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

Abstract

Provided are an information processing device, a method for controlling the information processing device, and a program, whereby the convenience of a user can be improved by using a plurality of language models to improve the efficiency of work for obtaining a response having a comprehensive viewpoint.　An information processing device 1 comprises: a response result acquisition unit 33 that transmits, to a plurality of language models 3 (for example, language models 3-1 to 3-3), a request prompt including content of a user request , and acquires a response result for the request content from each of the plurality of language models 3; and an integration processing unit 35 that generates an integration instruction prompt for instructing the language models 3 to integrate the response results of each of the plurality of language models 3. The response result acquisition unit 33 transmits the integration instruction prompt to the language models 3 to thereby acquire an integrated response result in which the response results of the plurality of language models 3 are reflected.

Description

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

[0001] The present invention relates to an information processing device, a control method for an information processing device, and a program.

[0002] Conventionally, information processing techniques have been known that use generative AI language models such as LLMs to analyze data and generate ideas. For example, Patent Document 1 describes this type of technique.

[0003] Patent Literature 1 relates to a data analysis device that analyzes various types of data. The data analysis device in Patent Literature 1 generates a system prompt by assigning input information in a natural language format by a user and analysis background information to a prompt template, transmits text information including the input information and the system prompt to a pre-trained language model that can interpret natural language, and generates display data based on a response output from the pre-trained language model after transmission.

[0004] JP 2025-067282 A

[0005] Generative AI language models such as LLMs output different answers depending on the training data and parameters, even when the same prompt is input. For example, even if the same prompt is input to GPT-4o and Claude (registered trademark), the answer results will be different between GPT-4o and Claude.

[0006] When a user wants to find comprehensive ideas or solutions, they may consider using multiple language models. However, to consolidate the answer results from multiple language models, they must send prompts, which are instructions, to each service that provides the language models, and then retrieve and consolidate the answer results from each language model, which is a time-consuming task.

[0007] The present invention has been made in consideration of the above circumstances, and aims to provide an information processing device, a control method for an information processing device, and a program that can improve user convenience by using multiple language models to streamline the process of obtaining answers with a comprehensive perspective.

[0008] In order to achieve the above object, one aspect of the present invention is an information processing device comprising: an answer result acquisition unit that sends a request prompt including a user's request content to a plurality of language models and acquires an answer result for the request content from each of the plurality of language models; and an integration processing unit that generates an integration instruction prompt that instructs the language models to integrate the answer results of each of the plurality of language models, wherein the answer result acquisition unit acquires an integrated answer result that reflects the answer results of the plurality of language models by sending the integration instruction prompt to the language models.

[0009] Another aspect of the present invention is a control method for an information processing device, including: an answer result acquisition step of sending a request prompt including a user's request content to a plurality of language models and acquiring an answer result for the request content from each of the plurality of language models; an integration processing step of generating an integration instruction prompt that instructs the language models to integrate the answer results of each of the plurality of language models; and an integrated answer result acquisition step of sending the integration instruction prompt to the language models to acquire an integrated answer result that reflects the answer results of the plurality of language models.

[0010] Another aspect of the present invention is a program for causing a computer to execute an answer result acquisition step of sending a request prompt including a user's request content to a plurality of language models and acquiring an answer result for the request content from each of the plurality of language models; an integration processing step of generating an integration instruction prompt that instructs the language model to integrate the answer results of each of the plurality of language models; and an integrated answer result acquisition step of sending the integration instruction prompt to the language model to acquire an integrated answer result that reflects the answer results of the plurality of language models.

[0011] According to the present invention, it is possible to provide an information processing device, a control method for an information processing device, and a program that can improve user convenience by using multiple language models to streamline the process of obtaining answers with a comprehensive perspective.

[0012] 1 is a diagram showing a language model utilization system to which an information processing device according to a first embodiment of the present invention is applied; FIG. 2 is a block diagram showing the hardware configuration of the information processing device according to the first embodiment; FIG. 3 is a functional block diagram showing an example of the functional configuration of the information processing device according to the first embodiment; FIG. 4 is a diagram showing an example of a request input screen displayed on a user terminal according to the first embodiment; FIG. 5 is a diagram showing an example of a language model selection screen displayed on a user terminal; FIG. 6 is a diagram showing an example of an answer result presentation screen displayed on a user terminal; FIG. 7 is a diagram showing an example of the configuration of an integration instruction prompt; FIG. 8 is a flowchart showing an example of the flow of an answer presentation process for a user executed by an information processing device; FIG. 9 is a functional block diagram showing an example of the functional configuration of an information processing device according to a second embodiment; FIG. 10 is a schematic diagram showing the flow of data in the answer presentation process by the information processing device according to the second embodiment;

[0013] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0014] First Embodiment First, the overall system configuration will be described. Fig. 1 is a diagram showing a language model utilization system 100 to which an information processing device 1 according to a first embodiment of the present invention is applied. The language model utilization system 100 provides an idea-raising service that uses a plurality of language models 3-1 to 3-n to generate and present answers that serve as clues for organizing ideas or solving problems in response to requests from users. Note that the purpose of the service is merely an example, and the present invention can be applied to purposes other than the idea-raising service.

[0015] The language model utilization system 100 is realized by an information processing device 1 that communicates with a user terminal 2 via a communication network such as the Internet.

[0016] The user terminal 2 is an information processing device used by a user. The user terminal 2 is configured by a personal computer, a tablet, a smartphone, etc. The user terminal 2 may exchange various types of information through a web browser, or may exchange various types of information with the information processing device 1 using a pre-installed program.

[0017] The information processing device 1 is a server that generates a response to a user request by communicating with a plurality of language models 3-1 to 3-n for a user terminal 2. The information processing device 1 cooperates with the language models 3-1 to 3-n, which are external services, via, for example, an API (Application Programming Interface) or the like.

[0018] The language models 3-1 to 3-n are not necessarily limited to those provided as external services. For example, some or all of the language models 3-1 to 3-n may be implemented within the information processing device 1, or may be implemented on another server in the same system as the information processing device 1. Also, some of the multiple language models 3-1 to 3-n may be implemented within the information processing device 1, and the other parts may be used as external services. In either configuration, the basic processes of sending a request prompt and obtaining an answer result by the answer result acquisition unit 33 are executed in the same manner.

[0019] The language models 3-1 to 3-n represent different language models of the generation AI. The language models 3-1 to 3-n may be LLMs (Large Language Models) or SLMs (Small Language Models). The language models 3-1 to 3-n in FIG. 1 represent candidates to be used, and in this example, the language models 3-1 to 3-3 are used to present answers to the user.

[0020] The language models 3-1 to 3-n are, for example, OpenAI's GPT-3.5 Turbo, GPT-4o, Google's Gemini 1.0 Pro, Gemini 1.5 Pro, Gemini 1.5 Flash, Anthropic's Cloude 3 Haiku, Cloude 3 Sonnet, Cloude 3 Opus, and the like.

[0021] In the following description, when there is a commonality between the language models 3-1 to 3-n and no need to distinguish between them, they may be simply referred to as the language model 3.

[0022] <Hardware Configuration> Next, an example of hardware constituting the information processing device 1 will be described. Fig. 2 is a block diagram showing the hardware configuration of the information processing device 1 according to the first embodiment. The information processing device 1 includes a CPU (Central Processing Unit) 11 as a processor, a ROM (Read Only Memory) 12, a RAM (Random Access Memory) 13, a bus 14, an input / output interface 15, an output unit 16, an input unit 17, a storage unit 18, a communication unit 19, and a drive 20.

[0023] The CPU 11 executes various processes in accordance with programs recorded in the ROM 12 or programs loaded from the storage unit 18 into the RAM 13. The RAM 13 also stores data and the like required for the CPU 11 to execute various processes. The CPU 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output interface 15 is also connected to this bus 14.

[0024] The input / output interface 15 is connected to an output unit 16, an input unit 17, a storage unit 18, a communication unit 19, and a drive 20. The output unit 16 is composed of a display, a speaker, etc., and outputs various information as images and sounds. The input unit 17 is composed of a keyboard, a mouse, etc., and inputs various information. The storage unit 18 is composed of a hard disk, a DRAM (Dynamic Random Access Memory), etc., and stores various data. The communication unit 19 communicates with other devices via a network including the Internet.

[0025] Removable media 21, such as a magnetic disk, optical disk, magneto-optical disk, or semiconductor memory, is appropriately attached to the drive 20. Programs read from the removable media 21 by the drive 20 are installed in the storage unit 18 as needed. The removable media 21 can also store various data stored in the storage unit 18 in the same way as the storage unit 18.

[0026] The hardware configuration described here is merely an example. The computer described in this embodiment, including the information processing device 1, may have a configuration common to that shown in FIG. 2 or a different configuration. The computer may also be composed of two or more computers. The user terminal 2 in FIG. 1 is, for example, a personal computer, tablet, or personal computer having a configuration similar to the hardware configuration shown in FIG. 2.

[0027] <Functional Configuration> Next, a description will be given of the functional configuration of the information processing device 1. Fig. 3 is a functional block diagram showing an example of the functional configuration of the information processing device 1 according to the first embodiment. As shown in Fig. 3, the information processing device 1 includes an input / output processing unit 31, a language model designation unit 32, an answer result acquisition unit 33, an integration target setting unit 34, and an integration processing unit 35 as functional units realized on a processor (CPU 11).

[0028] The input / output processing unit 31 executes a process of receiving information input by the user and a process of presenting information to the user. In this embodiment, the input / output processing unit 31 executes a process of receiving information input by the user through a website on the Internet that the user accesses through the user terminal 2, and a process of presenting information to the user.

[0029] For example, the input / output processing unit 31 executes a process of receiving information input by a user and a process of presenting information to a user. Specifically, the input / output processing unit 31 generates display data for various screens (a request input screen, a language model selection screen, an answer result presentation screen, etc., which will be described later) to be displayed on the user terminal 2, and controls the display of these screens by transmitting the data to the user terminal 2. The input / output processing unit 31 also executes control to display the integration instruction prompt generated by the integration processing unit 35 in a format that can be confirmed and edited by the user.

[0030] 4 is a diagram showing an example of a request input screen displayed on the user terminal 2 of the first embodiment. The screen in Fig. 4 displays a used language model display section 101, a request content input field 102, a language model designation operation section 103, a user information display section 104, etc. The user accesses, for example, a URL (Uniform Resource Locator) to connect to a website that displays the request input screen.

[0031] The language model display unit 101 displays the language model 3 used when the user responds to a request. In this example, ChatGPT-4.1 nano, Gemini 2.0 Flash, and Gemini 2.5 Flash are specified as the language models 3 to be used. The request content input field 102 is a section where the user inputs the request content in natural language text. In this example, the user inputs the request content, "Please provide name candidates for a new function using generative AI." By operating the send button 124 in the request content input field 102, the request content is transmitted to the information processing device 1. The language model designation operation unit 103 is an operation button for transitioning to a screen where the user designates the language model 3 to be used. The language model designation operation unit 103 can also save combinations of language models 3 frequently used by the user. The user information display unit 104 displays information such as the user ID and tickets held by the user for using the language models 3. The ticket is given by the information processing device 1 in exchange for money, points, or the like.

[0032] The language model designation unit 32 executes a process of setting a language model 3 designated by the user from among a plurality of candidate language models 3 as the destination of the request prompt. In this embodiment, the language model 3 is designated based on the user's selection on a language model selection screen that is displayed by selecting the language model designation operation unit 103 on the request input screen in Fig. 4. The language model designation unit 32 acquires the user's selection of the language model 3 via the input / output processing unit 31.

[0033] FIG. 5 is a diagram showing an example of a language model selection screen displayed on the user terminal 2. The screen in FIG. 5 displays a basic language model selection unit 105, an advanced language model selection unit 106, a confirmation instruction unit 107, and the like. The basic language model selection unit 105 displays a list of language models 3 that the user can select even if the user does not have tickets. The advanced language model selection unit 106 displays a list of language models 3 that the user can select if the user has tickets. In this example, the language models 3 are displayed in three levels according to the amount of tickets consumed. The language models 3 that consume one ticket are listed in the top row of the advanced language model selection unit 106, language models 3 that consume 25 tickets are listed in the middle row, and language models 3 that consume 80 tickets are listed in the bottom row. The user determines the language model 3 to use by selecting a language model 3 displayed in the basic language model selection unit 105 or the advanced language model selection unit 106. In this example, the user can select up to six language models 3 to use simultaneously.

[0034] The confirmation instruction unit 107 is an operation button on the screen for issuing an instruction to set the language model 3 selected by the basic language model selection unit 105 or the advanced language model selection unit 106 to the language model 3 to be used. When the user operates the confirmation instruction unit 107 with a language model 3 selected, the language model 3 specified by the user is sent to the input / output processing unit 31. When the input / output processing unit 31 receives information about the language model 3 specified by the user, the language model designation unit 32 sets the language model 3 specified by the user to the language model 3 to be used. In this example, ChatGPT-4.1nano, Gemini 2.0Flash, and Gemini 2.5Flash are selected.

[0035] The answer result acquisition unit 33 executes a process of transmitting a request prompt including the content of the user's request to a plurality of language models 3 and acquiring an answer result for the request content from each of the plurality of language models 3. In this embodiment, "transmission" is not limited to external transmission via a communication network, but is used as a broad concept including data transfer between different modules in the same device, inter-process communication, data transfer to other servers in the system, etc.

[0036] As used herein, the term "prompt" refers to input information that requests processing from a language model or other natural language processing system. A prompt is not limited to textual instructions in natural language, but is a broad concept that includes input information in any format that can be interpreted by a language model, such as structured data, parameters, control signals, arguments for API calls, audio data, image data, or a combination of these.

[0037] A method for acquiring the user's request will now be described. In this embodiment, the user's request is acquired by the input / output processing unit 31 through the request content input field 102 on the request input screen of FIG. 4. The answer result acquisition unit 33 transmits a prompt requesting an answer to the request acquired by the input / output processing unit 31 to the specified language model 3. In the above example, a prompt including the request "Please provide candidate names for new functions using generative AI" is transmitted to each of ChatGPT-4.1 nano (language model 3-1), Gemini 2.0 Flash (language model 3-2), and Gemini 2.5 Flash (language model 3-3).

[0038] The answer result acquisition unit 33 acquires answer results from each language model 3. In the above example, answer results for the request content are acquired from the three specified language models 3-1 to 3-3. The answer results acquired by the answer result acquisition unit 33 are presented to the user via a website by the input / output processing unit 31.

[0039] Fig. 6 is a diagram showing an example of an answer result presentation screen displayed on the user terminal 2. The screen of Fig. 6 displays answer result presentation sections 111a to 111c, individual addition request input instruction sections 112a to 112c, a batch addition request input field 113, answer result selection sections 114a to 114c, a wall-hitting instruction section 115, etc.

[0040] The answer result presentation sections 111a to 111c are sections that display the answer results of each language model 3. The number of answer result presentation sections 111a to 111c displayed is the same as the number of language models 3 specified in advance. In this example, three language models 3-1 to 3-3 are specified, so three answer result presentation sections 111a to 111c are displayed. The answer result presentation section 111a corresponds to the language model 3-1, the answer result presentation section 111b corresponds to the language model 3-2, and the answer result presentation section 111c corresponds to the language model 3-3.

[0041] Each of the answer result presentation units 111a to 111c displays text indicating the user's request at the beginning, and the answers from language models 3-1 to 3-3 are displayed below the text indicating the request. In this example, the answer result presentation unit 111a corresponds to ChatGPT-4.1nano (language model 3-1), the answer result presentation unit 111b corresponds to Gemini 2.0Flash (language model 3-2), and the answer result presentation unit 111c corresponds to Gemini 2.5Flash (language model 3-3).

[0042] Furthermore, when the answer result acquisition unit 33 receives a further review request for the answer result from the user, it executes a process of transmitting a review instruction prompt including the review request to one or more language models 3 and acquiring a re-reviewed re-answer result. In this embodiment, the input / output processing unit 31 receives an additional review request from the user via a website, and the answer result acquisition unit 33 transmits a review instruction prompt including the received review request to the language model 3.

[0043] The individual addition request input instruction units 112a to 112c are on-screen operation buttons that allow the user to input a further review request for each language model 3. The individual addition request input instruction units 112a to 112c are set for each of the answer result presentation units 111a to 111c. When the individual addition request input instruction units 112a to 112c are operated, an input field 116 for inputting a question to the language model 3 is displayed. In this example, the individual addition request input instruction unit 112b is selected by the user, and the input field 116 for making a further review request is displayed. When the user inputs a review request in the input field 116, the answer result acquisition unit 33 acquires the review request through the input / output processing unit 31 and transmits a review instruction prompt including the review request to the corresponding language model 3 (in this example, Gemini 2.0 Flash (language model 3-2)). The re-answer result acquired from the language model 3 is displayed on the user terminal 2 in a format following the initial answer result of the answer result presentation units 111a to 111c.

[0044] The batch addition request input field 113 is a section where the user inputs a review request to all of the language models 3. When the user inputs a review request in the batch addition request input field 113 and operates the send button 125, the answer result acquisition unit 33 acquires the review request through the input / output processing unit 31 and transmits a review instruction prompt including the review request to all of the language models 3 (in this example, language models 3-1 to 3-3). The re-answer results acquired from each language model 3 are displayed by the input / output processing unit 31 in a format following the initial answer results of the corresponding answer result presentation units 111a to 111c.

[0045] The integration target setting unit 34 executes processing to exclude answer results excluded by the user from targets of integration instructions by the integration processing unit 35, or to set only answer results specified by the user as targets of integration. In this embodiment, when an operation to exclude an answer result by the user via a website is detected, the integration target setting unit 34 excludes the excluded answer result from targets of integration by the integration processing unit 35. The operation to exclude an answer result is performed via answer result selection units 114a to 114c.

[0046] The answer result selection sections 114a to 114c are operation buttons on the screen that allow the user to select the answer result to be used. The answer result selection sections 114a to 114c are set for each of the answer result presentation sections 111a to 111c. In this example, they are indicated by an x, and when selected by the user, the selected answer result presentation section 111a to 111c is deleted, and the integration target setting section 34 excludes the selected answer result presentation section 111a to 111c from the integration targets. For example, when the answer result selection section 114a corresponding to the answer result presentation section 111a is selected, the answer result presentation section 111a is deleted from the screen, and only the answer result presentation sections 111b and 111c are displayed, and the answer result of language model 3-1 is excluded from the integration targets.

[0047] 6, the integration target setting unit 34 performs a process of deleting answer results from the integration targets through selection by the answer result selection units 114a to 114c, but the process is not limited to this. For example, a check box may be provided for each of the answer result presentation units 111a to 111c, and the answer result with a check mark in the check box may be set as the answer result designated by the user, and the answer result with a check mark in the check box may be set as the integration target.

[0048] The integration processing unit 35 generates an integration instruction prompt that instructs the integration of the respective answer results of the multiple language models 3. In this embodiment, the integration processing unit 35 generates the integration instruction prompt based on the user selecting the hitting-against-the-wall instruction unit 115 shown on the screen in FIG. 6. Note that "integration" in this embodiment refers to the process of providing multiple answer results to a language model and generating a new answer by referring to these. The specific method of integration is left to the discretion of the language model, and as a result, it is intended to obtain a comprehensive answer that reflects the perspectives of the multiple answer results.

[0049] 7 is a diagram showing an example of the configuration of the integration instruction prompt. The integration instruction prompt includes a guideline transmission unit 120 that provides answer guidelines for multiple specified language models 3, and answer result transmission units 121 to 123 that transmit the answers of each language model.

[0050] The guideline transmitter 120 includes instructions such as "Analyze the AI ​​outputs listed below, extract the most important and useful information, and create one concise and comprehensive response," a first guideline "Read each AI output and extract the text, eliminating any erroneous or inaccurate information," a second guideline "Identify the main points of each text," a third guideline "Eliminate redundant information and integrate information to form a coherent message," and a fourth guideline "If necessary, restructure the information and extract and summarize key points in a clear and fluent manner."

[0051] The answer result transmission units 121 to 123 are units that show the answer results generated from the language models 3. In this example, the answer result transmission unit 121 shows the answer result of the language model 3-1, the answer result transmission unit 122 shows the answer result of the language model 3-2, and the answer result transmission unit 123 shows the answer result of the language model 3-3. In this way, the integration instruction prompt includes the answer results generated in each language model 3 along with the generation guidelines.

[0052] In this embodiment, the input / output processing unit 31 displays the integration instruction prompt generated by the integration processing unit 35 in the batch addition request input field 113. The integration instruction prompt shown in FIG. 7 is displayed in the batch addition request input field 113, allowing the user to confirm the contents of the integration instruction prompt. When the user operates the send button 125 in the batch addition request input field 113, the answer result acquisition unit 33 sends the integration instruction prompt to each specified language model 3. As described above, if the answer result acquisition unit 33 is performing a reexamination, the re-answer results will be included in the integration targets.

[0053] <Processing Flow> Next, a processing flow for presenting an answer using the language model 3 in response to a request from a user will be described with reference to Fig. 8. Fig. 8 is a flowchart showing an example of the flow of a processing for presenting an answer to a user, which is executed by the information processing device 1. Note that the flowchart in Fig. 8 is merely an example, and the processing order and processing contents may be changed.

[0054] In step S1, the language model designation unit 32 sets the language model 3 designated by the user from among a plurality of candidate language models 3 as the destination of the request prompt. The user may designate the language model 3 to be used from the language model selection screen of Fig. 5 displayed on the user terminal 2, or may use a pre-designated language model 3 as is.

[0055] In step S2, the answer result acquisition unit 33 acquires the request content input by the user. The request content is information entered in the request content input field 102 on the request input screen of FIG. 4, and is acquired via the input / output processing unit 31.

[0056] In step S3, the answer result acquisition unit 33 transmits a request prompt including the acquired request content to the specified plurality of language models 3.

[0057] In step S4, the answer result acquisition unit 33 acquires answer results from each language model 3 that has sent the request prompt, and transmits the answer results to the user terminal 2 via the input / output processing unit 31 to present them to the user. In this embodiment, the answer results from each language model 3 are displayed in each of the answer result presentation units 111a to 111c in FIG. 6.

[0058] 6 (Step S5; Yes), the answer result acquisition unit 33 executes the process of Step S6 in which it transmits a review instruction prompt including the contents of the reconsideration request to the corresponding language model 3. In the process of Step S6, if a review request is input into the input field 116 of the individual addition request input instruction units 112a to 112c, the review instruction prompt is transmitted only to the corresponding language model 3, whereas if a review request is input into the collective addition request input field 113, the review instruction prompt is transmitted to all the specified language models 3. After this process, the process proceeds to Step S7.

[0059] In step S5, if a review request has not been received from the user (step S5; No), the process proceeds to step S7 without passing through step S6.

[0060] In step S7, if an operation to remove an answer result from the user is received through the answer result selection units 114a to 114c in Fig. 6 (step S7; Yes), the integration target setting unit 34 executes the process of step S8, in which the answer result selected by the user is deleted from the integration targets. In the process of step S8, the answer result presentation units 111a to 111c are also deleted from the screen. After this process, the process proceeds to step S9.

[0061] In step S7, if an operation to exclude the answer result from the user is not received (step S7; No), the process proceeds to step S9 without going through step S8.

[0062] In step S9, if the user selects the hitting-to-wall instruction unit 115 in Fig. 6, the process proceeds to step S10 (step S9; Yes). In step S9, if the user does not select the hitting-to-wall instruction unit 115, the process returns to step S5 (step S9; No).

[0063] In step S10, the integration processing unit 35 generates an integration instruction prompt that instructs the specified plurality of language models 3 to integrate the answer results. In this embodiment, the integration instruction prompt is displayed in the batch addition request input field 113, and the user can check the content of the integration instruction prompt on the user terminal 2.

[0064] In step S11, the answer result acquisition unit 33 transmits an integration instruction prompt to the specified plurality of language models 3. The transmission of the integration instruction prompt may be triggered by the user's operation of the send button 125 as described above, or may be automatically executed when the user operates the hitting-to-wall instructing unit 115, triggered by the generation of the integration instruction prompt.

[0065] In step S12, the answer result acquisition unit 33 acquires the integrated answer result from each language model 3, and transmits the integrated answer result to the user terminal 2 via the input / output processing unit 31 to present it to the user. After the processing of step S12, the processing ends. In this embodiment, the integrated answer result from each language model 3 is displayed in each of the answer result presentation units 111a to 111c in FIG. 6.

[0066] As described above, the information processing device 1 of this embodiment includes an answer result acquisition unit 33 that sends a request prompt including the user's request content to a plurality of language models 3 (e.g., language models 3-1 to 3-3) and acquires answer results to the request content from each of the plurality of language models 3, and an integration processing unit 35 that generates an integration instruction prompt that instructs the language model 3 to integrate the answer results of each of the plurality of language models 3, and the answer result acquisition unit 33 acquires an integrated answer result that reflects the answer results of the plurality of language models 3 by sending the integration instruction prompt to the language model 3.

[0067] The flowchart in FIG. 8 illustrates the basic processing flow from the user's initial request to obtaining the integrated answer result. The user can continue to interact with the language model 3 even after the integrated answer result is presented in step S12. For example, the user can send additional questions or requests for consideration to each language model 3 individually through the individual addition request input instruction units 112a to 112c, or send additional requests for consideration collectively to all language models 3 through the bulk addition request input field 113. Furthermore, by operating the back-channel instruction unit 115 again, the integration process can be performed again based on new answer results or re-answer results. Thus, the information processing device 1 of this embodiment is configured to enable continuous dialogue even after the integrated answer result is presented, allowing the user to repeatedly interact with the language model 3 until a satisfactory result is obtained. This continuous dialogue function allows the user to confirm the integrated answer result and then dig deeper or ask questions from a new perspective, thereby obtaining a more accurate answer.

[0068] Furthermore, the control method of the information processing device 1 of this embodiment includes an answer result acquisition step of sending a request prompt including the content of the user's request to a plurality of language models 3 and acquiring an answer result for the request content from each of the plurality of language models 3; an integration processing step of generating an integration instruction prompt that instructs the language model 3 to integrate the answer results of each of the plurality of language models 3; and an integrated answer result acquisition step of sending the integration instruction prompt to the language model 3 to acquire an integrated answer result that reflects the answer results of the plurality of language models 3.

[0069] In addition, the program of this embodiment causes a computer to execute an answer result acquisition step of sending a request prompt including the user's request content to multiple language models 3 and acquiring an answer result for the request content from each of the multiple language models 3; an integration processing step of generating an integration instruction prompt that instructs the language model 3 to integrate the answer results of each of the multiple language models 3; and an integrated answer result acquisition step of sending the integration instruction prompt to the language model 3 to acquire an integrated answer result that reflects the answer results of the multiple language models 3.

[0070] In this way, the information processing device 1, the control method for the information processing device 1, or the program is configured, allowing a user to simultaneously pose questions to multiple language models 3 and compare their response results. By comparing the outputs of different language models 3, it is possible to obtain response results that take advantage of the strengths of each language model 3, and furthermore, it is possible to obtain an integrated response result, which is a more comprehensive response that integrates these response results. For example, even with the same prompt, GPT-4o may suggest mobile development or a thematic session, while Claude may suggest a speed challenge rally. Integrating such differences allows a broader perspective to be gained, enabling the user to acquire ideas from different perspectives during brainstorming. Furthermore, because the integration instruction prompt is automatically generated, the user can easily perform the necessary tasks to utilize the characteristics of multiple language models 3.

[0071] In addition, the information processing device 1 of this embodiment further includes a language model designation unit 32 that sets a language model 3 designated by the user from among multiple candidate language models 3 as the destination of the request prompt, and the answer result acquisition unit 33 sends the request prompt to the language model 3 set by the language model designation unit 32.

[0072] This allows the user to select language models 3 individually, thereby obtaining more appropriate answer results and integrated answer results by taking into consideration the content of the request to the language models 3 and the characteristics and features of each language model 3. For example, some language models 3 have a web search function, and by selecting such a language model 3, if the user wishes to provide answers including web search results as a target for bouncing questions, the user can generate answer results and integrated answer results that are more suited to the user's purpose.

[0073] In addition, the information processing device 1 of this embodiment further includes an integration target setting unit 34 that excludes answer results excluded by the user from the target of the integration instruction prompt, or that makes only the answer results specified by the user the target of the integration instruction prompt.

[0074] This allows the user to check the answer results and eliminate unnecessary ones, or to integrate only useful answer results, thereby realizing the generation of more accurate integrated answer results.

[0075] In addition, in this embodiment, when the answer result acquisition unit 33 receives a request from a user for further consideration of the answer result, it sends a consideration instruction prompt including the consideration request to one or more language models 3 to acquire the re-considered re-answer result, and the integration processing unit 35 generates an integration instruction prompt that instructs the integration of the answer result, the re-answer result, or both.

[0076] This allows the language model 3 to perform further review after viewing the answer result, thereby improving the accuracy of the answer result, and then generate an integrated answer result in the language model 3. Also, the accuracy of each answer result can be improved by individually drilling down into the answer result of each language model 3. Furthermore, since it is possible to perform re-examination on all of the multiple language models 3 at once, it is possible to save the effort of performing the same re-examination on each language model 3, and it is possible to further improve user convenience.

[0077] Second Embodiment Next, a configuration of a second embodiment, which is different from that of the first embodiment, will be described. Fig. 9 is a functional block diagram showing an example of the functional configuration of an information processing device 1a according to the second embodiment. Note that in the description of the second embodiment, components that are common or similar to those of the first embodiment may be assigned the same reference numerals and detailed description thereof may be omitted.

[0078] The information processing device 1a of the second embodiment is configured with the same hardware as that of the first embodiment (see FIG. 2). The information processing device 1a of the second embodiment includes an input / output processing unit 31, a language model designation unit 32, an answer result acquisition unit 33, and an integration processing unit 35 as functional units implemented on a processor (CPU 11). The information processing device 1a of the second embodiment differs from the information processing device 1 of the first embodiment in that it does not include an integration target setting unit 34.

[0079] The second embodiment differs from the first embodiment in that the processes from obtaining answer results from multiple language models 3 to integrating them are performed in the backend of the information processing device 1a. The answer presentation process in the second embodiment will be described with reference to Fig. 10. Fig. 10 is a schematic diagram showing the flow of data in the answer presentation process by the information processing device 1a of the second embodiment.

[0080] In step S21, the user displays a request input screen on the user terminal 2. Fig. 11 is a diagram showing an example of the request input screen displayed on the user terminal 2 of the second embodiment. The request input screen of the second embodiment is basically the same as that of the first embodiment. The request input screen of the second embodiment further displays a used language model display section 101, a request content input field 102, a language model designation operation section 103, a user information display section 104, and in addition, a setting instruction section 201 for the language model 3 that displays "Super AI."

[0081] In step S22, when the user operates the setting instruction unit 201, the language model designation unit 32 sets a plurality of pre-set language models 3 as destinations of the request prompt. Note that the language model 3 to be used may be designated by the user via the language model designation operation unit 103 or a language model selection screen such as that shown in FIG.

[0082] In step S23, the input / output processing unit 31 acquires the request content from the user. The request content is input by the user through the request content input field 102, similar to the first embodiment.

[0083] In step S24, the answer result acquisition unit 33 transmits a request prompt including the request content to the set plurality of language models 3. In this example, the answer result acquisition unit 33 transmits the request prompt to three language models 3-1 to 3-3.

[0084] In step S25, the answer result acquisition unit 33 acquires answer results from each of the language models 3-1 to 3-3. These answer results are not sent to the user terminal 2, and therefore, unlike the first embodiment, the answer results before integration are not presented to the user.

[0085] In step S26, the integration processing unit 35 generates an integration instruction prompt, and the answer result acquisition unit 33 transmits the integration instruction prompt to the set language model 3. In the second embodiment, this language model 3 is set in advance as the language model 3 to be integrated. Note that the language model 3 to be integrated may be selected by the user.

[0086] In step S27, the answer result acquisition unit 33 acquires the integrated answer result from the language model 3. In step S28, the input / output processing unit 31 transmits information for displaying the integrated answer result to the user terminal 2. In step S29, the integrated answer result is displayed on the user terminal 2.

[0087] In the second embodiment, similarly to the first embodiment, the placement of the language models 3-1 to 3-n can be flexibly configured. That is, some or all of the language models 3-1 to 3-n may be implemented within the information processing device 1a, and the language model 3 that performs the integration is not limited to an external service but may be placed within the system.

[0088] In the second embodiment, the processes from step S24 to step S28 are performed automatically in the backend, so that it appears to the user that one AI (language model 3) is providing the answer.

[0089] Although one embodiment of the present invention has been described above, the present invention is not limited to the above-described embodiment, and modifications, improvements, etc. within the scope of achieving the object of the present invention are included in the present invention.

[0090] An information display function for visually displaying the characteristics and strengths of the language model 3 may be added to the configurations of the first and second embodiments. For example, an icon indicating the characteristics may be displayed in the name of the language model 3 displayed on the request input screen or language model selection screen, or the characteristics of the language model 3 may be displayed by a hover function when the mouse cursor is placed over it.

[0091] Furthermore, conclusions or specific keywords may be highlighted in the answer results or integrated answer results. The specific keywords may be extracted from the words in the request content or may be set in advance.

[0092] In addition, automatic cooperation (sequence execution) between the language models 3 may be performed. For example, in an integration instruction prompt, an answer may be relayed from the language model 3-1 to the language model 3-2 to the language model 3-3 in this order to generate an integrated answer result. In addition, in an integration instruction prompt, a specific role may be set for each language model 3, and an integrated answer result may be generated in a discussion format.

[0093] Furthermore, the above-described series of processes can be executed by hardware or software. In other words, the above-described functional configuration is merely an example and is not particularly limited. That is, it is sufficient for the information processing device 1 to have the function of executing the above-described series of processes as a whole, and the type of functional block used to realize this function is not particularly limited to the above example. Furthermore, the location of the functional block is not particularly limited and may be arbitrary. For example, the functional block of the information processing device 1 may be transferred to another device, etc. Conversely, the functional block of another device may be transferred to a server, etc. Furthermore, one functional block may be configured as a single piece of hardware, a single piece of software, or a combination thereof.

[0094] When a series of processes is executed by software, the programs constituting the software are installed onto a computer or the like from a network or a recording medium. The computer may be a computer incorporated into dedicated hardware. The computer may also be a computer capable of executing various functions by installing various programs, such as a server, a general-purpose smartphone, or a personal computer.

[0095] The recording medium containing such a program may be configured as not only a removable medium (not shown) that is distributed separately from the device main body in order to provide the program, but also a recording medium that is provided in a state that it is pre-installed in the device main body, etc. Since the program can be distributed via a network, the recording medium may be installed in or accessible to a computer that is connected or connectable to the network.

[0096] In this specification, the steps describing the program recorded on the recording medium include not only processes that are performed in chronological order, but also processes that are not necessarily performed in chronological order but are performed in parallel or individually. Also, in this specification, the term "system" means an overall device composed of multiple devices or multiple means, etc.

[0097] REFERENCE SIGNS LIST 1, 1a Information processing device 2 User terminal 3-1 to 3-n Language model 31 Input / output processing unit 32 Language model designation unit 33 Answer result acquisition unit 34 Integration target setting unit 35 Integration processing unit 100 Language model utilization system

Claims

1. An information processing device comprising: an answer result acquisition unit that sends a request prompt including a user's request content to a plurality of language models and acquires an answer result for the request content from each of the plurality of language models; and an integration processing unit that generates an integration instruction prompt that instructs the language models to integrate the answer results of each of the plurality of language models, wherein the answer result acquisition unit acquires an integrated answer result that reflects the answer results of the plurality of language models by sending the integration instruction prompt to the language models.

2. The information processing device according to claim 1, further comprising a language model designation unit that sets the language model designated by the user from among a plurality of candidate language models as the destination of the request prompt, and the answer result acquisition unit transmits the request prompt to the language model set by the language model designation unit.

3. The information processing device according to claim 1, further comprising an integration target setting unit that excludes the answer results excluded by the user from the target of the integration instruction prompt, or that makes only the answer results specified by the user the target of the integration instruction prompt.

4. An information processing device as described in any one of claims 1 to 3, wherein the answer result acquisition unit, upon receiving a request from the user for further consideration of the answer result, sends a consideration instruction prompt including the consideration request to one or more of the language models to acquire a re-answer result that has been reconsidered, and the integration processing unit generates the integration instruction prompt that instructs the integration of the answer result, the re-answer result, or both.

5. A control method for an information processing device, comprising: an answer result acquisition step of sending a request prompt including a user's request content to a plurality of language models and acquiring an answer result for the request content from each of the plurality of language models; an integration processing step of generating an integration instruction prompt that instructs the language models to integrate the answer results of each of the plurality of language models; and an integrated answer result acquisition step of sending the integration instruction prompt to the language models to acquire an integrated answer result in which the answer results of the plurality of language models are reflected.

6. A program for causing a computer to execute the following steps: an answer result acquisition step of sending a request prompt including a user's request content to a plurality of language models and acquiring an answer result for the request content from each of the plurality of language models; an integration processing step of generating an integration instruction prompt that instructs the language model to integrate the answer results of each of the plurality of language models; and an integrated answer result acquisition step of sending the integration instruction prompt to the language model to acquire an integrated answer result that reflects the answer results of the plurality of language models.

Images