Information processing device, information processing method, and program

Abstract

To provide a technique that can generate explanations for data. [Solution] An information processing device according to one embodiment of the present disclosure has a selection unit that selects one or more types as the type of explanation to be generated for data related to equipment, a data acquisition unit that acquires the data according to the one or more selected types, an explanation generation unit that generates one or more explanations corresponding to each of the one or more types based on the acquired data and a generation AI realized by a machine learning model including a large-scale language model, and an integration unit that creates a first integrated explanation representing an explanation that integrates the one or more explanations corresponding to each of the one or more types.

Description

[Technical field]

[0001] The present disclosure relates to an information processing device, an information processing method, and a program. [Background technology]

[0002] There are known methods for determining an optimal operation plan for equipment such as various plants (e.g., power plants, steel plants, chemical plants, petroleum plants, energy plants, etc.) by solving optimization problems. For example, Patent Document 1 discloses a method for calculating an optimal operation plan for a heat source system. Also, for example, Patent Document 2 discloses a method for creating an optimal production plan for production equipment. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 6810597 [Patent Document 2] Patent No. 7349608 Summary of the Invention [Problem to be solved by the invention]

[0004] However, since it is generally difficult to confirm the correctness of an operation plan obtained by solving an optimization problem, a technology is required to generate an explanation for data representing the operation plan. This is not limited to data representing an operation plan obtained by solving an optimization problem, but also applies to data representing operation results, for example.

[0005] The present disclosure has been made in consideration of the above points, and aims to provide a technology that can generate explanations for data. [Means for solving the problem]

[0006] An information processing device according to one embodiment of the present disclosure has a selection unit that selects one or more types as the type of explanation to be generated for data related to equipment, a data acquisition unit that acquires the data according to the one or more selected types, an explanation generation unit that generates one or more explanations corresponding to each of the one or more types based on the acquired data and a generation AI realized by a machine learning model including a large-scale language model, and an integration unit that creates a first integrated explanation representing an explanation that integrates the one or more explanations corresponding to each of the one or more types. Effect of the Invention

[0007] Explanations can be generated for the data. [Brief description of the drawings]

[0008] [Figure 1] FIG. 2 is a diagram illustrating an example of a configuration of a target facility. [Diagram 2] FIG. 2 is a diagram illustrating an example of a hardware configuration of the data explanation device according to the present embodiment. [Diagram 3] FIG. 2 is a diagram illustrating an example of a functional configuration of the data explanation device according to the present embodiment. [Figure 4] 13 is a flowchart illustrating an example of a data explanation process according to the embodiment. [Diagram 5] FIG. 13 is a diagram showing an example of driving performance data in an array format. [Figure 6] FIG. 11 is a diagram showing an example of optimal operation plan data in an array format. [Figure 7] 13 is a flowchart illustrating an example of an explanation generation process according to the present embodiment. [Figure 8] FIG. 13 is a diagram showing an example (part 1) of a prompt. [Figure 9] FIG. 13 is a diagram showing an example (part 2) of a prompt. [Figure 10] FIG. 13 is a diagram showing an example (part 3) of a prompt. [Figure 11] FIG. 13 is a diagram showing an example (part 1) of data explanation information. [Figure 12]FIG. 13 is a diagram showing an example (part 2) of data explanation information. [Figure 13] 13 is a flowchart showing a first modified example of the explanation generating process according to the present embodiment. [Figure 14] 13 is a flowchart showing a second modification of the explanation generating process according to the present embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] An embodiment of the present invention will be described in detail below with reference to the drawings. In the following embodiment, a data explanation device 10 capable of generating an explanation for data related to a certain predetermined facility (hereinafter referred to as "target facility") will be described. Examples of the target facility include various plants such as power plants, steel plants, chemical plants, petroleum plants, and energy plants. Examples of the data related to the target facility include data showing the operation record of the target facility (hereinafter referred to as "operation record data") and data showing an optimal operation plan for the target facility (hereinafter referred to as "optimal operation plan data"). The optimal operation plan data is, for example, data obtained by solving an optimization problem for determining an optimal operation plan for the target facility.

[0010] The format of the optimization problem for obtaining the optimal operation plan data is not limited to a specific format, and it is possible to use optimal operation plan data obtained by solving an optimization problem of any format. That is, the optimal operation plan data may be, for example, data obtained by solving a continuous optimization problem, or data obtained by solving a discrete optimization problem (also called a "combinatorial optimization problem" or the like). In addition, various methods (e.g., metaheuristics, reinforcement learning, other existing methods, etc.) can be used to solve the optimization problem depending on the problem. In the following, as an example, it is assumed that the optimal operation plan data is data obtained by solving a continuous optimization problem.

[0011] Furthermore, targeting data related to facilities is just one example, and it is also possible to target data related to devices, data related to systems, etc.

[0012] Here, the data explanation device 10 according to the present embodiment generates an explanation for data related to the target facility using a generation AI (Artificial Intelligence). At this time, the data explanation device 10 according to the present embodiment generates a plurality of explanations for the data, and then creates an explanation that integrates these plurality of explanations. As a result, by referring to this explanation, the user can, for example, grasp problems in the current operation from the operation record, or verify the validity of the optimal operation plan. In addition, it is also possible to confirm improvement measures for problems in the current operation, or compare the current operation with the optimal operation plan. Therefore, by using the data explanation device 10 according to the present embodiment, it is possible to support the user in making some decision (for example, deciding whether or not to actually apply the optimal operation plan) regarding the target facility.

[0013] The data explanation device 10 according to the present embodiment is realized, for example, by an information processing device (computer) or an information processing system (computer system) composed of one or more information processing devices. Specific examples of the information processing device include a personal computer, a general-purpose server, a smartphone, and a tablet terminal.

[0014] Generative AI refers to a technology or mechanism that generates content such as natural language sentences or images that satisfy given instructions or questions (these instructions or questions are called "prompts") Generative AI is generally realized by machine learning models including, for example, large language models (LLMs).

[0015] <Example of target equipment configuration> Below, the energy plant shown in Figure 1 will be taken as an example of the target facility. The energy plant shown in Figure 1 is composed of a gas turbine (+exhaust gas boiler), a once-through boiler, N turbo chillers, and N steam absorption chillers. The gas turbine (+exhaust gas boiler) is composed of a gas turbine that generates electricity from city gas, and an exhaust gas boiler that outputs steam from the gas exhausted from the gas turbine. The once-through boiler outputs steam from city gas. The turbo chiller outputs cold energy using the electricity generated by the gas turbine and purchased electricity as needed. The steam absorption chiller outputs cold energy using the steam output from the exhaust gas boiler and the steam output from the once-through boiler. The electricity, cold energy, and steam output from the energy plant are used for the power load, heat load, and steam load, respectively.

[0016] The configuration of the energy plant shown in FIG. 1 is just an example, and the present invention is not limited to this.

[0017] <Example of hardware configuration of data explanation device 10> An example of the hardware configuration of the data explanation device 10 according to this embodiment is shown in Fig. 2. As shown in Fig. 2, the data explanation device 10 according to this embodiment has an input device 101, a display device 102, an external I / F 103, a communication I / F 104, a RAM (Random Access Memory) 105, a ROM (Read Only Memory) 106, an auxiliary storage device 107, and a processor 108. Each of these pieces of hardware is connected to each other via a bus 109 so as to be able to communicate with each other.

[0018] The input device 101 is, for example, a keyboard, a mouse, a touch panel, a physical button, etc. The display device 102 is, for example, a display, a display panel, etc. Note that the data explanation device 10 does not necessarily have at least one of the input device 101 and the display device 102, for example.

[0019] The external I / F 103 is an interface with an external device such as a recording medium 103a. Examples of the recording medium 103a include a CD (Compact Disc), a DVD (Digital Versatile Disk), an SD memory card (Secure Digital memory card), and a USB (Universal Serial Bus) memory card.

[0020] The communication I / F 104 is an interface for connecting to a communication network. The RAM 105 is a volatile semiconductor memory (storage device) that temporarily stores programs and data. The ROM 106 is a non-volatile semiconductor memory (storage device) that can store programs and data even when the power is turned off. The auxiliary storage device 107 is a non-volatile storage device such as a hard disk drive (HDD), a solid state drive (SSD), or a flash memory. The processor 108 is an arithmetic device such as a central processing unit (CPU) or a graphic processing unit (GPU).

[0021] 2 is an example, and the hardware configuration of the data explanation device 10 is not limited to this. For example, the data explanation device 10 may have multiple auxiliary storage devices 107 and multiple processors 108, may not have some of the hardware shown in the figure, or may have various hardware other than the hardware shown in the figure.

[0022] <Example of functional configuration of data explanation device 10> An example of the functional configuration of the data explanation device 10 according to this embodiment is shown in Fig. 3. As shown in Fig. 3, the data explanation device 10 according to this embodiment has an explanation type selection unit 201, a data acquisition unit 202, an information sorting unit 203, an explanation generation processing unit 204, an explanation output unit 205, and a user input acquisition unit 206. The data explanation device 10 according to this embodiment also has a data storage unit 301.

[0023] The explanation type selection unit 201, the data acquisition unit 202, the information organization unit 203, the explanation generation processing unit 204, the explanation output unit 205, and the user input acquisition unit 206 are realized, for example, by processing executed by the processor 108 or the like of one or more programs installed in the data explanation device 10. Note that at least one of these units may be realized, for example, by processing executed by an external cloud server or the like.

[0024] The data storage unit 301 is realized, for example, by a storage area of ​​the auxiliary storage device 107. The data storage unit 301 may be realized by a storage area of ​​a storage device (e.g., a storage device provided in a database server) communicably connected to the data explanation device 10.

[0025] The explanation type selection unit 201 selects one or more explanation types for which the user wants to receive an explanation from one or more predetermined explanation types. The explanation type is a type of explanation for data related to the target equipment. Hereinafter, as an example, four explanation types are assumed: "problems in the current operation", "measures to improve the problems", "comparison and consideration between the current operation and the optimization result", and "reason for the optimization result". Here, "problems in the current operation" is an explanation type when an explanation is desired regarding problems in the current operation of the target equipment. "Measures to improve the problems" is an explanation type when an explanation is desired regarding measures to improve problems in the current operation of the target equipment. "Comparison and consideration between the current operation and the optimization result" is an explanation type when an explanation is desired regarding comparison and consideration between the current operation of the target equipment and the optimal operation plan. "Reason for the optimization result" is an explanation type when an explanation is desired regarding the reason for the optimal operation plan of the target equipment. However, all of these four explanation types are examples, and the explanation types are not limited to these four.

[0026] The data acquiring unit 202 acquires at least one of the driving record data and the optimal operation plan data from the data storage unit 301 according to the explanation type selected by the explanation type selecting unit 201. Here, the data acquiring unit 202 acquires driving record data from the data storage unit 301 when the explanation type selected by the explanation type selecting unit 201 includes "problems with the current operation" or "measures to improve the problems", acquires optimal operation plan data from the data storage unit 301 when the explanation type selected by the explanation type selecting unit 201 includes "reason for optimization result", and acquires driving record data and optimal operation plan data from the data storage unit 301 when the explanation type selected by the explanation type selecting unit 201 includes "comparison and consideration between the current operation and the optimization result".

[0027] The information organizing unit 203 organizes or converts the data acquired by the data acquisition unit 202 (operation performance data or optimal operation plan data, or both) into a predetermined format (e.g., array format, etc.) that can be input to the generation AI.

[0028] The explanation generation processing unit 204 generates data explanation information that represents an explanation for the data acquired by the data acquisition unit 202. Here, the explanation generation processing unit 204 includes a prompt generation unit 211, an explanation generation unit 212, and an integration unit 213. The prompt generation unit 211 creates a prompt corresponding to each explanation type selected by the explanation type selection unit 201. The explanation generation unit 212 generates data explanation information that represents an explanation for the data acquired by the data acquisition unit 202, using the prompt generated by the prompt generation unit 211 and a generation AI for each explanation type selected by the explanation type selection unit 201. That is, the explanation generation unit 212 generates data explanation information by the generation AI by inputting the prompt generated by the prompt generation unit 211 to the generation AI. The integration unit 213 creates data explanation information (hereinafter, also referred to as "integrated data explanation information") that integrates one or more pieces of data explanation information generated by the explanation generation unit 212.

[0029] In addition, the large-scale language models that realize generative AI acquire various domain knowledge during training, so when data explanation information is generated by generative AI, it is expected that the data explanation information will be generated using this domain knowledge.

[0030] Furthermore, the explanation generating unit 212 generates new data explanation information by the generation AI by inputting user input information, which will be described later, as a prompt to the generation AI. This new data explanation information is, for example, information that represents an additional explanation for the data explanation information previously generated by the generation AI, information that represents a summary of the data explanation information previously generated by the generation AI, information that represents a response to some question or instruction for the data explanation information previously generated by the generation AI, etc.

[0031] The explanation output unit 205 outputs the data explanation information or the integrated data explanation information generated by the explanation generation unit 212 to a predetermined output destination. Examples of the predetermined output destination include the display device 102 such as a display, a storage area such as the auxiliary storage device 107, other devices or equipment connected in a communicable manner, and the like.

[0032] The user input acquisition unit 206 inputs information such as a natural language sentence input by a user (hereinafter, also referred to as "user input information"). Examples of the user input information include information expressing a natural language sentence requesting additional explanation for the data explanatory information, information expressing a natural language sentence requesting a summary of the data explanatory information, information expressing a natural language sentence such as a standard question for the data explanatory information, and information expressing a natural language sentence meaning some kind of question or instruction for other data explanatory information.

[0033] The data storage unit 301 stores data (operation performance data, optimal operation plan data, etc.) related to the target facility. The operation performance data is collected from the target facility (more precisely, collected from sensors, etc. installed in the target facility or the devices or equipment constituting the target facility). On the other hand, the optimal operation plan data is obtained by solving an optimization problem for obtaining an optimal operation plan for the target facility. The subject that solves the optimization problem may be the data explanation device 10, or may be a server (e.g., a cloud server) that is communicably connected to the data explanation device 10. When the data explanation device 10 solves the optimization problem, the data explanation device 10 may have a functional unit called, for example, an "optimization calculation unit" as a functional unit for solving the optimization problem.

[0034] <Example of data explanation processing> An example of the data explanation process according to this embodiment will be described with reference to Fig. 4. For example, steps S101 to S105 in Fig. 4 are executed each time data explanation information is needed, and steps S106 to S108 are executed repeatedly one or more times as necessary after the data explanation information is obtained.

[0035] The explanation type selection unit 201 selects one or more explanation types for which the user wants to receive an explanation from one or more predetermined explanation types (step S101). Note that the explanation type selection unit 201 may select, for example, an explanation type designated by the user from one or more predetermined explanation types. In the following, as an example, it is assumed that four explanation types, "problems with the current operation", "measures to improve the problems", "comparison and consideration between the current operation and the optimization result", and "reason for the optimization result", have been selected.

[0036] The data acquisition unit 202 acquires at least one of the operation record data and the optimal operation plan data from the data storage unit 301 according to the explanation type selected in the above step S101 (step S102). For example, when the explanation type selected in the above step S101 includes "problems with the current operation" or "measures to improve the problem", the data acquisition unit 202 acquires the operation record data from the data storage unit 301. Also, for example, when the explanation type selected in the above step S101 includes "reason for the optimization result", the data acquisition unit 202 acquires the optimal operation plan data from the data storage unit 301. Furthermore, for example, when the explanation type selected in the above step S101 includes "comparison and consideration between the current operation and the optimization result", the data acquisition unit 202 acquires the operation record data and the optimal operation plan data from the data storage unit 301.

[0037] Here, the operation record data and the optimal operation plan data are expressed as time-series data of data (hereinafter referred to as "state data") consisting of one or more variables (when the target facility is a plant or the like, these variables are also called "state variables" or the like). When the target facility is the energy plant shown in Fig. 1, examples of the state variables include the amount of electricity purchased, the amount of city gas purchased, the amount of city gas consumed by the gas turbine, the power output of the gas turbine, the steam output of the exhaust gas boiler, the amount of electricity consumed by the turbo chiller, the heat output of the turbo chiller, the amount of steam consumed by the steam absorption chiller, the heat output of the steam absorption chiller, the amount of city gas consumed by the once-through boiler, the steam output of the once-through boiler, the surplus steam of the once-through boiler, the power load, the heat load, the steam load, etc.

[0038] When the operating performance data is acquired in step S102, the data acquisition unit 202 may, for example, sequentially acquire the status data constituting the operating performance data from the target equipment in real time.

[0039] The information organizing unit 203 organizes or converts the data (operation performance data or optimal operation plan data, or both) acquired in the above step S102 into a predetermined format (e.g., array format, etc.) that can be input to the generation AI (step S103). As an example, FIG. 5 shows data in which operation performance data in the case where the energy plant shown in FIG. 1 is the target facility is organized or converted into an array format. Similarly, FIG. 6 shows data in which optimal operation plan data in the case where the energy plant shown in FIG. 1 is the target facility is organized or converted into an array format. FIGS. 5 and 6 show examples in which the time series data of each state variable is expressed as an array. In FIGS. 5 and 6, each row represents one array, and in each array, the name of the state variable corresponding to the array and the unit of the value set for the state variable are set. In addition, in each array, the time series data of the state variables from time s=1 to time s=t is stored.

[0040] In addition, if the data acquired in the above step S102 (operation performance data or optimal operation plan data, or both) is in a format that can be input to the generation AI, the above step S103 does not need to be executed. Hereinafter, the data after sorting or conversion in the above step S103 (operation performance data or optimal operation plan data, or both) will be simply referred to as "operation performance data" or "optimum operation plan data".

[0041] The explanation generation processing unit 204 executes an explanation generation process for generating data explanation information that represents an explanation for the data organized or converted in step S103 (step S104). As a result, integrated data explanation information is obtained. The explanation generation process will be described in detail later.

[0042] The explanation output unit 205 outputs the integrated data explanation information obtained in the above step S104 to a predetermined output destination (step S105).

[0043] The user input acquisition unit 208 acquires user input information input by the user (step S106).

[0044] The explanation generation unit 212 of the explanation generation processing unit 204 generates new data explanation information by using the user input information acquired in the above step S106 and the generation AI (step S107). That is, the explanation generation unit 212 generates new data explanation information by the generation AI by inputting the user input information acquired in the above step S106 to the generation AI as a prompt.

[0045] The explanation output unit 205 outputs the data explanation information generated in the above step S107 to a predetermined output destination (step S108).

[0046] <Example of explanation generation process> An example of the explanation generation process according to this embodiment will be described with reference to FIG. 7. Hereinafter, the explanation type selected in step S101 in FIG. 4 will be referred to as the "explanation target type". Also, the number of explanation target types will be referred to as the "number of explanation target types" and represented by I. Furthermore, the i-th explanation target type will be represented as the "explanation target type i". Hereinafter, as an example, the first explanation target type will be "problems with the current operation", the second explanation target type will be "measures to remedy the problems", the third explanation target type will be "comparison and consideration of the current operation and the optimization result", and the fourth explanation target type will be "reason for the optimization result".

[0047] The explanation generation processor 204 initializes a variable i, which indicates the number of an explanation object type, to 1 (step S201).

[0048] The prompt generating unit 211 of the explanation generation processing unit 204 generates a prompt corresponding to the explanation object type i (step S202). The prompt generating unit 211 may generate a prompt corresponding to the explanation object type i, for example, by following the steps 1 to 3 below.

[0049] Step 1: The prompt generation unit 211 acquires a prompt template that is a model (template) of a prompt corresponding to the explanation object type i. The prompt template is template information that includes a natural language sentence that represents an instruction for generating an explanation related to the explanation object type i, and zero or more variable parts. The variable parts are parts where data corresponding to the explanation object type i is set (i.e., variable parts of the prompt template). The prompt template is created in advance by, for example, a user and then saved in a predetermined storage area.

[0050] Step 2: If the prompt template acquired in step 1 above includes a variable portion, the prompt generation unit 211 sets data (operation record data or optimal operation plan data, or both) corresponding to the explanation object type i to the variable portion. Hereinafter, the data set to the variable portion included in the prompt template will be referred to as "setting information."

[0051] Step 3: If the prompt template acquired in step 1 above does not include a variable portion, the prompt generation unit 211 sets the prompt template as the prompt corresponding to the explanation object type i.

[0052] A prompt corresponding to the explanation object type i is generated by the above steps 2 or 3. Note that examples of prompt templates including a variable portion include a prompt template corresponding to "Problems with the current operation," a prompt template corresponding to "Comparison and consideration between the current operation and the optimization result," and a prompt template corresponding to "Reasons for the optimization result." On the other hand, examples of prompt templates not including a variable portion include a prompt template corresponding to "Measures to remedy the problems."

[0053] As an example, a prompt corresponding to the explanation object type "problems with current operation" is shown in Fig. 8. Prompt 1100 shown in Fig. 8 is a prompt for generating data explanation information based on the results of identifying problems with the current operation of the target equipment (i.e., the operational performance data shown in Fig. 5) from the perspective of energy cost efficiency. Note that in prompt 1100 shown in Fig. 8, the setting information is shown in bold and underlined.

[0054] As another example, a prompt corresponding to the explanation object type "reason for optimization result" is shown in Fig. 9. Prompt 1200 shown in Fig. 9 is a prompt for generating the reason for the optimal solution (i.e., the optimal operation plan data shown in Fig. 6) obtained by the mathematical optimization technology as data explanation information. Note that in the prompt 1200 shown in Fig. 9, the setting information is shown in bold and underlined.

[0055] As another example, a prompt corresponding to the explanation object type "remedy for problem" is shown in Fig. 10. Prompt 1300 shown in Fig. 10 is a prompt for generating a remedy for a problem obtained as an explanation for the explanation object type "problem in current operation" as data explanation information. Note that no setting information exists in prompt 1300 shown in Fig. 10.

[0056] The explanation generation unit 212 of the explanation generation processing unit 204 generates data explanation information corresponding to the explanation object type i using the prompt generated in the above step S202 (step S203). That is, the explanation generation unit 212 inputs the prompt generated in the above step S202 to the generation AI, and generates data explanation information corresponding to the explanation object type i by the generation AI. Hereinafter, the data explanation information corresponding to the explanation object type i will also be referred to as "data explanation information i".

[0057] In addition, when generating data explanatory information i (where i≧2), the generation AI generates data explanatory information i based on data explanatory information 1 to data explanatory information i-1 as well. Therefore, when generating data explanatory information i (where i≧2), data explanatory information i can be generated taking into account the contents of data explanatory information 1 to data explanatory information i-1.

[0058] As an example, data explanation information corresponding to the explanation object type "problems with current operation" is shown in Fig. 11. In the data explanation information 2100 shown in Fig. 11, problems with the current operation of the target equipment are explained in natural language sentences from the viewpoint of energy cost efficiency. Note that the data explanation information 2100 shown in Fig. 11 is data explanation information generated by inputting the prompt 1100 shown in Fig. 8 to the generation AI.

[0059] As another example, data explanation information corresponding to the explanation object type "reason for optimization result" is shown in Fig. 12. In the data explanation information 2200 shown in Fig. 12, the reason why the optimal solution obtained by the mathematical optimization technology is optimal is explained in natural language sentences. Note that the data explanation information 2200 shown in Fig. 12 is data explanation information generated by inputting the prompt 1200 shown in Fig. 9 to the generation AI.

[0060] The explanation generation processor 204 adds 1 to the variable i representing the number of the explanation object type (step S204).

[0061] The explanation generation processor 204 determines whether i is equal to or smaller than the number of explanation target types (step S205). That is, the explanation generation processor 204 determines whether i≦I.

[0062] When it is determined in step S205 that i is equal to or smaller than the number of explanation target types, the explanation generating processor 204 returns to step S202. As a result, steps S202 to S204 are repeatedly executed until i>I.

[0063] On the other hand, if it is not determined in step S205 that i is equal to or less than the number of types to be explained, the integration unit 213 of the explanation generation processing unit 204 creates integrated data explanation information by integrating the data explanation information i (1≦i≦I) (step S206). The method of integrating the data explanation information i (1≦i≦I) is not limited to a specific method, but as an example, for i=1,...,I-1, it is possible to concatenate data explanation information i+1 after the data explanation information i. As another example, it is possible to create integrated data explanation information by summarizing the data explanation information i (1≦i≦I) using a generation AI or an existing document summarization technology.

[0064] The above explanation generation process produces integrated data explanation information that integrates data explanation information corresponding to, for example, the explanation object types “problems with the current operation”, “measures to remedy the problems”, “comparison and consideration of the current operation and the optimization result”, and “reason for the optimization result”. Therefore, the user can consistently know, for example, the problems with the current operation, the measures to remedy the problems, the comparison and consideration of the current operation and the optimization result, and the reason for the optimization result.

[0065] <Modification> Variation 1 As a variant of the explanation generation process shown in Figure 7, we will explain the explanation generation process in which multiple pieces of data explanation information are generated by the same generation AI using the same prompt, and then these pieces of data explanation information are integrated, with reference to Figure 13.

[0066] The explanation generation processor 204 initializes a variable i representing the number of an explanation object type to 1, similarly to step S201 in FIG. 7 (step S301).

[0067] The prompt generating unit 211 of the explanation generating processor 204 generates a prompt corresponding to the explanation object type i (step S302), similarly to step S202 in FIG.

[0068] The explanation generation processing unit 204 initializes a variable j, which indicates the number of times data explanation information is generated in response to a prompt corresponding to the explanation object type i, to 1 (step S303).

[0069] The explanation generation unit 212 of the explanation generation processing unit 204 generates the j-th data explanation information corresponding to the explanation object type i using the prompt generated in the above step S302, similar to step S203 in FIG. 7 (step S304). Hereinafter, the j-th data explanation information corresponding to the explanation object type i is referred to as "data explanation information i". j " will also be expressed as ".

[0070] The explanation generation processing unit 204 adds 1 to the variable j that indicates the number of times data explanation information is generated in response to the prompt corresponding to the explanation object type i (step S305).

[0071] The explanation generation processor 204 determines whether j is equal to or less than a predetermined designated number of times (step S306). That is, for example, when the designated number of times corresponding to the explanation object type i is J(i), the explanation generation processor 204 determines whether j≦J(i).

[0072] If it is determined in step S306 that j is equal to or less than the predetermined designated number, the explanation generation processor 204 returns to step S304, whereby steps S304 to S305 are repeatedly executed until j>J(i).

[0073] On the other hand, if it is not determined in step S306 that j is equal to or less than the predetermined designated number, the integration unit 213 of the explanation generation processing unit 204 integrates the data explanation information i j (1≦j≦J(i)) is integrated to generate a first integrated data explanation information (step S307). j The method of integrating (1≦j≦J(i)) is not limited to a specific method. For example, for j=1, ,J(i)-1, data description information i j followed by data description information i j+1Another example is the use of generative AI or existing document summarization technology to generate data description information i j It is possible to create a first integrated data description by summarizing (1≦j≦J(i)). Hereinafter, data description information i j The first integrated data explanation information obtained by integrating (1≦j≦J(i)) will also be referred to as "first integrated data explanation information i."

[0074] The explanation generation processor 204 adds 1 to the variable i representing the number of the explanation object type, similarly to step S204 in FIG. 7 (step S308).

[0075] The explanation generation processor 204 determines whether i is equal to or smaller than the number of explanation target types (step S309), similarly to step S205 in Fig. 7. That is, the explanation generation processor 204 determines whether i≦I.

[0076] If it is determined in step S309 that i is equal to or smaller than the number of explanation target types, the explanation generation processor 204 returns to step S302. As a result, steps S302 to S308 are repeatedly executed until i>I.

[0077] On the other hand, if it is not determined in step S309 that i is equal to or less than the number of types to be explained, the integration unit 213 of the explanation generation processing unit 204 creates second integrated data explanation information by integrating the first integrated data explanation information i (1≦i≦I) (step S310). The method of integrating the first integrated data explanation information i (1≦i≦I) is not limited to a specific method, but as an example, for i=1,...,I-1, it is possible to link the first integrated data explanation information i+1 after the first integrated data explanation information i. As another example, it is possible to create the second integrated data explanation information by summarizing the first integrated data explanation information i (1≦i≦I) using a generation AI or an existing document summarization technology. Note that in step S105 of FIG. 4, the second integrated data explanation information is output to a predetermined output destination.

[0078] In the explanation generation process shown in FIG. 13, data explanation information i j Since the first integrated data explanatory information i is created by integrating (1≦j≦J(i)), the first integrated data explanatory information i is obtained as a robust explanation corresponding to the explanation target type i. Therefore, even if the generation AI uses a probabilistic method to generate an explanation, it is expected that the first integrated data explanatory information i will be generated with high reliability.

[0079] Variation 2 As a variant of the explanation generation process shown in Figure 7, we will explain the explanation generation process in which multiple data explanation information is generated by multiple generation AIs from the same prompt, and then these data explanation information are integrated, with reference to Figure 14.

[0080] The explanation generation processor 204 initializes a variable i representing the number of an explanation object type to 1, similarly to step S201 in FIG. 7 (step S401).

[0081] The prompt generating unit 211 of the explanation generating processor 204 generates a prompt corresponding to the explanation object type i (step S402), similarly to step S202 in FIG.

[0082] The explanation generation unit 212 of the explanation generation processing unit 204 generates a plurality of pieces of data explanation information by a plurality of generation AIs using the prompt generated in the above step S402 (step S403). That is, the explanation generation unit 212 generates a plurality of pieces of data explanation information by a plurality of generation AIs by inputting the prompt generated in the above step S402 to each of the plurality of generation AIs. Hereinafter, the number of generation AIs used when generating data explanation information from a prompt corresponding to the explanation object type i is defined as K(i), and the data explanation information generated by the kth generation AI is referred to as "data explanation information i". k ". Note that the K(i) generation AIs may all be different generation AIs, or some of the generation AIs may be the same generation AI.

[0083] The integration unit 213 of the explanation generation processing unit 204 integrates the data explanation information i k (1≦k≦K(i)) is integrated to generate a first integrated data explanation information (step S404). k The method of integrating (1≦k≦K(i)) is not limited to a specific method. For example, for k=1, ,K(i)-1, data description information i k followed by data description information i k+1 Another example is the use of generative AI or existing document summarization technology to generate data description information i k It is possible to create a first integrated data description by summarizing (1≦k≦K(i)). Hereinafter, data description information i k The first integrated data explanation information obtained by integrating (1≦k≦K(i)) will also be referred to as “first integrated data explanation information i”.

[0084] The explanation generation processor 204 adds 1 to the variable i representing the number of the explanation object type, similarly to step S204 in FIG. 7 (step S405).

[0085] The explanation generation processor 204 determines whether i is equal to or smaller than the number of explanation target types (step S406), similarly to step S205 in Fig. 7. That is, the explanation generation processor 204 determines whether i≦I.

[0086] If it is determined in step S406 that i is equal to or smaller than the number of explanation target types, the explanation generation processor 204 returns to step S402. As a result, steps S402 to S405 are repeatedly executed until i>I.

[0087] On the other hand, if it is not determined in step S406 that i is equal to or less than the number of types to be explained, the integration unit 213 of the explanation generation processing unit 204 creates second integrated data explanation information by integrating the first integrated data explanation information i (1≦i≦I) (step S407), similar to step S310 in Fig. 13. Note that in step S105 in Fig. 4, the second integrated data explanation information is output to a predetermined output destination.

[0088] In the explanation generation process shown in FIG. 14, data explanation information i k Since the first integrated data explanatory information i is created by integrating (1≦k≦K(i)), the first integrated data explanatory information i is obtained as an explanation corresponding to the explanation object type i, which is highly robust. Therefore, as in the above-mentioned modification example 1, even if the generation AI generates an explanation using a probabilistic method, it is expected that the first integrated data explanatory information i with high reliability will be generated.

[0089] Variation 3 The prompt is not limited to a natural language sentence, and may include content such as an image, a video, a sound, etc. Furthermore, it may include a mathematical formula, a program code, etc.

[0090] In particular, when instructing the AI ​​to perform some kind of numerical calculation or processing, the prompt may contain the formula or program code used in the numerical calculation or processing, which makes it possible to reduce calculation errors when the AI ​​performs numerical calculations.

[0091] <Summary> As described above, the data explanation device 10 according to the present embodiment uses a generation AI to generate and output an explanation for data related to the target facility. At this time, the data explanation device 10 according to the present embodiment generates multiple explanations for the data, and then creates an explanation that integrates these multiple explanations. This allows the user of the data explanation device 10 according to the present embodiment to make various decisions by referring to this explanation.

[0092] The present invention is not limited to the above-described embodiments specifically disclosed, and various modifications, changes, combinations with known technologies, etc. are possible without departing from the scope of the claims. [Explanation of symbols]

[0093] 10 Data Explanation Device 101 Input Device 102 Display device 103 External I / F 103a Recording media 104 Communication I / F 105 RAM 106 ROM 107 Auxiliary storage 108 processors 109 Bus 201 Explanation Type Selection Section 202 Data Acquisition Department 203 Information Organizing Department 204 Explanation Generation Processing Unit 205 Explanation output section 206 User Input Acquisition Unit 211 Prompt Generation Unit 212 Explanation Generation Unit 213 Integration Department 301 Data storage unit

Claims

1. A selection unit that selects one or more types of explanation to be generated for data related to the facility; a data acquisition unit that acquires the data according to the one or more selected types; An explanation generation unit that generates one or more explanations corresponding to each of the one or more types based on the acquired data and a generation AI realized by a machine learning model including a large-scale language model; an integration unit that creates a first integrated explanation that represents an explanation obtained by integrating the one or more explanations corresponding to each of the one or more types; An information processing device having the above configuration.

2. a creation unit that creates a prompt corresponding to the one or more types by using the data corresponding to the type, The explanation generation unit is The information processing apparatus according to claim 1 , further comprising: an information generating unit configured to generate an explanation corresponding to the type by inputting the prompt into the generating AI.

3. The one or more types include a first type representing an explanation regarding a problem with the current operation of the equipment, a second type representing an explanation regarding a remedy for the problem, a third type representing an explanation regarding a comparison and consideration between the current operation of the equipment and an optimal operation plan for the equipment, and a fourth type representing an explanation regarding the reason for the optimal operation plan for the equipment, The explanation generation unit is The information processing apparatus according to claim 2 , further comprising: generating the prompt in order from the first type to the fourth type; and generating an explanation corresponding to the type.

4. The explanation generation unit is inputting the prompt to the generating AI a plurality of times to generate a plurality of descriptions corresponding to the type; The integration unit includes: creating a second integrated explanation representing an explanation obtained by integrating a plurality of explanations corresponding to the type; The information processing apparatus according to claim 2 , further comprising: a processor configured to generate the first integrated explanation by integrating second integrated explanations corresponding to the one or more types.

5. The explanation generation unit is inputting the prompt to a plurality of generating AIs to generate a plurality of descriptions corresponding to the type; The integration unit includes: creating a second integrated explanation representing an explanation obtained by integrating a plurality of explanations corresponding to the type; The information processing apparatus according to claim 2 , further comprising: a processor configured to generate the first integrated explanation by integrating second integrated explanations corresponding to the one or more types.

6. The integration unit includes: The information processing device according to claim 1 , which uses generative AI or document summarization technology to create an explanation that integrates multiple explanations.

7. A selection step of selecting one or more types of explanation to be generated for the data related to the facility; a data acquisition step of acquiring the data in accordance with the one or more selected types; An explanation generation step of generating one or more explanations corresponding to each of the one or more types based on the acquired data and a generation AI realized by a machine learning model including a large-scale language model; an integration step of creating a first integrated explanation representing an explanation obtained by integrating the one or more explanations corresponding to each of the one or more types; An information processing method implemented by a computer.

8. A selection step of selecting one or more types of explanation to be generated for the data related to the facility; a data acquisition step of acquiring the data in accordance with the one or more selected types; An explanation generation step of generating one or more explanations corresponding to each of the one or more types based on the acquired data and a generation AI realized by a machine learning model including a large-scale language model; an integration step of creating a first integrated explanation representing an explanation obtained by integrating the one or more explanations corresponding to each of the one or more types; A program that causes a computer to execute the following.

Images