Support device, program, recording medium, and method for creating materials

Abstract

Enables document creation through collaboration between humans and AI. [Solution] This is a device that generates materials in multiple steps while accepting user operations and inputs based on basic data, and is equipped with a first acquisition unit for acquiring basic data, a display unit that displays a screen for inputting review result data including answers to questions asked to the user at the current step regarding the basic data at each step, a second acquisition unit that acquires the review result data input by the user via the screen at each step, and a generation unit that causes the LLM to generate candidates for the review result at the current step regarding the basic data or the basic data and at least some of the review result data acquired by the second acquisition unit at at least some steps prior to the current step, and if candidates for the current step have been generated by the generation unit, the display unit can provide an operation on the screen that allows the user to include some or all of the candidates for the current step in the review result data for the current step.

Description

[Technical Field]

[0001] The present invention relates to a support device, a program, a recording medium, and a method for creating materials. [Background technology]

[0002] In the business field, it is common to create documents for new projects, such as new business plans. However, it is not easy to create such documents so that they can be evaluated as business documents. Therefore, there is a demand for technology that supports the creation of documents that can be evaluated as business documents.

[0003] In relation to technologies for supporting the creation of materials that are evaluated as business documents in conventional services and conventional technologies, Patent Document 1 discloses a creation support device that receives a new inquiry letter including information on required specifications for bidding in a project such as plant design and construction, and supports the creation of a proposal including information on bidding specifications that correspond to the new inquiry letter, the creation support device comprising: a calculation unit that calculates a similarity between the content of the new inquiry letter and the content of past inquiries or past proposals, and uses the similarity to extract content that is useful for creating the proposal from history data associated with the past inquiries; and a providing unit that uses the results of the calculation unit to generate and provide auxiliary data to be used in creating the proposal.

[0004] The technology described in Patent Document 1 can efficiently understand the specifications presented when bidding is requested, and can support bidders in creating accurate proposals based on the specifications. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Patent Publication No. 2021-120803 Summary of the Invention [Problem to be solved by the invention]

[0006] In bidding for projects such as plant design and construction, specifications similar to past specifications are expected to be submitted. Therefore, the more similar the proposal for such a bid is to the quotations for past specifications similar to the current specifications, the more likely it is to be successful in the bid and is expected to be evaluated as a business document.

[0007] On the other hand, new projects are required to be different from past projects. Therefore, if the documents for a new project are similar to those for past projects, it is not guaranteed that they will be evaluated as business documents.

[0008] The technology of Patent Document 1 uses similarity to extract content useful for creating proposals from historical data associated with past inquiries. Therefore, the technology of Patent Document 1 can only support the creation of proposals for bidding on projects such as plant design and construction, and it is believed that there is room for further improvement in terms of supporting the creation of materials for new projects.

[0009] The objective of the present invention is to support the creation of materials through collaboration between humans and artificial intelligence (AI). [Means for solving the problem]

[0010] As a result of intensive research into solving the above-mentioned problems, the inventors have found that the above-mentioned object can be achieved by, for example, providing an operation to include, in an input screen for inputting the results of a study on a new project, candidate study results generated by a large-scale language model (LLM) based on basic data for the new project and data on previously input study results. As a result, the inventors have completed the present invention.

[0011] According to an embodiment of the present invention, Based on the basic data, multiple review processing steps are performed while accepting operations and inputs from users. cumulatively A support device for generating materials using a basic data acquisition unit that acquires the basic data; an input screen display unit that commands the display of an input screen for inputting review result data including answers to questions posed to a user in a current review processing step regarding the basic data in each review processing step; a review result acquisition unit that acquires the review result data input by the user via the input screen in each review processing step; In each review processing step, the basic data, or the basic data and the review result data acquired by the review result acquisition unit in at least a part of the review processing steps prior to the current review processing step. Or candidate results of the study and a candidate for the review result (hereinafter referred to as "candidate for review result") including an answer to a question posed to the user in the current review processing step regarding at least a part of the above. As candidates for the results of each examination process step a study result candidate generation unit that generates a large-scale language model; Equipped with the input screen display unit is configured to provide, when the consideration result candidates for the current consideration processing step are generated by the consideration result candidate generation unit, an operation by the user to include part or all of the consideration result candidates for the current consideration processing step in the consideration result data for the current consideration processing step on the input screen; the consideration result candidate generation unit causes the large-scale language model to generate consideration result candidates including answers to questions posed to the user in the (n+1)th consideration processing step regarding consideration result candidates in the nth consideration processing step and consideration result data selected by the user, as consideration result candidates in the (n+1)th consideration processing step; a data generation unit that causes the large-scale language model to generate data that includes at least a portion of the basic data and the review result data or review result candidates after the review result data or review result candidates in each review processing step are obtained, When causing the large-scale language model to generate the consideration result candidates, the consideration result candidate generation unit performs the following on the large-scale language model: The basic data, or the basic data and the review result data acquired by the review result acquisition unit in at least a part of the review processing steps prior to the current review processing step and one of the candidate results At least a portion of the above (hereinafter referred to as "input data"); The content of the question (hereinafter referred to as the "specific question") to the user in the current review processing step regarding the input data, instructions for generating an answer to the specific question based on the input data; Enter death, When causing the large-scale language model to generate the material, the material generation unit The basic data, or the basic data and at least a portion of the review result data or review result candidates acquired by the review result acquisition unit in at least a portion of the review processing steps included in all of the review processing steps (hereinafter referred to as "total input data"); instructions to generate the material based on the total input data; Enter the A document creation support device is provided.

[0012] The input screen display unit of this embodiment displays an input screen for inputting the results of consideration for the new project. In this support device, the material generation unit uses a large-scale language model to generate materials for the new project based on the basic data for the new project acquired by the basic data acquisition unit and the consideration result data entered from this input screen. However, it is not easy for users who are unfamiliar with creating materials for new projects to appropriately list and input the necessary consideration items.

[0013] Here, the input screen display unit of this embodiment displays an input screen offering an operation for including some or all of the consideration result candidates generated by the consideration result candidate generation unit by inputting a prompt including basic data and the consideration result data into a large-scale language model. That is, the assistance device of this embodiment presents the consideration result candidates proposed by the large-scale language model to the user, allowing the user to add some or all of them to their own consideration results. This allows the user to add the AI's consideration results to their own consideration results, enabling co-creation between humans and AI. The assistance device of this embodiment then generates materials for new projects through this co-creation. While there is a concern that relying solely on AI to generate materials for new projects will result in materials similar to those created in the past, the assistance device of this embodiment, through co-creation between humans and AI, can provide materials that reflect consideration results that differ from past materials input by humans.

[0014] As described above, this aspect can support the efficient creation of materials and the improvement of their quality through co-creation between humans and AI.

[0015] In addition, the present invention can take various forms, exemplified by a form that advances co-creation between humans and AI through questions posed to the inputter, a form that allows input of review results divided into three stages: advantages and prospects for a new project, concerns about the new project, and measures to overcome the concerns, a form that shows an AI's evaluation of the generated materials, a form that enables co-creation with others in addition to co-creation with AI, etc. These forms, by virtue of the effects brought about by the addition of their respective unique configurations, contribute to supporting the efficient creation of materials for new projects and the improvement of their quality through co-creation between humans and AI. [Effects of the Invention]

[0016] As described above, the present invention can support the efficient creation of materials and the improvement of their quality through co-creation between humans and AI. [Brief explanation of the drawings]

[0017] [Figure 1] 1 is a block diagram showing an example of the hardware configuration and software configuration of a system S of this embodiment. [Figure 2] 1 is a main flowchart showing an example of a preferable flow of the assistance process executed by the assistance device 1 of this embodiment. [Figure 3] This is a continuation of the previous figure. [Figure 4] 10 is a flowchart of a review process in the main flowchart. [Figure 5] This is a continuation of the previous figure. [Figure 6] 10 is a display example of an input screen E. [Figure 7] 10 is a display example of an evaluation screen A. [Figure 8] FIG. 10 is a diagram showing a partial configuration of a support device according to a second embodiment. [Figure 9] FIG. 10 is a diagram showing the remaining part of the configuration of the support device according to the second embodiment. [Figure 10] FIG. 10 is a diagram showing a screen displayed by the support device according to the second embodiment. [Figure 11] FIG. 10 is a diagram showing a screen displayed by the support device according to the second embodiment. [Figure 12] FIG. 10 is a diagram showing a screen displayed by the support device according to the second embodiment. [Figure 13] FIG. 10 is a diagram showing a screen displayed by the support device according to the second embodiment. [Figure 14] FIG. 10 is a diagram showing a screen displayed by the support device according to the second embodiment. [Figure 15] FIG. 10 is a diagram showing a screen displayed by the support device according to the second embodiment. [Figure 16] FIG. 10 is a diagram showing the remaining part of the configuration of the support device according to the second embodiment. [Figure 17] FIG. 10 is a diagram showing a screen displayed by the support device according to the second embodiment. [Figure 18] FIG. 10 is a diagram showing a screen displayed by the support device according to the second embodiment. [Figure 19] FIG. 10 is a diagram showing a screen displayed by the support device according to the second embodiment. [Figure 20] FIG. 10 is a diagram showing a screen displayed by the support device according to the second embodiment. [Figure 21] FIG. 10 is a diagram showing a screen displayed by the support device according to the second embodiment. [Figure 22] FIG. 10 is a diagram showing a screen displayed by the support device according to the second embodiment. [Figure 23] FIG. 10 is a diagram showing a screen displayed by the support device according to the second embodiment. [Figure 24] FIG. 10 is a diagram showing a screen displayed by the support device according to the second embodiment. [Figure 25] FIG. 10 is a diagram showing a screen displayed by the support device according to the second embodiment. [Figure 26] FIG. 10 is a diagram showing a screen displayed by the support device according to the second embodiment. [Figure 27] FIG. 10 is a diagram showing a screen displayed by the support device according to the second embodiment. [Figure 28] FIG. 10 is a diagram showing a screen displayed by the support device according to the second embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0018] First, although the following disclosure, diagrams, and / or claims may be described as being presented alone or in combination with one or more other aspects, the subject matter of the immediate disclosure is not intended to be so limited. That is, the immediate disclosure, diagrams, and claims are intended to encompass the various aspects described herein, each alone or in one or more combinations with each other. For example, even if the immediate disclosure describes and illustrates a first, second, and third embodiment in such a way that the first embodiment is described and illustrated specifically in conjunction with the second embodiment, or the second embodiment is described and illustrated only in conjunction with the third embodiment, the immediate disclosure and illustrations are not so limited and may include only the first embodiment, only the second embodiment, only the third embodiment, or one or more combinations of the first, second, and / or third embodiments, such as the first and second embodiments, the first and third embodiments, the second and third embodiments, or the first, second, and third embodiments.

[0019] The use of the phrase "or" in this document shall mean a "non-exclusive" arrangement unless expressly specified otherwise. For example, when we say "item x is A or B," we mean either: (1) item x is either A or B, but not both; or (2) item x is both A and B. In other words, the word "or" is not used to define an "exclusive" arrangement.

[0020] Additionally, the phrases "comprising at least one of" and "comprising at least one of the following," when used in conjunction with a system or element, mean that the system or element includes one or more of the elements listed after the phrase. For example, if there are three types of elements, element 1 through element 3, the phrases "comprising at least one of" and "comprising at least one of the following" are to be interpreted as any of the following structural arrangements: a device including the first element, a device including the second element, a device including the third element, a device including the first and second elements, a device including the first and third elements, a device including the second and third elements, or a device including the first, second, and third elements.

[0021] A similar interpretation is intended when the phrase "used in at least one of the following" is used in this context. Furthermore, as used in this context, "and / or" is used as a verbal conjunction to indicate that one or more of the listed elements or conditions are included or occur. For example, a device including a first element, a second element, and / or a third element is to be interpreted as any of the following structural arrangements: a device including the first element, a device including the second element, a device including the third element, a device including the first element and the second element, a device including the first element and the third element, a device including the second element and the third element, or a device including the first element, the second element, and the third element.

[0022] In addition, the use of the phrase "and / or" in the text means a "non-exclusive" agreement, as stipulated in the Japanese Industrial Standards (JIS) "Format and preparation method of standard sheets JIS Z 8301."

[0023] Hereinafter, an example of an embodiment of the present invention will be described in detail with reference to the drawings.

[0024] [First embodiment] <System S>

[0025] 1 is a block diagram showing an example of the hardware and software configurations of a system S according to this embodiment. The support system (system S) for creating materials for new projects according to this embodiment includes a support device 1 for creating materials for new projects (hereinafter also simply referred to as "support device 1"). The support device 1 is configured to communicate with a terminal T via a network N. [New project materials]

[0026] In this embodiment, the "materials for a new project" include, for example, materials such as drafts or strategic review materials related to new projects such as new business development or project development. In addition, the "materials for a new project" may also be materials related to the consideration of future visions, problem identification, idea creation, etc. [Support device 1]

[0027] The assistance device 1 includes various hardware components such as a control unit 11, a storage unit 13, and a communication unit 14. The assistance device 1 provides a unique user interface and processing related to a large-scale language model, thereby assisting in the creation of materials for new projects through co-creation between humans and AI. The type of assistance device 1 is not particularly limited, and may be, for example, a server device, a cloud server, or the like. [Control unit 11]

[0028] The control unit 11 includes a central processing unit (CPU), a random access memory (RAM), a read only memory (ROM), and the like.

[0029] The control unit 11 cooperates with at least one of the storage unit 13 and the communication unit 14 as necessary. The control unit 11 then realizes the software components of the program of this embodiment executed by the support device 1, such as a basic data acquisition unit 111, an input screen display unit 112, a review result acquisition unit 113, a review result candidate generation unit 114, an information provision unit 115, a collaborator review result acquisition unit 116, a material generation unit 117, an evaluation generation unit 118, and a comparison display command unit 119. The control unit 11 may be capable of realizing its functions by a processor such as a CPU executing a program recorded on a computer-readable recording medium.

[0030] The support processes realized by the above-mentioned software components will be described in detail later with reference to FIGS. [Storage section 13]

[0031] The memory unit 13 is a device in which data and / or files are stored, and has a storage unit that non-temporarily stores data using a hard disk, semiconductor memory, recording medium, memory card, etc. The memory unit 13 stores programs and the like that are executed by the microcomputer. (Large-scale language model)

[0032] The large-scale language model of this embodiment is not particularly limited. Examples of the large-scale language model of this embodiment include GPT-4 and GPT-4o used in ChatGPT (registered trademark). Note that, hereinafter, the large-scale language model may be simply referred to as "AI." [Communications Section 14]

[0033] The communication unit 14 is not particularly limited as long as it can connect the support device 1 to the network N and enable communication. Examples of the communication unit 14 include a network card compatible with the Ethernet standard and a communication device compatible with wireless LAN. [Network N]

[0034] The type of the network N is not particularly limited as long as it enables communication between the support device 1, etc. The type of the network N is, for example, the Internet, a mobile phone network, a wireless LAN, etc. [Terminal T]

[0035] Terminal T executes processes such as transmitting data related to user input to the support device 1 and performing various displays based on display-related commands received from the support device 1. The type of terminal T is not particularly limited, and may be, for example, a stationary terminal such as a personal computer, or various terminals such as a mobile terminal such as a tablet terminal. [Main flowchart of support processing]

[0036] Fig. 2 is a main flowchart showing an example of a preferred flow of the support processing executed by the support device 1 of this embodiment. Fig. 3 is a diagram continuing from the previous figure. The following is an example of a preferred flow of the support processing executed by the support device 1 of this embodiment using Figs. 2 and 3. [Step S1: Obtain basic data]

[0037] The control unit 11 cooperates with the storage unit 13 and the communication unit 14 to execute the basic data acquisition unit 111. Then, the control unit 11 executes a process of acquiring basic data of a new project by the basic data acquisition unit 111 (basic data acquisition step). The control unit 11 shifts the process to step S2.

[0038] In the basic data acquisition step, the basic data acquisition unit 111 acquires basic data input at, for example, terminal T. If some or all of the basic data is stored in advance in the storage unit 13 to enable advance input or reuse of the basic data, the basic data acquisition unit 111 may acquire the data. The basic data acquisition unit 111 may also acquire the data from the Internet. In this case, the basic data acquisition unit 111 searches the Internet using search words based on the headline of the new project input at terminal T, for example. (Basic data for new projects)

[0039] The basic data of the new project acquired in the basic data acquisition step includes, for example, a headline related to the new project, a basic idea, a target market, strengths of the project, and assets of the project.

[0040] After acquiring the basic data, the support process executes a review process in which humans and AI work together to conduct various reviews related to the new plan. The following is an example of a case in which the review process is divided into a first review process, a second review process, and a third review process. Note that the number of divisions into the review process and the content of each of the divided partial review processes are not limited to the following example. [Step S2: First examination process]

[0041] The control unit 11 cooperates with the storage unit 13 and the communication unit 14 to execute a first review process for the advantages and prospects of the new project (first review step). The control unit 11 moves the process to step S3. In the first review process, the input screen display unit 112, the review result acquisition unit 113, and the review result candidate generation unit 114, etc., perform co-creation between humans and AI for reviewing the advantages and prospects of the new project. Furthermore, the first review process may take a form in which the information provision unit 115 and the collaborator review result acquisition unit 116 support co-creation with further collaborators. Details of the first review process will be described later using Figures 4 and 5. [Step S3: Second Consideration Process]

[0042] The control unit 11, in cooperation with the storage unit 13 and the communication unit 14, executes a second review process related to the concerns of the new project (second review step). The control unit 11 moves the process to step S4. In the second review process, the input screen display unit 112, the review result acquisition unit 113, and the review result candidate generation unit 114, etc., perform co-creation between humans and AI related to the review of the concerns of the new project. Furthermore, the second review process may take a form in which the information provision unit 115 and the collaborator review result acquisition unit 116 support co-creation with further collaborators. Details of the second review process will be described later using Figures 4 and 5. [Step S4: Third examination process]

[0043] The control unit 11, in cooperation with the storage unit 13 and the communication unit 14, executes a third examination process related to measures to overcome the above-mentioned concerns (third examination step). The control unit 11 moves the process to step S5. In the third examination process, the input screen display unit 112, the examination result acquisition unit 113, the examination result candidate generation unit 114, etc., perform co-creation between humans and AI related to the examination of measures to overcome the above-mentioned concerns. Furthermore, the third examination process may take a form in which the information provision unit 115 and the collaborator examination result acquisition unit 116 support co-creation with further collaborators. Details of the third examination process will be described later using Figures 4 and 5. [Step S5: Generate materials]

[0044] The control unit 11 cooperates with the storage unit 13 and the communication unit 14 to execute the material generation unit 117. Then, the control unit 11 inputs a prompt including basic data and review result data related to the review process into the large-scale language model using the material generation unit 117, and executes a process of generating materials (material generation step). The control unit 11 moves the process to step S6. The review result data related to the review process here refers to data that summarizes the review result data related to each of the first to third review processes. [Editing Steps]

[0045] To make the document appear as if it were being completed by a human, the assistance process preferably further includes an editing step of editing the generated document based on user input. In this case, the assistance device 1 preferably includes a display that visualizes the degree of contribution of AI to the edited document. This allows the assistance device 1 to increase the reliability of the generated and edited document and further encourage the user to proceed in a way that does not rely solely on AI generation. [Work time display step]

[0046] To make the review process transparent, it is preferable that the support process display the amount of time spent by the user on each of the first review process, second review process, third review process, and process related to creating and editing materials. This allows the user to see how much time was spent on each stage and refine their review process. [Explanation Addition Step]

[0047] To demonstrate that the AI's review process is logical, it is preferable that the support process display text explaining the data sources relied upon by the AI ​​and the logic used by the AI ​​for each of the first review process, second review process, third review process, and process related to the generation and editing of materials. This allows users to refer to how the AI ​​conducted its review at each stage and refine their review process.

[0048] The support process preferably includes a series of processes for displaying an evaluation of the generated materials, steps S6 to S9 being an example of such processes. [Step S6: Determine whether to evaluate]

[0049] The control unit 11 cooperates with the storage unit 13 and the communication unit 14 to execute the evaluation generation unit 118. Then, the control unit 11 executes a process of determining whether to evaluate the generated material by the evaluation generation unit 118 (evaluation determination step). If the control unit 11 determines that the material should be evaluated, it proceeds to step S7, and if not, it proceeds to step S9.

[0050] The evaluation generating unit 118 realizes the above-mentioned determination by, for example, a procedure of determining that an evaluation should be made if data requesting an evaluation has been received from the terminal T. [Step S7: Generate evaluation]

[0051] The control unit 11 executes a process in which the evaluation generation unit 118 inputs the material generated by the material generation unit 117 and a prompt including an index of the evaluation into the large-scale language model, and generates an evaluation for the material (evaluation generation step). The control unit 11 proceeds to step S8. (Evaluation indicators)

[0052] The evaluation index for the evaluation generation step includes, for example, text indicating the items to be evaluated and the score range. To stabilize the evaluation, the index preferably includes text indicating the correspondence between the contents described in the materials and the scores. To enable evaluations from various perspectives, it is preferable that the evaluation index be selectable from multiple sets. [Step S8: Display the evaluation]

[0053] The control unit 11 cooperates with the storage unit 13 and the communication unit 14 to execute the comparison display command unit 119. Then, the control unit 11 executes a process of commanding the comparison display command unit 119 to display the evaluation of the material (display command step). The control unit 11 moves the process to step S9. (Regarding comparative display)

[0054] The comparison display command unit 119 preferably commands a comparative display of the generated evaluations of the plurality of materials, allowing the user to consider what kind of material will receive a higher evaluation based on the comparatively displayed evaluations.

[0055] The plurality of materials preferably includes a history of materials generated by the same user, allowing the user to consider which materials would receive a higher rating based on differences in inputs in or between past generations and their respective ratings.

[0056] It is preferable that the plurality of materials include materials from different users. This allows the user to incorporate the knowledge of other users who are expected to have different background knowledge or viewpoints. This also allows the user to feel competitive with other users and use it as motivation. A specific example of comparative display will be explained later using FIG. 7. [Step S9: Determine whether to return to the first review]

[0057] The control unit 11, in cooperation with the storage unit 13 and the communication unit 14, executes a process of determining whether to return to the first review (first review continuation determination step). If the control unit 11 determines that the process should be returned, it shifts the process to step S2, and if not, it shifts the process to step S10.

[0058] The control unit 11 realizes the above-mentioned determination by, for example, a procedure of determining to return if data requesting continuation or redoing of the first examination has arrived from the terminal T. [Step S10: Determine whether to return to the second examination]

[0059] The control unit 11, in cooperation with the storage unit 13 and the communication unit 14, executes a process of determining whether to return to the second review (second review continuation determination step). If the control unit 11 determines that the process should be returned, it shifts the process to step S3, and if not, it shifts the process to step S11.

[0060] The control unit 11 realizes the above-mentioned determination by, for example, a procedure of determining to return if data requesting continuation or redoing of the second examination has arrived from the terminal T. [Step S11: Determine whether to return to the third consideration]

[0061] The control unit 11, in cooperation with the storage unit 13 and the communication unit 14, executes a process of determining whether to return to the third examination (third examination continuation determination step). If the control unit 11 determines that it will return, it moves the process to step S4, and if it does not, it returns the process to step S1 and repeats the processes from step S1 to step S11.

[0062] The control unit 11 realizes the above-mentioned determination by, for example, a procedure of determining to return if data requesting continuation or redo of the third examination has arrived from the terminal T. [Flowchart of review process]

[0063] FIG. 4 is a flowchart of the review process in the main flowchart. FIG. 5 is a continuation of the previous figure. The following is an example of a preferred flow of the review process using FIGS. 4 and 5. The review process described below corresponds to various review processes such as the first review process (step S2), the second review process (step S3), and the third review process (step S4). Therefore, in the following explanation of each step, we will first explain the common parts of the first review process (step S2), the second review process (step S3), and the third review process (step S4), and then explain the parts that differ between each process. [Step S21: Command to display input screen]

[0064] The control unit 11 cooperates with the storage unit 13 and the communication unit 14 to execute the input screen display unit 112. Then, the control unit 11 executes a process of instructing the input screen display unit 112 to display an input screen for inputting the review result data related to the new plan (input screen display command step). The control unit 11 shifts the process to step S22. (If the step to generate candidate results has already been executed)

[0065] When the consideration result candidate generation step (described later) has been executed, the input screen display unit 112 commands the display of an input screen that provides an operation for including some or all of the consideration result candidates generated in this step in the input of consideration result data. Such an input screen is realized, for example, by a user interface element that displays a list of consideration result candidates and pastes some or all of the candidates into a field where the consideration results are to be input, exemplified by a copy button, a paste button, etc. Note that when consideration result candidates including questions for the inputter have been generated, the input screen is preferably configured to allow the input of consideration result data including answers to the questions.

[0066] The input screen preferably includes an operation for not including some or all of the consideration result candidates in the input of the consideration result data. In addition, in order to reduce the user's effort involved in co-creation, it is preferable that the operation for including some or all of the consideration result candidates in the input of the consideration result data be an operation that can be completed with a single basic operation procedure, such as one click. This allows the user to switch between data to be handed over to the AI ​​and data not to be handed over with a single operation.

[0067] In order to be able to incorporate AI suggestions for each cohesive candidate, when the review result candidates are divided into multiple groups, it is preferable that the input screen display unit 112 instructs the display of an input screen that provides an operation to select one of the multiple divided candidates and include some or all of them in the input of the review result data. (In the case of the first review process)

[0068] In the first review process, the input screen display unit 112 commands the display of a first input screen for inputting first review result data including advantages and prospects of the new project. This enables the support device 1 to support co-creation between humans and AI regarding the advantages and prospects of the new project. (In the case of the second review process)

[0069] In the second examination process, the input screen display unit 112 issues a command to display a second input screen for inputting second examination result data including concerns about the new project. This enables the support device 1 to support co-creation between humans and AI regarding concerns about the new project. (In the case of the third review process)

[0070] In the third examination process, the input screen display unit 112 commands the display of a third input screen for inputting third examination result data including measures to overcome the concerns. This enables the support device 1 to support co-creation between humans and AI regarding measures to overcome the concerns of a new project. [Step S22: Obtain the examination results]

[0071] The control unit 11 cooperates with the storage unit 13 and the communication unit 14 to execute the review result acquisition unit 113. Then, the control unit 11 executes a process of acquiring the review result data input via the above-mentioned input screen by the review result acquisition unit 113 (review result acquisition step). The control unit 11 shifts the process to step S23. (In the case of the first review process)

[0072] In the first review process, the review result acquisition unit 113 acquires the above-mentioned first review result data. (In the case of the second review process)

[0073] In the second review process, the review result acquisition unit 113 acquires the above-mentioned second review result data. (In the case of the third review process)

[0074] In the third review process, the review result acquisition unit 113 acquires the above-mentioned third review result data. [Step S23: Determine whether to generate study result candidates]

[0075] The control unit 11 cooperates with the storage unit 13 and the communication unit 14 to execute the consideration result candidate generation unit 114. Then, the control unit 11 executes a process of determining whether to generate consideration result candidates, which are candidates for the above-mentioned consideration results, by the consideration result candidate generation unit 114 (consideration result candidate generation determination step). If the control unit 11 determines that a consideration result candidate should be generated, it moves the process to step S24, and if not, it moves the process to step S25. [Step S24: Generate candidate study results]

[0076] The control unit 11 executes a process of inputting a prompt including the above-mentioned basic data or a prompt including the above-mentioned basic data and the above-mentioned consideration result data into a large-scale language model by the consideration result candidate generation unit 114, and generating candidates for the above-mentioned consideration results (consideration result candidates) (consideration result candidate generation step). The control unit 11 proceeds to step S25.

[0077] In the consideration result candidate generation step, the consideration result candidate generation unit 114 inputs, for example, prompts including input data from among the basic data and various consideration result data, into the large-scale language model. (Questions for the person entering the data)

[0078] In the consideration result candidate generation step, the consideration result candidate generation unit 114 preferably generates consideration result candidates including questions to the person inputting data regarding data including basic data or consideration result data. This enables the support device 1 to promote co-creation between humans and AI through questions to the person inputting data. (In the case of the first review process)

[0079] In the first review process, the review result candidate generation unit 114 inputs a prompt including basic data into the large-scale language model to generate candidates for the above-mentioned first review results (first review result candidates). This allows the review result candidate generation unit 114 to generate candidates for review results related to the advantages and prospects of a new project based on the basic data. If first review result data has already been input, the review result candidate generation unit 114 may input a prompt that further includes the already-input first review result data into the large-scale language model. This allows the review result candidate generation unit 114 to generate candidates for review results related to the advantages and prospects of a new project by utilizing the already-input first review result data. (In the case of the second review process)

[0080] In the second review process, the review result candidate generation unit 114 inputs a prompt including the basic data and the first review result data into the large-scale language model to generate candidates for the second review result (second review result candidates). This allows the review result candidate generation unit 114 to generate candidates for review results related to the concerns of the new project based on the basic data and the first review result data. If the second review result data has already been input, the review result candidate generation unit 114 may input a prompt that further includes the already-input second review result data into the large-scale language model. This allows the review result candidate generation unit 114 to generate candidates for review results related to the concerns of the new project by utilizing the already-input second review result data. (In the case of the third review process)

[0081] In the third review process, the review result candidate generation unit 114 inputs a prompt including the basic data, the first review result data, and the second review result data into the large-scale language model to generate candidates for the third review result (third review result candidates). As a result, the review result candidate generation unit 114 can generate candidates for review results related to measures for overcoming concerns about the new project based on the basic data, the first review result data, and the second review result data. If the third review result data has already been input, the review result candidate generation unit 114 may input a prompt that further includes the already-input third review result data into the large-scale language model. As a result, the review result candidate generation unit 114 can generate candidates for review results related to measures for overcoming concerns about the new project by utilizing the already-input third review result data.

[0082] The review process preferably includes a series of processes related to co-creation with a collaborator different from the person who inputs the data such as the basic data and the review result data. Steps S25 to S28 are an example of such processes. [Step S25: Determine whether to provide information to collaborators]

[0083] The control unit 11 cooperates with the storage unit 13 and the communication unit 14 to execute the information providing unit 115. Then, the control unit 11 executes a process of determining whether to provide information to a collaborator other than the person who inputted the data, using the information providing unit 115 (information provision determination step). If the control unit 11 determines that information should be provided, it proceeds to step S26, and if not, it proceeds to step S28.

[0084] The control unit 11 realizes the above-mentioned determination by, for example, a procedure of determining that the information is to be provided if a helper has been designated from the terminal T. [Step S26: Provide to collaborators]

[0085] The control unit 11 executes a process of providing data including the basic data or the examination result data to a collaborator other than the person who inputted the data (information providing step) by the information providing unit 115. The control unit 11 moves the process to step S27. [Step S27: Obtain collaborator review results]

[0086] The control unit 11 cooperates with the storage unit 13 and the communication unit 14 to execute the collaborator review result acquisition unit 116. Then, the control unit 11 executes a process of acquiring the collaborator review results by the collaborators using the collaborator review result acquisition unit 116 (collaborator review result acquisition step). The control unit 11 shifts the process to step S28. (Providing collaborator review results on the input screen)

[0087] On the input screen, the collaborator review results are preferably provided in a manner that provides an operation for including a part or all of them in the input of review result data, similar to the review result candidates. [Step S28: Determine whether to continue the study]

[0088] The control unit 11 executes a process of determining whether to continue the review in cooperation with the storage unit 13 and the communication unit 14 (a review continuation determination step). If the control unit 11 determines to continue, the process returns to step S21, and if not, the process proceeds to the step following the step that transitioned to the review process. (In the case of the first review process)

[0089] In the case of the first examination process, if it is not determined to continue, the transition destination is step S3. (In the case of the second review process)

[0090] In the case of the second examination process, if it is not determined to continue, the transition destination is step S4. (In the case of the third review process)

[0091] In the case of the third examination process, if it is not determined to continue, the transition destination is step S5. [Effects of support processing]

[0092] In the support process, the basic data acquisition unit 111 acquires basic data of a new project (step S1). The input screen display unit 112 issues a command to display an input screen E for inputting the results of consideration of the new project (step S21).

[0093] However, for users who are not sufficiently skilled in creating materials for new projects, it is not easy to properly list the various points that need to be considered and input the results of their consideration of those points. On the other hand, AI can output a large amount of consideration results in a short time compared to humans. AI also has a wider range of knowledge than humans. Therefore, if human ideas and choices can be effectively combined with AI consideration results, it is possible to create excellent materials for new projects in a short time through co-creation between humans and AI.

[0094] To realize such co-creation, the consideration result candidate generation unit 114 inputs a prompt including basic data and consideration result data into a large-scale language model to generate consideration result candidates (steps S23 to S24). Then, the input screen display unit 112 displays an input screen that provides an operation to include some or all of the consideration result candidates in the consideration result data (step S21).

[0095] Then, the consideration result acquisition unit 113 acquires the consideration result data input from the input screen (step S22). In the support device 1, the material generation unit 117 generates materials for the new project using a large-scale language model based on the basic data of the new project acquired by the basic data acquisition unit 111 and the consideration result data input from this input screen (step S5).

[0096] That is, the support device 1 presents the user with candidate study results proposed by the large-scale language model, and allows the user to add some or all of them to the results of their own study. This allows the user to add the results of the AI's study to their own study results, enabling co-creation between humans and AI. The support device 1 then generates materials for new projects through this co-creation.

[0097] There is a concern that if AI alone were to generate the materials, they would end up being similar to materials for new projects created in the past, but the support device 1 that executes the above-mentioned support process can provide materials that reflect different review results from past materials input by humans in a shorter time than if created by humans alone by co-creating with humans.As described above, the support device 1 that executes the support process can efficiently create materials for new projects and support improving their quality through co-creation between humans and AI.

[0098] That is, the assistance device 1 of this embodiment can realize Human in the Loop (HITL), an approach in which humans and AI work together to solve problems. In this approach, AI makes judgments and analyses, and humans evaluate and correct the results. As a result, more accurate and efficient results can be obtained.

[0099] The support process may also take the form of providing consideration result candidates including questions to the person inputting data (step S24). This allows the person inputting data to determine the content to be input as consideration result data by asking the questions. Therefore, the support device 1 in this form can promote co-creation between humans and AI through questions to the person inputting data.

[0100] In addition, the support process may have the user input the results of the study in three stages: advantages and prospects of the new project, concerns about the new project, and measures to overcome the concerns (steps S2 to S4). This allows the support device 1 to support the co-creation of the results of the study related to these important points in the materials for the new project in a step-by-step manner.

[0101] However, it is not easy for users like those mentioned above to judge whether the generated materials are good or bad. The support process provides the user with evaluation criteria for the generated materials by processing that shows the AI's evaluation of the generated materials (steps S6 to S8).

[0102] Co-creation between multiple people and AI is expected to produce materials that reflect a wider range of knowledge. The support process can take the form of providing various data to a collaborator other than the person who inputted the data and obtaining the collaborator's review results (steps S25 to S27). This allows the support device 1 to perform co-creation with AI as well as co-creation involving other people. <Usage example>

[0103] The following is an example of how the system S of this embodiment is used. [Basic data input]

[0104] A user inputs basic data related to the creation of materials for a new project into terminal T. Terminal T transmits the data to support device 1. Support device 1 receives the data. [Entering the first review result]

[0105] The support device 1 instructs the terminal T to display an input screen E that prompts the user to input the first review results. The terminal T displays the input screen E. The user inputs a draft of the first review results that includes the advantages and prospects of the new project.

[0106] The user instructs the support device 1 to conduct a review using AI via terminal T. The support device 1 instructs the display of an input screen E including review result candidates. Terminal T displays the input screen E.

[0107] Figure 6 is a display example of input screen E. At the top of this example, there is a display showing the flow of steps to be taken in order: "Preparation" for inputting basic data, "Advantages and prospects" for inputting the first study result data, "Concerns" for inputting the second study result data, "Measures to overcome" for inputting the third study result data, and "Proposal draft" for generating materials, and also showing that the current step is "Advantages and prospects."

[0108] Below the flow chart is an input field with the heading "Highlight Advantages." In the input field, users can write down their drafts: "Economic value: Efficiency will increase significantly, saving on human resources and reducing costs. Functional value: High-quality proposals can be created in a short amount of time, which will improve work efficiency." The first review result data shows that "the process will be smoother."

[0109] Further below that is the heading "List of AI Study Results," and below that are two candidate study results, along with buttons for copying the candidate study results to the input field mentioned above, editing the candidate study results, printing the candidate study results or outputting them as a PDF file, and hiding the candidate study results.

[0110] The first of the candidate results of the study begins with the following statement: "Social value: The widespread adoption of efficient business processes will contribute to improved productivity and strengthen the competitiveness of the entire company. Organizational culture change: Increased work efficiency will free up more time for new..."

[0111] The second one begins with the words, "· Time saving · User friendly UI · Output tailored to your needs..."

[0112] Further below that is a heading that reads "List of Other Users' Ideas." Below that are the results of the two collaborators' reviews, along with buttons for copying them to the input field mentioned above, editing them, printing them or outputting them as a PDF file, and hiding them.

[0113] The first result of the collaborator review includes a heading identifying the collaborator, "User 1," and states, "Why not have AI evaluate the proposal as well, and then rewrite it multiple times while looking at the AI's evaluation, thereby creating a collaborative brush-up?"

[0114] The second article includes a headline identifying the collaborator, "User 2," along with the results of a review of the collaborators, stating, "Wouldn't it be more reassuring for readers if the degree of AI's contribution was made visible? I believe there was research like that done in the Department of Psychiatry and Neurology at Duke University."

[0115] In this example, the user can refine the first review result data by incorporating standard review result candidates generated by the AI ​​based on extensive knowledge and collaborator review results presented from a perspective not available from the AI. [Entering the second review result]

[0116] The support device 1 instructs the terminal T to display an input screen E that prompts the user to input the second review results. The terminal T displays the input screen E. The user inputs a draft of the second review results, including concerns about the new project. As with inputting the first review results, the user also works to refine the second review result data by incorporating standard review result candidates based on extensive knowledge generated by the AI ​​and the collaborator review results presented from a perspective not available from the AI. [Third input of review results]

[0117] The support device 1 instructs the terminal T to display an input screen E that prompts the user to input the third study results. The terminal T displays the input screen E. The user inputs a draft of the third study results, including measures to overcome concerns about the new project. As with the input of the first study results, the user also works to refine the third study result data by incorporating standard study result candidates based on extensive knowledge generated by the AI ​​and the collaborator study results presented from a perspective not available from the AI. [Data generation]

[0118] The user instructs the support device 1 to generate materials for the new project based on the above input via the terminal T. The support device 1 generates the materials and displays them on the terminal T. [Brushing up using evaluation]

[0119] The user instructs the support device 1 to evaluate the generated materials via the terminal T. The support device 1 instructs the terminal T to comparatively display the evaluations of the multiple materials. The terminal T displays the evaluation screen A.

[0120] Figure 7 is an example of the display of evaluation screen A. In this example, the heading "AI Evaluation" is followed by the scores and evaluation reasons for items 1 to 7. Further below that, it is shown that the total score is 25 out of 35 points.

[0121] Further below that are general comments beginning with "This proposal utilizes new technology..." and advice for improving the score beginning with "The problems the customer faces..." Users can refer to these and brush up their materials.

[0122] At the bottom of Evaluation Screen A, a summary of previously created documents is displayed under the heading "Draft History." The first entry in the history shows a document with a "Total Score of 26 / 35" created on "2024 / 10 / 01 12:00:00," along with a button to view the document. The second entry in the history shows a document with a "Total Score of 20 / 35" created on "2024 / 10 / 01 11:50:00," along with a button to view the document. Users can refer to these to refine their documents to obtain a higher score.

[0123] Further below that, under the heading "Proposals from Other Users," a list of documents created by other users is displayed. The first item in the list is a heading indicating that there is a document titled "Iterative Processing Using AI" with a total score of 27 / 35, created by a user identified as "User 3," along with a button to view the document. The second item in the list is a heading indicating that there is an untitled document with a total score of 27 / 35, created by a user identified as "User 4," along with a button to view the document. Users can refer to these documents to refine their documents and achieve a higher score.

[0124] In the above embodiment, if S8 is completed and the branch at S9, S10, or S11 returns NO, the process returns to S2, S3, or S4 and restarts from there. However, this is not a limitation; the process may be returned even if S8 is not reached. For example, if the user selects the "Preparation" tab, the "Advantages and Prospects" tab, the "Concerns" tab, the "Overcoming Strategies" tab, the "Proposal Draft" tab, or the "Proposal F" tab on the screen of FIG. 6, the process may return to the step corresponding to the selected tab and restart from there. For example, if the user selects the "Advantages and Prospects" tab while S4 corresponding to the "Concerns" tab is being executed, the process may return to S3 and resume from there.

[0125] [Second embodiment]

[0126] 9 and 10 are block diagrams showing the configuration of an assistance device 900 according to a second embodiment. As shown in FIGS. 9 and 10, the assistance device 900 basically includes six LLM utilization units and a group of blocks sandwiched between two adjacent LLM utilization units. The six LLM utilization units include a multi-output LLM utilization unit 902, a multi-output LLM utilization unit 909, a multi-output LLM utilization unit 916, a single-output LLM utilization unit 923, a single-output LLM utilization unit 926, and a single-output LLM utilization unit 929. The six LLM utilization units utilize an external LLM (not shown). The multi-output LLM utilization units 902, 909, and 915 output multiple data sets generated by LLM. The single-output LLM utilization units 923, 926, and 929 output a single data set generated by LLM.

[0127] As will be described in more detail below, the multiple-output LLM utilization unit 902 outputs three data sets related to market segments per output. The multiple-output LLM utilization unit 909 outputs three data sets related to issues per output. The multiple-output LLM utilization unit 915 outputs three data sets related to solutions per output. Note that the number "3" is merely an example, and any integer greater than or equal to 2 may be used.

[0128] The single-output LLM utilization unit 923 outputs one data set relating to a concept and its evaluation at a time.

[0129] The single-output LLM utilization unit 926 outputs one dataset related to a specific concept at a time, where the dataset output by the single-output LLM utilization unit 923 includes questions related to the concept, and the dataset output by the single-output LLM utilization unit 926 includes answers to the questions as specific concepts.

[0130] The single-output LLM utilization unit 929 outputs one data set constituting a business concept proposal at a time.

[0131] There are multiple blocks between each LLM user and the next LLM user, and these blocks basically operate according to user operations. Therefore, this configuration makes it possible to create a Human in the Loop (HITL) where the user is interposed between LLMs and this process is repeated.

[0132] The storage units 903 - 1 to 903 - 3 store the respective data sets output by the multiple-output LLM utilization unit 902 .

[0133] The duplication unit 904 duplicates a dataset held in the holding unit 903-1 to the editing unit 906, for example, based on a user's operation. Similarly, the duplication unit 904 duplicates a dataset held in the holding units 903-2 and 903-3 to the editing unit 906, for example, based on a user's operation. Furthermore, the input unit 905 supplies a dataset input by a user to the editing unit 906 based on the user's operation. Therefore, the editing unit 906 can input a dataset output by the duplicate output LLM utilization unit 302 or a dataset input by a user and edit it.

[0134] Based on user operations, storage units 907-1 to 907-n1 store data sets currently stored in editing unit 906. Therefore, for example, storage unit 907-1 can store the data set stored in storage unit 903-1, storage unit 907-2 can store the data set stored in storage unit 903-3, and storage unit 907-3 can store a data set input by the user.

[0135] Basically, the selection unit 908 selectively supplies, based on a user operation, one of the data sets held in the holding unit 907-1 and the data set held in the holding unit 907-n1 to the multi-output LLM utilization unit 909. However, the selection unit 908 can also selectively supply, based on a user operation, any number of data sets from the data sets held in the holding unit 907-1 to the data sets held in the holding unit 907-n1 to the multi-output LLM utilization unit 909.

[0136] Next, each functional block from the multiple output LLM utilization unit 909 to the selection unit 915 is similar to each functional block from the multiple output LLM utilization unit 902 to the selection unit 918, so duplicated explanations will be omitted.

[0137] Each functional block from the multiple output LLM utilization unit 916 to the selection unit 922 is similar to each functional block from the multiple output LLM utilization unit 902 to the selection unit 918, so duplicated explanations will be omitted.

[0138] As described above, the single-output LLM utilization unit 923 outputs one data set related to a concept and its evaluation for each output. The selection unit 922 sequentially selects the data sets stored in the storage units 921-1 to 921-n3 and supplies them to the single-output LLM utilization unit 923, thereby allowing the data sets sequentially output by the single-output LLM utilization unit 923 to be stored in each of multiple storage units 924. The number of storage units 924 is represented by n4 instead of n3 because there are various ways of using them, and the number varies depending on the way they are used.

[0139] As described above, the single-output LLM utilization unit 926 outputs one data set related to a specific concept per output. The selection unit 922 sequentially selects the data sets stored in the storage units 924-1 to 924-n4 and supplies them to the single-output LLM utilization unit 926, thereby allowing the data sets sequentially output by the single-output LLM utilization unit 926 to be stored in multiple storage units 927. The number of storage units 927 is represented by n5 rather than n3 or n4 because there are various ways to use them, and the number varies depending on the way they are used.

[0140] As described above, the single-output LLM utilization unit 929 outputs one dataset constituting the business concept proposal at a time. The selection unit 928 sequentially selects datasets stored in the storage units 927-1 to 927-n5 and supplies them to the single-output LLM utilization unit 929, allowing the datasets sequentially output by the single-output LLM utilization unit 929 to be stored in multiple storage units 930. The number of storage units 930 is represented by n6 rather than n3, n4, or n5 because there are various ways to use them, and the number varies depending on the way they are used. The output unit 931 outputs the datasets stored in the storage unit 930, and is, for example, a file generation device, an image display device, or a printer.

[0141] Note that the selection unit 908 may select data sets held in multiple holding units 907-i1 to 907-ix instead of one holding unit 907-i and supply them to the multiple-output utilization unit 908. In this case, the multiple-output LLM utilization unit 909 may, for example, separate the data sets into cases for each data set and output the data sets. For example, the selection unit 908 may select data sets held in three holding units 907-i1 to 907-i3 and supply them to the multiple-output LLM utilization unit 908. In this case, the multiple-output LLM utilization unit 909 may, for example, separate the data sets into three cases and output three data sets for each case (i.e., nine data sets in total).

[0142] The multi-output LLM utilization unit 909 outputs, for example, three data sets for each data set selected by the selection unit 909. Therefore, if the selection unit 909 sequentially selects three data sets, the multi-output LLM utilization unit 909 outputs a total of 3 × 3 = 9 data sets. The multi-output LLM utilization unit 916 outputs, for example, three data sets for each data set selected by the selection unit 915. Therefore, if the selection unit 9159 sequentially selects nine data sets, the multi-output LLM utilization unit 916 outputs a total of 9 × 3 = 27 data sets. Therefore, for example, each data set is stored in each of the storage units 921-1 to 921-27. However, by using the data sets in a time-division manner, there is no need to increase the number of storage units.

[0143] Next, detailed operations of the assistance device 900 according to the second embodiment will be described with reference to FIGS. 10 to 28 showing display screens.

[0144] The support device 900 supports the creation of a business concept proposal through the following steps 0 to 5.

[0145] Step 0: Assets / Strategy Step 1: Market Selection Step 2: Identify the issue Step 3: Idea discovery Step 4: Concept evaluation Step 5: Report

[0146] Fig. 10 shows the screen for step 0. Figs. 11 to 14 show the screen for step 1. Figs. 15 to 17 show the screen for step 2. Figs. 18 to 22 show the screen for step 3. Figs. 23 to 26 show the screen for step 4. Figs. 27 and 28 show the screen for step 5.

[0147] 10, the screen for step 0 has a text box 1001 for entering the industry of the company conducting business development. Figure 10 shows an example in which the user enters "zabuton maker" in text box 1001 as the industry of the company conducting business development and then presses "Submit" button 1002, causing the character string "zabuton maker" 1003 to be displayed. The character string "zabuton maker" 1003 is stored in storage unit 901.

[0148] To the left of the character string 1003, there are displayed an icon 1004 for activating a copy function, an icon 1005 for activating an edit function, an icon 1006 for hiding an associated display object (also called a "switching component"), and an icon 1007 for activating a delete function. When the user selects icon 1004, the character string 1003 is copied to the clipboard. When the user selects icon 1005, a text box (not shown) for editing the character string 1003 is displayed, and the user can edit the character string 1003 using this text box. When the user selects icon 1006, the character string 1003 and icons 1004 to 1007 are hidden. When the user selects icon 1007, the character string 1003 is deleted. Similar icons displayed in other locations have similar functions.

[0149] The screen for step 0 has a text box 1008 for inputting strengths and assets that could be utilized in business development. Fig. 10 shows an example in which a user inputs "Able to design, manufacture, and sell cushions with new shapes and new materials" into text box 1008 as a strength or asset that could be utilized in business development, and then presses "Submit" button 1009, causing a character string 1010 of "Able to design, manufacture, and sell cushions with new shapes and new materials" to be displayed. The character string 1010 of "Able to design, manufacture, and sell cushions with new shapes and new materials" is stored in storage unit 901.

[0150] The screen for step 0 has a text box 1011 for inputting a policy or strategy for business development or innovation. Figure 10 shows an example in which a user inputs "Design, manufacture, and sell new doughnut-shaped cushions" as a policy or strategy for business development or innovation in text box 1011 and then presses a "Submit" button 1012, causing a character string 1013, "Design, manufacture, and sell new doughnut-shaped cushions," to be displayed. The character string 1013, "Design, manufacture, and sell new doughnut-shaped cushions," is stored in storage unit 901.

[0151] The screen for step 0 has a text box 1014 for the user to confirm and modify the session name. Fig. 10 shows an example in which "doughnut-shaped cushion," which the user has already entered on a screen not shown, is displayed in text box 1014. When the user selects icon 1013, the user can edit the session name in text box 1014.

[0152] 11, the screen for step 1 has a text box 1101 for inputting the target market for business development. FIG. 11 shows an example in which a user inputs "office," "home," "nursing home," and "hospital" in text box 1101 as the target markets for business development, and presses a "Submit" button 1102 after each input, thereby displaying the character string "office" 1103, the character string "home" 1104, the character string "nursing home" 1105, and the character string "hospital" 1106. Each of the character string "office" 1103, the character string "home" 1104, the character string "nursing home" 1105, and the character string "hospital" 1106 is stored in storage unit 901.

[0153] The Step 1 screen has a "Run" button (also called a "launch component") 1107 that causes the LLM to generate candidate market segments.

[0154] When the user presses the "Execute" button 1107, the multiple output LLM utilization unit 902 is started. The multiple output LLM utilization unit 902 causes the LLM to generate advantageous segments for each market based on the conditions of the following items (a) to (d):

[0155] (a) "Zabuton manufacturer" held in holding unit 901 as the industry of the company conducting business development (b) "We can design, manufacture, and sell cushions with new shapes and new materials" held in holding unit 901 as a strength or asset that can be utilized in business development. (c) "Design, manufacture, and sell a new donut-shaped cushion" held in holding section 901 as a policy or strategy for business development and innovation. (d) "Offices," "Homes," "Nursing Homes," and "Hospitals" held in the holding unit 901 as markets targeted for business development

[0156] The multiple output LLM utilization unit 902 displays a results list 1108 based on the information generated by the LLM regarding advantageous segments in each market.

[0157] As shown in Figures 12 and 13, the results list 1108 includes a table 1109 of candidate market segments, a first table 1110 of generated segments, a second table 1111 of generated segments, and a table 1112 of trace logs.

[0158] Since the specific contents of tables 1109 to 1111 are shown in FIG. 12, a detailed description will be omitted. Table 1109 lists the industry and market entry. Table 1110 lists five segments. Table 1111 lists three segments. There are a total of eight records, and accordingly, eight storage units 903 store data sets for each record. The trace log will be described later.

[0159] Referring to FIG. 13, the Step 1 screen has a text box 1113 below the results list 1108 for setting the market segment.

[0160] The user copies the character string of one of the market segments listed in table 1110 and the market segments listed in table 1111, pastes it into text box 1113, and then presses submit button 1114. The character string is then displayed below text box 1113. Note that each specific market segment in the market segment column of table 1110 and each specific segment in the market segment column of table 1111 may be displayed in a manner that distinguishes them from other character strings, so that they can be selected as market segments to be pasted into text box 1113. A manner that distinguishes them from other character strings may be, for example, a manner in which the font is bold or colored, or a manner in which they are highlighted.

[0161] For example, if the "Office Cushion Market" in table 1110 is the target of the above operation, the character string 1115 of "Office Cushion Market" is displayed, as shown in Figure 13. Similarly, if the "Health-Oriented Cushion Market" in table 1111 is the target of the above operation, the character string 1116 of "Health-Oriented Cushion Market" is displayed, as shown in Figure 13. If the "Eco-Friendly Cushion Market" in table 1111 is the target of the above operation, the character string 1117 of "Eco-Friendly Cushion Market" is displayed, as shown in Figure 13.

[0162] The user can toggle between displaying and hiding character string 1115 by clicking on the corresponding icon 1006B or on the open / close screen control (also called a "switching component") 1118. The same applies to character strings 1116 and 1117.

[0163] FIG. 14(a) shows an example in which only character string 1115 is displayed by such an operation, FIG. 14(b) shows an example in which only character string 1116 is displayed by such an operation, and FIG. 14(c) shows an example in which only character string 1117 is displayed by such an operation.

[0164] A text box 1113 corresponds to the editing unit 906 shown in Fig. 8. Furthermore, a storage unit 907 shown in Fig. 8 stores character strings 1115 to 1117. A selection unit 908 shown in Fig. 8 selects a character string that has become displayed by the above operation.

[0165] Referring to FIG. 15, the Step 2 screen has a "Run" button 1501 that causes the LLM to generate candidate assignments.

[0166] When the user presses the "Execute" button 1501, the multiple output LLM utilization unit 909 is started. The multiple output LLM utilization unit 909 causes the LLM to generate a problem in the "Office Cushion Market" based on the content of the necessary items from items (a) to (e) below.

[0167] (a) "Zabuton manufacturer" held in holding unit 901 as the industry of the company conducting business development (b) "We can design, manufacture, and sell cushions with new shapes and new materials" held in holding unit 901 as a strength or asset that can be utilized in business development. (c) "Design, manufacture, and sell a new donut-shaped cushion" held in holding section 901 as a policy or strategy for business development and innovation. (d) "Offices," "Homes," "Nursing Homes," and "Hospitals" held in the holding unit 901 as markets targeted for business development (e) The "office cushion market" held by holding unit 907 as a favorable segment

[0168] The necessary items are, for example, items (a), (d), and (e). However, the configuration of the present support device 900 is such that each of the items (a) to (e) can be switched between selected and unselected.

[0169] As shown in Figure 14(a), it is assumed that only the character string 1115 "office cushion market" is displayed as an advantageous segment, and the selection unit 908 selects only the advantageous segment indicated by the character string 1115 "office cushion market."

[0170] The multiple output LLM utilization unit 909 displays a list of results 1502 based on the information generated by the LLM about issues in the office cushion market.

[0171] As shown in FIG. 15, the result list 1502 includes a table 1503 of prerequisites, a table 1504 of generated tasks, and a trace log 1505 .

[0172] 15, so a detailed explanation of the specific contents of tables 1503 and 1504 will be omitted, but table 1503 lists the industry, market entry, and segment. Table 1504 lists three issues.

[0173] 14(b), if only the character string 1116 "healthy zabuton market" is displayed on the screen in step 1 and the selection unit 908 selects only the character string 1116 "healthy zabuton market," the contents of tables 1503 and 1504 will be different, but a detailed explanation will be omitted. Similarly, if only the character string 1117 "eco-friendly zabuton market" is displayed and the selection unit 908 selects only the character string 1117 "eco-friendly zabuton market," the contents of tables 1503 and 1504 will be different, but a detailed explanation will be omitted.

[0174] Referring to FIG. 16, the screen for step 2 has a text box 1506 below the results list 1502 for setting the challenge.

[0175] The user copies a character string from one of the multiple sets of situations, issues, targets, conflicts, and obstacles listed in table 1504, pastes it into text box 1506, and then presses submit button 1507. This causes the character string to be displayed below text box 1506. Note that the individual specific situations, issues, targets, conflicts, and obstacles in the status, issue, target, conflict, and obstacle columns of table 1504 may be displayed in a manner that distinguishes them from other character strings, so that they can be selected as the situations, issues, targets, conflicts, and obstacles to be pasted into text box 1506. Examples of a manner that distinguishes them from other character strings include a bold font, a colored font, or a highlighted font.

[0176] For example, if the set in the first record of table 1504 (the set of "Remote work is becoming more common and work from home is increasing," "I want to maintain a comfortable posture while working," "Business person working from home," "Office chair or regular chair," and "Sitting for long periods of time is straining.") is targeted for the above operation, a string 1508 corresponding to the set is displayed, as shown in FIG. 16, such as "Remote work is becoming more common and work from home is increasing. I want to maintain a comfortable posture while working. Business person working from home. Office chair or regular chair. Sitting for long periods of time is straining." Similarly, if the set in the second record of table 1504 is targeted for the above operation, a string 1509 corresponding to the set is displayed, as shown in FIG. 16. If the set in the third record of table 1504 is targeted for the above operation, a string 1510 corresponding to the set is displayed, as shown in FIG. 16.

[0177] The user can toggle between displaying and hiding character string 1508 by clicking the corresponding icon 1006C or by clicking the open / close screen control 1511. The same applies to character strings 11509 and 11510.

[0178] FIG. 17(a) shows an example in which only character string 1508 is displayed by such an operation, FIG. 17(b) shows an example in which only character string 1509 is displayed by such an operation, and FIG. 17(c) shows an example in which only character string 1510 is displayed by such an operation.

[0179] Text box 1506 corresponds to editing unit 913 shown in Fig. 8. Furthermore, storage unit 914 shown in Fig. 8 stores character strings 1508 to 1510. Selection unit 915 shown in Fig. 8 selects the character string that has become displayed by the above operation.

[0180] Referring to FIG. 18, the Step 3 screen has a "Run" button 1801 that causes the LLM to generate potential solutions.

[0181] When the user presses the "Execute" button 1801, the multiple output LLM utilization unit 916 is started. The multiple output LLM utilization unit 916 causes the LLM to generate a solution based on the contents of the required items from the following items (a) to (f).

[0182] (a) "Zabuton manufacturer" held in holding unit 901 as the industry of the company conducting business development (b) "We can design, manufacture, and sell cushions with new shapes and new materials" held in holding unit 901 as a strength or asset that can be utilized in business development. (c) "Design, manufacture, and sell a new donut-shaped cushion" held in holding section 901 as a policy or strategy for business development and innovation. (d) "Offices," "Homes," "Nursing Homes," and "Hospitals" held in the holding unit 901 as markets targeted for business development (e) The "office cushion market" held in the holding unit 907 as an advantageous segment is a problem in the "office cushion market." (f) A dataset related to the problem, stored in the storage unit 914, is: “In a situation where remote work is widespread and work at home is increasing, it is necessary to maintain a comfortable posture while working. Business people working at home, whether in an office chair or a regular chair, can feel strain when sitting for a long time.”

[0183] The necessary items are, for example, items (a), (b), (c), (d), (e), and (f). However, the configuration of the present support device 900 is such that each of the items (a) to (f) can be switched between selected and unselected.

[0184] As shown in FIG. 17(a), it is assumed that only the task indicated by the character string 1508 "Remote work is becoming more common and work from home is increasing. I want to maintain a comfortable posture while working. Business people working from home. Office chairs and regular chairs. Sitting for long periods of time can be straining." is displayed on the screen in step 2, and the selection unit 915 selects only the task indicated by the character string 1508 "Remote work is becoming more common and work from home is increasing. I want to maintain a comfortable posture while working. Business people working from home. Office chairs and regular chairs. Sitting for long periods of time can be straining."

[0185] The multiple output LLM utilization unit 916 displays a solution 1802 based on information generated by the LLM about a problem in the office cushion market.

[0186] As shown in FIGS. 18 and 19, a solution 1802 includes a table 1803 of prerequisites, tables 1804, 1805 and 1806 relating to generated solutions, and a trace log 1807.

[0187] 18 and 19, so a detailed explanation of the specific contents of tables 1803 to 1806 will be omitted, but table 1803 lists the industry, market entry, segment, and issues. Tables 1804 to 1806 each list one solution.

[0188] 17(b), when only character string 1509 is displayed and selection unit 915 selects only character string 1509, the contents of tables 1803 to 1806 will be different, but a detailed description will be omitted. Similarly, when only character string 11510 is displayed and selection unit 915 selects only character string 11510, the contents of tables 1803 to 1806 will be different, but a detailed description will be omitted.

[0189] The user can toggle between displaying and hiding the contents of solution 1802 by clicking the corresponding icon 1006D (see Figure 19) or by clicking the open / close screen control 1807 (see Figure 18). When the contents of solution 1802 are hidden, only the title of solution 1802 is displayed on the screen. In the screen shown in Figure 20, only the title of solution 1802B is displayed.

[0190] The screen shown in Fig. 20 also displays solutions 2001 and 2002, which are titles only. Solution 2001 is a solution generated when the selection unit 915 selects only the character string 1509. Solution 2002 is a solution generated when the selection unit 915 selects only the character string 1510. The user can switch between displaying and hiding the contents of each of the three solutions by operating the open / close screen controls 1807, 2003, and 2004.

[0191] Referring to FIG. 20, the Step 3 screen has a text box 2006 for setting a solution below a results list 2005 that includes solutions 1807, 2003, and 2004.

[0192] For example, the user copies the character string of the set of provision means, purpose, provision form, and supplementary information listed in table 1804, pastes it into text box 2006, and then presses submit button 2007. Then, a character string 2008 identical to the copied character string is displayed below text box 1806. Similarly, a character string 2009 identical to the character string of the set of provision means, purpose, provision form, and supplementary information listed in table 1805 can be displayed below that. Furthermore, a character string 2010 identical to the character string of the set of provision means, purpose, provision form, and supplementary information listed in table 1806 can be displayed below that. Note that the individual specific provision means, purpose, provision form, and supplementary information listed in the provision means, purpose, provision form, and supplementary information columns of table 1804 may be displayed in a manner that distinguishes them from other character strings so that they can be selected as the provision means, purpose, provision form, and supplementary information to be pasted into text box 2006. The same applies to tables 1805 and 1806. The manner in which the character string is distinguished from other character strings is, for example, a manner in which the font is bold or colored, or a manner in which a highlight is added.

[0193] The screen shown in Figure 20 displays three solutions included in Solution 1807. Similarly, the screen shown in Figure 21 displays nine solutions, including three solutions 2101-2103 included in Solution 2003 and three solutions 2104-2106 included in Solution 2004. Users can toggle the display of these nine solutions, 2008-2010 and 2101-2106, by selecting the corresponding hide icon or operating the open / close screen control. For example, Figure 22 shows the display of only one of these solutions, Solution 2008: "Ergonomically Designed Floor Cushion: Maintaining a Comfortable Posture for Long Periods, Home Floor Cushion Product, Focusing on Ergonomics as a Standard."

[0194] Referring to FIG. 23, the screen for step 4 has a "Run" button 2301 for causing the LLM to generate concepts and then evaluate the generated concepts.

[0195] When the user presses the "Execute" button 2301, the single-output LLM utilization unit 923 is started. The single-output LLM utilization unit 923 causes the LLM to generate concepts and evaluations based on the contents of the necessary items among the following items (a) to (g):

[0196] (a) "Zabuton manufacturer" held in holding unit 901 as the industry of the company conducting business development (b) "We can design, manufacture, and sell cushions with new shapes and new materials" held in holding unit 901 as a strength or asset that can be utilized in business development. (c) "Design, manufacture, and sell a new donut-shaped cushion" held in holding section 901 as a policy or strategy for business development and innovation. (d) "Offices," "Homes," "Nursing Homes," and "Hospitals" held in the holding unit 901 as markets targeted for business development (e) The "office cushion market" held in the holding unit 907 as an advantageous segment is a problem in the "office cushion market." (f) A dataset related to the problem, stored in the storage unit 914, is: “In a situation where remote work is widespread and work at home is increasing, it is necessary to maintain a comfortable posture while working. Business people working at home, whether in an office chair or a regular chair, can feel strain when sitting for a long time.” (g) As a solution, the "Ergonomically designed cushion for maintaining a comfortable posture for a long time, a cushion product for home use, focusing on ergonomics as a standard" is held in the holding part 921.

[0197] The necessary items are, for example, all items (a) to (g). However, the support device 900 is configured so that each of the items (a) to (g) can be switched between selected and unselected.

[0198] As shown in FIG. 22(a), it is assumed that only the solution indicated by the character string 2008 "Ergonomically designed cushion, maintains a comfortable posture for a long time, home cushion product, focuses on ergonomics as a standard" is displayed on the screen of step 3, and the selection unit 922 selects only the solution indicated by the character string 2008 "Ergonomically designed cushion, maintains a comfortable posture for a long time, home cushion product, focuses on ergonomics as a standard."

[0199] The single-output LLM utilization unit 923 displays concepts and ratings 2302 based on the information about concepts and ratings generated by the LLM.

[0200] As shown in Fig. 23, concept and evaluation 2302 includes title 2303, summary 2304, table 2305, overall evaluation 2306, and trace log 2307. These contents are as shown in Fig. 23. To briefly explain the contents, the summary includes the above-mentioned problem (f) and solution (g). Table 2305 includes five items as evaluation items. Table 2305 describes the item name, concept, evaluation criteria, score, and question for each item.

[0201] The user can toggle between displaying and hiding the contents of concept and rating 2302 by clicking the corresponding icon 1006E (see FIG. 23) or by clicking the open / close screen control 2007 (see FIG. 23). When the contents of concept and rating 2302 are hidden, only the title of concept and rating 2302 is displayed on the screen. In the screen shown in FIG. 24, only the title of concept and rating 2302B is displayed.

[0202] The screen shown in Fig. 24 also displays title-only concepts and ratings 2401 to 2408. The ith concept and rating among the nine concepts and ratings shown in Fig. 24 are the concept and rating generated when the ith set of character strings among the nine sets of character strings of means of provision, purpose, form of provision, and supplementary information shown in Fig. 21 is selected.

[0203] The user can switch between displaying and hiding the contents of each of the nine concepts and evaluations by operating the open / close screen controls 2008, 2409 to 2416.

[0204] 24 and 25, the screen for step 4 has a "Run" button 2418 for prompting the LLM to answer questions below a results list 2417 including concepts and assessments 2302 (2302B), 2401 to 2408. The questions here refer to the five questions in the question column of table 2305 (see FIG. 23) included in any of the concepts and assessments.

[0205] When the user presses the "Execute" button 2418, the single-output LLM utilization unit 926 is started. The single-output LLM utilization unit 926 generates an answer based on the contents of the necessary items among the following items (a) to (h):

[0206] (a) "Zabuton manufacturer" held in holding unit 901 as the industry of the company conducting business development (b) "We can design, manufacture, and sell cushions with new shapes and new materials" held in holding unit 901 as a strength or asset that can be utilized in business development. (c) "Design, manufacture, and sell a new donut-shaped cushion" held in holding section 901 as a policy or strategy for business development and innovation. (d) "Offices," "Homes," "Nursing Homes," and "Hospitals" held in the holding unit 901 as markets targeted for business development (e) The "office cushion market" held in the holding unit 907 as an advantageous segment is a problem in the "office cushion market." (f) A dataset related to the problem, stored in the storage unit 914, is: “In a situation where remote work is widespread and work at home is increasing, it is necessary to maintain a comfortable posture while working. Business people working at home, whether in an office chair or a regular chair, can feel strain when sitting for a long time.” (g) As a solution, the "Ergonomically designed cushion for maintaining a comfortable posture for a long time, a cushion product for home use, focusing on ergonomics as a standard" is held in the holding part 921. (h) Information stored in the storage unit 924 as concepts and questions

[0207] It is assumed that the first concept and the evaluation details are displayed on the screen in Step 4, but the second to ninth concepts and the evaluation details are not displayed.

[0208] The necessary items are, for example, items (a) to (e) and (h). However, the support device 900 is configured so that each of the items (a) to (h) can be switched between selected and unselected.

[0209] The single-output LLM utilization unit 926 displays the answer 2419B based on the information about the answer generated by the LLM.

[0210] As shown in Fig. 25, answer 2419B includes title 2501, Q&A 2502, and trace log 2503. These contents are as shown in Fig. 25. To briefly explain the contents, the Q&A includes two questions for each of the five evaluation items included in concept table 2305, and the answers to each question.

[0211] The user can toggle between displaying and hiding the content of answer 2419B to a question by clicking the corresponding icon 1006F (see FIG. 25) or by clicking the open / close screen control 2504 (see FIG. 25). When the content of answer 2419B is hidden, only the title of answer 2419B is displayed on the screen. In the screen shown in FIG. 24, the concept and rating 2419 with only the title are displayed. In the screen shown in FIG. 26, the concept and rating 2419 with only the title are also displayed.

[0212] Answers 2601 to 2608, which are titles only, are also displayed on the screen shown in Fig. 26. The ith answer of the nine answers shown in Fig. 26 corresponds to the ith concept and evaluation of the nine concepts and evaluations shown in Fig. 24.

[0213] The user can switch between displaying and hiding the contents of each of the nine answers by operating the open / close screen controls 2504, 2609 to 2616.

[0214] Referring to FIG. 27, the screen for step 5 includes a "run" button 2701 that causes the LLM to generate a business concept proposal.

[0215] When the user presses the "Execute" button 2701, the single-output LLM utilization unit 929 is launched. The single-output LLM utilization unit 929 causes the LLM to generate a business concept proposal based on the contents of the necessary items from the following items (a) to (i):

[0216] (a) "Zabuton manufacturer" held in holding unit 901 as the industry of the company conducting business development (b) "We can design, manufacture, and sell cushions with new shapes and new materials" held in holding unit 901 as a strength or asset that can be utilized in business development. (c) "Design, manufacture, and sell a new donut-shaped cushion" held in holding section 901 as a policy or strategy for business development and innovation. (d) "Offices," "Homes," "Nursing Homes," and "Hospitals" held in the holding unit 901 as markets targeted for business development (e) The "office cushion market" held in the holding unit 907 as an advantageous segment is a problem in the "office cushion market." (f) A dataset related to the problem, stored in the storage unit 914, is: “In a situation where remote work is widespread and work at home is increasing, it is necessary to maintain a comfortable posture while working. Business people working at home, whether in an office chair or a regular chair, can feel strain when sitting for a long time.” (g) As a solution, the "Ergonomically designed cushion for maintaining a comfortable posture for a long time, a cushion product for home use, focusing on ergonomics as a standard" is held in the holding part 921. (h) Information stored in the storage unit 924 as concepts and questions (i) Information stored in the storage unit 927 as a response

[0217] The necessary items are, for example, items (a) to (c), (e), and (g) to (i). However, the support device 900 is configured so that each of the items (a) to (i) can be switched between selected and unselected.

[0218] It is assumed that the content of the first answer is displayed on the screen in step 4, but the contents of the second to ninth answers are not displayed.

[0219] The single-output LLM utilization unit 929 displays a business concept proposal 2702 based on information about the business concept proposal generated by the LLM.

[0220] 27, the business concept proposal 2702 includes a title 2703, an outline 2704, contents 2705, and a trace log 2710. A detailed explanation of the contents 2705 will be omitted, but to put it simply, the contents 2705 are divided into contents for each of the five evaluation items included in the concept table 2305 (target customers and issues 2705, business concept 2706, solutions 2707, market competitiveness 2708, and worldview 2709).

[0221] The user can toggle between displaying and hiding the content of answer 2419B to the question by clicking the corresponding icon 1006G (see FIG. 27) or by clicking the open / close screen control 2712 (see FIG. 27). When the content of business concept proposal 2702 is hidden, only the title of business concept proposal 2702 is displayed on the screen. The screen shown in FIG. 28 displays only the title of business concept proposal 2702B.

[0222] The screen shown in Fig. 28 also displays business concept proposals 2801 to 2808, which are titles only. The i-th answer among the nine business concept proposals shown in Fig. 28 corresponds to the answer to the i-th question among the answers to the nine questions shown in Fig. 26.

[0223] The user can switch between displaying and hiding the contents of each of the nine business proposals by operating the open / close screen controls 2809 to 2817.

[0224] <About the prompt>

[0225] The prompt that the multiple output LLM utilization unit 902 provides to the LLM includes the following description: A description that the input items are items (a) to (d) stored in the storage unit 901 · Statement of LLM industry professionalism A statement instructing you to identify candidate segments for the market you want to enter, and then identify candidate market segments in the market you want to enter and the candidate market segments your company should enter from those candidate segments. A statement instructing you to present the top five market segments to enter based on your company's strengths, assets, and strategies. A statement instructing you to present the top three candidate market segments with the most potential customers and likely needs based on the external environment of the industry and recent trends. - Description that instructs trace log output Output format description

[0226] The prompt that the multiple output LLM utilization unit 909 gives to the LLM includes the following description: A description that the input items are items (a) and (d) stored in the storage unit 901 and item (e) selected by the selection unit 908. · Statement of LLM industry professionalism Statements regarding LLM's deep market and segment experience and knowledge A description that instructs you to identify the specific situation of the target customers in a market or segment, their background and context, and the combination of the goals they want to achieve and the problems they want to solve at that time, and present them in the specified output format. - Description that instructs trace log output Output format description

[0227] The prompts that the multiple output LLM utilization unit 916 provides to the LLM include the following: A description that the input items are the items (a) to (d) stored in the storage unit 901, the item (e) selected by the selection unit 908, and the item (f) selected by the selection unit 915. A statement directing the creation of three solutions -Description specifying the pattern that generates each solution A description that instructs you to identify the specific situation of the target customers in a market or segment, their background and context, and the combination of the goals they want to achieve and the problems they want to solve at that time, and present them in the specified output format. - Description that instructs trace log output Output format description

[0228] The prompt that the multiple output LLM utilization unit 923 gives to the LLM includes the following description: A description that the input items are the items (a) to (d) stored in the storage unit 901, the item (e) selected by the selection unit 908, the item (f) selected by the selection unit 915, and the item (g) selected by the selection unit 922. · Statement of LLM industry professionalism Instructions for creating a concept proposal Instructions for evaluating concept proposals and formulating questions - Description that instructs trace log output Output format description

[0229] The prompts that the multiple output LLM utilization unit 926 provides to the LLM include the following: A description that the input items are the items (a) to (d) stored in the storage unit 901, the item (e) selected by the selection unit 908, and the item (h) selected by the selection unit 925. · Statement of LLM industry professionalism A statement instructing you to write answers to questions posed by evaluating the concept proposal - Description that instructs trace log output Output format description

[0230] The prompts that the multiple output LLM utilization unit 929 provides to the LLM include the following: A description that the input items are the items (a) to (c) stored in the storage unit 901, the item (e) selected by the selection unit 908, the item (g) selected by the selection unit 922, the item (h) selected by the selection unit 925, and the item (i) selected by the selection unit 928. · Statement of LLM industry professionalism - Instructions for creating a new business proposal - Description that instructs trace log output Output format description

[0231] <About trace logs>

[0232] The trace log allows the user to understand the position of the current step relative to other steps.

[0233] The trace log contains a description of which step's dataset the current step's LLM consumer is providing to the LLM.

[0234] The trace log contains a description of whether the current step is a divergence step, a convergence & evaluation step, or a convergence step.

[0235] The trace log includes an explanation of which step selects the dataset generated in the current step if the current step is a divergence step, and also includes an explanation of which subsequent step handles the step selected in the step that selects the dataset generated in the current step.

[0236] If the current step is a convergence & evaluation step or a convergence step, the trace log includes an explanation of which step will consider the data set generated in the current step, and an explanation of which subsequent step will handle the consideration results generated in the step that considers the data set generated in the current step.

[0237] For example, the input of trace log 1505 (FIG. 15) for step 2-1 is "0-1 / 0-2 / 0-3," which indicates that the multiple-output LLM utilization unit 909 corresponding to step 2-1 supplies the items output from these three steps to the LLM. Trace log 1505 (FIG. 15) for step 2-1 also indicates that step 2-1 is a "divergence" step. Furthermore, trace log 1505 for step 2-1 indicates that one of the multiple outputs from step 2-1 is selected in step 2-2. Furthermore, trace log 1505 for step 2-1 indicates that the item selected in step 2-2 is used in steps 3-1, 4-1, and 5-1.

[0238] For example, the input of trace log 2307 (FIG. 23) for step 4-1 is "0-1 / 0-2 / 0-3 / 1-1 / 1-3 / 2-2 / 3-2," which indicates that the single-output LLM utilization unit corresponding to step 4-1 supplies the items output from these seven steps to the LLM. Trace log 2307 for step 4-1 also indicates that step 4-1 is a "convergence & evaluation" step. Furthermore, trace log 2307 for step 4-1 indicates that the output from step 4-1 is selected in step 4-3 (not shown). Furthermore, trace log 4-1 indicates that the item selected in step 4-3 is used in step 5-1.

[0239] <About Loop>

[0240] The user can freely select the steps to be executed by the assistance device 900 by selecting a tab located at the top of the screen (for example, the tab labeled "1 Market Selection"). Basically, the user causes the assistance device 900 to execute steps 0 to 5 in sequence. However, the user can also cause the assistance device 900 to repeatedly execute a specific range of steps. For example, the selection unit 908 can switch the data set selected and repeatedly execute a predetermined range of steps from the multi-output LLM utilization unit 909 onward. Any portion may be selected as the end of the predetermined range. The operation may be repeated up to the storage unit 914, with intermediate results stored in multiple storage units 914, or the operation may be repeated up to the storage unit 921, with intermediate results stored in multiple storage units 921, or the operation may be repeated up to the storage unit 930. Furthermore, any portion may be selected as the starting point of the repeated operation.

[0241] <About input by third parties> As shown in FIG. 10 and other figures, a set 1015 of an icon and a character string for inviting participants is displayed in the upper right corner of the screen. When a user selects this set, a screen (not shown) for the user to invite a third party is displayed. This screen displays invitation information such as a URL for accessing this screen. When a user invites a third party using the invitation information, the third party becomes able to display and operate the screen, input character strings, and the like. In other words, the third party becomes able to do the same things as the user. Using this function, the third party can input character strings in text boxes 1001, 1008, 1011, 1014, 1101, 1113, 1506, and 20006. Therefore, in addition to answers by the LLM and the user, answers by third parties can be used.

[0242] [Other embodiments] The configuration described in the second embodiment is merely an example. That is, for example, the use of three multi-output LLM utilization units and three single-output LLM utilization units as LLM utilization units is merely an example. In general, the number of these may be changed depending on the application, and the order of these may also be changed. The number of parallel storage units is also arbitrary. In particular, if a list is used, it does not need to be determined in advance. Furthermore, the allocation of steps to each screen is free. For example, the steps on the multiple screens in the above embodiment may be consolidated into a single screen. Furthermore, the prompt may be changed while maintaining the configuration depending on the application. The configuration may be changed and a new prompt may be applied depending on the application.

[0243] It should be noted that within the scope of the concept of the present invention, those skilled in the art may conceive of various modifications and alterations. Therefore, it is understood that such modifications and alterations fall within the scope of the present invention. For example, even if a person skilled in the art appropriately adds, deletes, or modifies components of the above-described embodiment, or adds, omits, or modifies the conditions of a process, such modifications are also included within the scope of the present invention as long as they maintain the gist of the present invention. [Explanation of symbols]

[0244] S System 1 Support equipment 11 Control section 111 Basic Data Acquisition Department 112 Input screen display section 113 Study Results Acquisition Department 114 Study result candidate generation unit 115 Information Provision Department 116 Collaborator Review Results Acquisition Department 117 Material generation department 118 Evaluation Generation Unit 119 Comparison display command section 13 Storage section 14 Communications Department N Network T-Terminal E Input screen A. Evaluation screen

Claims

1. A support device for generating materials by cumulatively using a plurality of review processing steps while accepting operations and inputs by a user based on basic data, a basic data acquisition unit that acquires the basic data; an input screen display unit that commands the display of an input screen for inputting review result data including answers to questions posed to a user in a current review processing step regarding the basic data in each review processing step; a review result acquisition unit that acquires the review result data input by the user via the input screen in each review processing step; a review result candidate generation unit that causes a large-scale language model to generate review result candidates (hereinafter referred to as "review result candidates") including answers to questions posed to a user in the current review processing step regarding the basic data, or the basic data and at least some of the review result data or review result candidates acquired by the review result acquisition unit in at least some review processing steps prior to the current review processing step, as review result candidates in each review processing step; Equipped with the input screen display unit is configured to provide, when the consideration result candidates for the current consideration processing step are generated by the consideration result candidate generation unit, an operation by the user to include part or all of the consideration result candidates for the current consideration processing step in the consideration result data for the current consideration processing step on the input screen; the consideration result candidate generation unit causes the large-scale language model to generate consideration result candidates including answers to questions posed to the user in the (n+1)th consideration processing step regarding consideration result candidates in the nth consideration processing step and consideration result data selected by the user, as consideration result candidates in the (n+1)th consideration processing step; a data generation unit that causes the large-scale language model to generate data that includes at least a portion of the basic data and the review result data or review result candidates after the review result data or review result candidates in each review processing step are obtained, When causing the large-scale language model to generate the consideration result candidates, the consideration result candidate generation unit performs the following on the large-scale language model: The basic data, or the basic data and at least a part of the review result data and review result candidates acquired by the review result acquisition unit in at least a part of the review processing steps prior to the current review processing step (hereinafter referred to as "input data"); The content of the question (hereinafter referred to as the "specific question") to the user in the current review processing step regarding the input data, - instructions to generate an answer to the specific question with reference to the input data; Enter When causing the large-scale language model to generate the material, the material generation unit The basic data, or the basic data and at least a portion of the review result data or review result candidates acquired by the review result acquisition unit in at least a portion of the review processing steps included in all of the review processing steps (hereinafter referred to as "total input data"); - instructions to generate the material with reference to the total input data; Enter the A support device for document creation.

2. The input screen display unit A command to display a first input screen for inputting first examination result data; A command to display a second input screen for inputting second examination result data; A command to display a third input screen for inputting third examination result data; configured to: the review result acquisition unit is configured to acquire the first review result data, the second review result data, and the third review result data, respectively; The consideration result candidate generation unit a first generation means for supplying the basic data to the large-scale language model and causing the large-scale language model to generate a first review result candidate including an answer to a question posed to a user in a first review processing step regarding the basic data; a second generation means for supplying the basic data, the first review result data, or the first review result candidate to the large-scale language model, and causing the large-scale language model to generate a second review result candidate including an answer to a question posed to a user in a second review processing step regarding at least a part of the data supplied to the large-scale language model; a third generation means for supplying the large-scale language model with the basic data, the first review result data, and the first review result candidates that have been shared with the large-scale language model by the second generation means, and any of the second review result data and the second review result candidates, and causing the large-scale language model to generate third review result candidates including answers to questions posed to a user in a third review processing step regarding at least a portion of the data supplied to the large-scale language model; Equipped with The input screen display unit a means for providing, on the first input screen, an operation for including a part or all of the first review result candidates in the input of the first review result data when the first review result candidates have been generated; A means for providing, on the second input screen, an operation for including a part or all of the second consideration result candidates in the input of the second consideration result data when the second consideration result candidates have been generated; a means for providing, on the third input screen, an operation for including a part or all of the third review result candidates in the input of the third review result data when the third review result candidates have been generated; Equipped with The consideration result candidate generation unit When the large-scale language model generates the first consideration result candidate, the large-scale language model The basic data (hereinafter referred to as "first input data"), The content of a question (hereinafter referred to as a "first specific question") posed to the user in the first review processing step regarding at least a part of the first input data, instructions for generating an answer to the first specific question with reference to at least a portion of the first input data; Enter When the large-scale language model generates the second consideration result candidate, the large-scale language model The basic data and the first review result data (hereinafter referred to as "second input data"), The content of a question (hereinafter referred to as a "second specific question") posed to the user in the second review processing step regarding at least a part of the second input data, instructions for generating an answer to the second specific question with reference to at least a portion of the second input data; Enter When the large-scale language model generates the third consideration result candidate, the large-scale language model The basic data, the first review result data, and the second review result data (hereinafter referred to as "third input data"); The content of a question (hereinafter referred to as a "third specific question") to the user in the third review processing step regarding at least a part of the third input data, instructions for generating an answer to the second specific question with reference to at least a portion of the third input data; Enter the The support device according to claim 1 .

3. an information providing unit that provides data including the basic data or the review result data of the current review processing step to a collaborator different from the person who inputted the data; a collaborator review result acquisition unit that acquires a collaborator review result by the collaborator; Furthermore, The input screen display unit is configured to provide, when the collaborator review results have been acquired, an operation to include part or all of the collaborator review results in the input of the review result data of the current review processing step on the input screen. The support device according to claim 1 .

4. A program for causing a computer to function as the support device according to any one of claims 1 to 3.

Images