time to the previous time (Nth time) is positive, and the state information for the current time (N+1 time) is "decrease", the correction amount determination unit 108 determines the correction amount for the current time (N+1 time) so as to satisfy the third correction condition.

[0120] The third modification condition is expressed by the following formula (3).

[0121]

[0122] In the above equation (3), d ｐ ｋ Δd is the amount of the k-th correction. ｐ Ｎ＋１ This is the correction amount for the current (N+1)th time. In other words, the correction amount determination unit 108 determines the absolute value of the correction amount for the current (N+1)th time |Δd ｐ Ｎ＋１ The amount of correction for the current (N+1)th time is determined such that | is smaller than the absolute value of the sum of the correction amounts from the first time to the previous (Nth) time.

[0123] The correction amount determination unit 108 may determine the correction amount so as to satisfy the third correction condition of Modification 1 instead of the first correction condition of this embodiment, or it may determine the correction amount so as to satisfy both the first correction condition of this embodiment and the third correction condition of Modification 1.

[0124] (Modification 2) In modification 2 of this embodiment, the response information (second modification information) may further include modification instruction information indicating whether or not the parameter modification is explicitly indicated in the prompt (first modification information). The modification amount determination unit 108 may then determine the modification amount of the parameter in the current response information (second modification information) to satisfy the fourth modification condition if the state information included in the previous response information (second modification information) is an increase, the state information included in the current response information (second modification information) is a decrease, the degree information included in the previous response information (second modification information) is the same as the degree information included in the current response information (second modification information), the modification instruction information included in the previous response information (second modification information) indicates that the parameter modification is not explicit, and the modification instruction information included in the current response information (second modification information) indicates that the parameter modification is explicit. Furthermore, the correction amount determination unit 108 may determine the correction amount for the current parameter so as to satisfy the fourth correction condition if the state information contained in the previous response information (second correction information) is a decrease, the state information contained in the current response information (second correction information) is an increase, the degree information contained in the previous second correction information is the same as the degree information contained in the current response information (second correction information), the correction instruction information contained in the previous response information (second correction information) indicates that the parameter correction is not explicit, and the correction instruction information contained in the current response information (second correction information) indicates that the parameter correction is explicit. The fourth correction condition is that the absolute value of the correction amount for the current parameter is the same as the absolute value of the correction amount for the previous parameter.

[0125] Figure 12 is a schematic diagram illustrating the fourth modification condition in a modified example 2 of this embodiment.

[0126] In Figure 12, the vertical axis represents the value of the nth parameter pn, and the horizontal axis represents the number of parameter changes. Arrow 41 indicates the modification of the nth parameter pn in the previous (Nth) change, and arrow 51 indicates the modification of the nth parameter pn in the current (N+1) change. The state information for the previous (N) change is "decrease", and the degree information for the previous (N) change is "medium", while the state information for the current (N+1) change is "increase", and the degree information for the current (N+1) change is "medium". Furthermore, the modification instruction information for the previous (N) change indicates that the parameter modification is not explicit, while the modification instruction information for the current (N+1) change indicates that the parameter modification is explicit.

[0127] The generating AI 106 determines whether the parameter modification is explicitly indicated in the prompt and generates response information that further includes modification instruction information indicating whether the parameter modification is explicit or not. For example, the prompt may further include the task: "Also, if you change a value, please specify whether you are changing a parameter that is explicitly indicated in the user's request or a parameter that is not explicitly indicated in the user's request, using either 'explicit' or 'implicit'." The modification instruction information is expressed as either "explicit" or "implicit."

[0128] If the state information for the previous (Nth) time was "decrease", the degree information for the previous (Nth) time was "medium", the state information for the current (N+1) time was "increase", the degree information for the current (N+1) time was "medium", the correction instruction information for the previous (Nth) time indicated that the parameter correction was not explicit (implicit), and the correction instruction information for the current (N+1) time indicated that the parameter correction was explicit (explicit), then the correction amount determination unit 108 determines the correction amount for the current (N+1) time so as to satisfy the fourth correction condition.

[0129] The fourth modification condition is expressed by the following formula (4).

[0130] |Δd ｐ Ｎ |=|Δd ｐ Ｎ＋１ | ... (4)

[0131] In equation (4) above, Δd ｐＮ Δd is the amount of correction from the previous (Nth) time. ｐ Ｎ＋１ This is the correction amount for the current (N+1)th time. In other words, the correction amount determination unit 108 determines the correction amount Δd for the current (N+1)th time so that the absolute value of the correction amount for the current (N+1)th time is the same as the absolute value of the correction amount for the previous (N)th time. ｐ Ｎ＋１ To decide.

[0132] Thus, if a parameter not explicitly stated in the previous prompt entered by the user is changed, and an instruction to explicitly modify this parameter is entered in the current prompt, the modification amount determination unit 108 determines the current modification amount such that the absolute value of the current modification amount is the same as the absolute value of the previous modification amount, even if the previous state information was an increase (decrease) and the current state information is a decrease (increase).

[0133] Next, other modifications of this embodiment will be described.

[0134] In this embodiment, the memory may store a combination of the parameter modification amount and the result of creating the manufacturing plan (e.g., overtime hours). For example, if a prompt is entered that includes a user request such as "I want the increase in overtime hours from the manufacturing plan without DR adjustments to be 25% or less," the modification amount determination unit 108 may read the parameter modification amount associated with the overtime hours with an increase of 25% or less from the memory. This makes it possible to determine the parameter modification amount more accurately and reduces the number of times prompts are created.

[0135] Alternatively, a learning model may be created using the parameter adjustment amounts and the results of the manufacturing plan creation as training data. In this case, when the desired manufacturing plan is input to the learning model, the learning model may estimate and output the parameter adjustment amounts.

[0136] Furthermore, the memory may store the natural language entered by the user at the prompt and the final determined parameter modification amount in association with each other. If the prompt contains the same natural language as the natural language stored in the memory, the modification amount determination unit 108 may read the parameter modification amount associated with the natural language from the memory.

[0137] In this embodiment, the information processing device 10 may further include an answer presentation unit that presents the answer information acquired by the answer acquisition unit 107 to the user. The answer presentation unit may display the answer information on the display device 20. The display device 20 may display the answer information.

[0138] Furthermore, while a manufacturing plan is created in this embodiment, the disclosure is not limited thereto, and a delivery plan for delivering goods or people may also be created.

[0139] Furthermore, some or all of the functions of the apparatus according to the embodiments of this disclosure may be realized by a processor such as a CPU executing a program.

[0140] Furthermore, all figures used above are illustrative examples provided to illustrate this disclosure, and this disclosure is not limited to these illustrative figures.

[0141] Furthermore, the order in which the steps shown in the flowchart above are performed is illustrative for the purpose of specifically illustrating this disclosure, and other orders are acceptable as long as similar effects are achieved. Also, some of the steps above may be performed simultaneously (in parallel) with other steps.

[0142] The technology described herein is useful as a technique for creating an optimal solution to an objective function with parameters, because it allows for easy modification of the parameters of the objective function.

Claims

1. An information processing method performed by a computer, comprising: creating an optimal solution to an objective function having parameters; presenting the created solution to a user; obtaining first modification information for modifying the solution, instructed by the user in natural language; inputting the obtained first modification information to a generating artificial intelligence (AI) and obtaining second modification information from the generating AI indicating a policy for modifying the parameters; determining the amount of modification for the parameters based on the obtained second modification information; and changing the current parameters based on the determined amount of modification for the parameters.

2. The information processing method according to claim 1, wherein the solution is a manufacturing plan for manufacturing a product.

3. The information processing method according to claim 1 or 2, wherein the second modification information includes state information indicating that the parameter is increased, decreased, or left unchanged, and degree information indicating the degree to which the parameter is increased or decreased.

4. The information processing method according to claim 3, further comprising storing the second modification information and the modification amount in memory each time the parameter is changed, wherein the determination of the modification amount includes: obtaining a history of the second modification information and the modification amount from the memory; and determining the modification amount of the parameter in the current case such that the absolute value of the modification amount of the parameter in the current case is smaller than the absolute value of the modification amount of the parameter in the previous case, if the state information included in the previous second modification information is an increase and the state information included in the current second modification information is a decrease, or if the state information included in the previous second modification information is a decrease and the state information included in the current second modification information is an increase.

5. The information processing method according to claim 3, further comprising storing the second modification information and the modification amount in memory each time the parameter is changed, wherein the determination of the modification amount includes: obtaining a history of the second modification information and the modification amount from the memory; and determining the modification amount of the parameter in the current case such that the absolute value of the modification amount of the parameter in the current case is the same as the absolute value of the modification amount of the parameter in the previous case, if the state information included in the previous second modification information is an increase, the state information included in the current second modification information is an increase, and the degree information included in the previous second modification information is the same as the degree information included in the current second modification information, or if the state information included in the previous second modification information is a decrease, the state information included in the current second modification information is a decrease, and the degree information included in the previous second modification information is the same as the degree information included in the current second modification information.

6. The information processing method according to claim 3, further comprising storing the second modification information and the modification amount in memory each time the parameter is changed, wherein the determination of the modification amount includes: obtaining a history of the second modification information and the modification amount from the memory; and determining the modification amount for the parameter in the current case such that the absolute value of the modification amount for the parameter in the current case is smaller than the absolute value of the sum of the modification amounts from the first to the previous case, if the sign of the sum of the modification amounts from the first to the previous case is positive and the state information included in the current second modification information is decreasing, or if the sign of the sum of the modification amounts from the first to the previous case is negative and the state information included in the current second modification information is increasing.

7. The second modification information further includes modification instruction information indicating whether or not the modification of the parameter is explicitly indicated in the first modification information, and further includes storing the second modification information and the modification amount in memory each time the parameter is changed, and the determination of the modification amount includes obtaining the history of the second modification information and the modification amount from the memory, The information processing method according to claim 3, comprising: determining the amount of modification for the parameter in the current case such that the absolute value of the amount of modification for the parameter in the current case is the same as the absolute value of the amount of modification for the parameter in the previous case; when the state information included in the previous second modification information is an increase, the state information included in the current second modification information is a decrease, the degree information included in the previous second modification information is the same as the degree information included in the current second modification information, the modification instruction information included in the previous second modification information indicates that the modification of the parameter is not explicit, and the modification instruction information included in the current second modification information indicates that the modification of the parameter is explicit; or when the state information included in the previous second modification information is a decrease, the state information included in the current second modification information is an increase, the degree information included in the previous second modification information is the same as the degree information included in the current second modification information, the modification instruction information included in the previous second modification information indicates that the modification of the parameter is not explicit, and the modification instruction information included in the current second modification information indicates that the modification of the parameter is explicit.

8. An information processing device comprising a processor, wherein the processor creates an optimal solution to an objective function having parameters, presents the created solution to a user, obtains first modification information for modifying the solution as instructed by the user in natural language, inputs the obtained first modification information to a generating artificial intelligence (AI), obtains second modification information from the generating AI indicating a policy for modifying the parameters, determines the amount of modification of the parameters based on the obtained second modification information, and modifies the current parameters based on the determined amount of modification of the parameters.

9. An information processing program that creates an optimal solution to an objective function having parameters, presents the created solution to a user, obtains first modification information for modifying the solution as instructed by the user in natural language, inputs the obtained first modification information to a generating artificial intelligence (AI), obtains second modification information from the generating AI indicating a policy for modifying the parameters, determines the amount of modification for the parameters based on the obtained second modification information, and causes a computer to function to change the current parameters based on the determined amount of modification for the parameters.

Images