Document creation device, document creation method, and recording medium

Abstract

This document creation device comprises: a document creation means for creating a document by using information relating to the document and an agent model that mimics the thought of a person assigned a role relating to a proposal using the document; and a presentation means for presenting the created document.

Description

Document Creation Device, Document Creation Method, and Recording Medium

[0001] This disclosure relates to the technical field of document creation devices, document creation methods, and recording media.

[0002] For example, Patent Document 1 discloses a system for creating a plan.

[0003] International Publication No. 2024 / 166521

[0004] An object of this disclosure is to provide a document creation device, a document creation method, and a recording medium that aim to improve the technology related to the above-mentioned prior art documents.

[0005] One aspect of the document creation device includes a document creation means for creating the document using information related to the document and an agent model that mimics the thinking of a person given a role related to a proposal using the document, and a presentation means for presenting the created document.

[0006] One aspect of the document creation method is a document creation method executed by a computer, including creating the document using information related to the document and an agent model that mimics the thinking of a person given a role related to a proposal using the document, and presenting the created document.

[0007] One aspect of the recording medium has a computer program recorded thereon for causing a computer to execute a document creation method including creating the document using information related to the document and an agent model that mimics the thinking of a person given a role related to a proposal using the document, and presenting the created document.

[0008] This is a block diagram showing an example of the configuration of a document creation device according to the embodiment. This is a flowchart showing an example of the flow of document creation operations in the document creation device according to this disclosure. This is a block diagram showing an example of the configuration of a document creation device according to the embodiment. This is a flowchart showing an example of the flow of agent model creation operations in the document creation device according to this disclosure. This is a flowchart showing an example of the flow of document creation operations in the document creation device according to this disclosure. This is a block diagram showing an example of the configuration of a document creation device according to the embodiment. This is a flowchart showing an example of the flow of document creation operations in the document creation device according to this disclosure.

[0009] The following describes embodiments of the document creation apparatus, document creation method, and recording medium with reference to the drawings. [1: First Embodiment]

[0010] A first embodiment relating to a document creation device, a document creation method, and a recording medium will be described with reference to Figures 1 and 2. In the following description, the first embodiment relating to a document creation device, a document creation method, and a recording medium will be described using the document creation device 10.

[0011] As shown in Figure 1, the document creation device 10 comprises a document creation unit 11 and a document presentation unit 12. The operations performed by the document creation device 10 will be explained with reference to the flowchart in Figure 2.

[0012] As shown in Figure 2, the document creation unit 11 generates the document using information about the document and an agent model (step S11). The document creation unit 11 may automatically generate the document. A person who is given a "role" related to the proposal using the document is referred to as the "target person". The agent model may be a model that mimics the thinking of the "target person (i.e., the "person who has been given a role")".

[0013] The "role" may indicate what the person assigned the corresponding "role" is responsible for and what they are responsible for. For example, the "role" may include "proposal" using materials. In this case, the "target audience" may include the "proposer." Also, the "role" may include "approval" of a proposal using materials. In this case, the "target audience" may include the "approver."

[0014] There may be multiple "targets" who are given the same "role." The agent model may be a model that mimics the thinking of "targets" given the same "role," the thinking expected of "targets" given the same "role," the thinking that "targets" given the same "role" should do, the average thinking of "targets" given the same "role," etc.

[0015] Individuals assigned the same "role" may hold different "positions." For example, a "superior" and a "subordinate" assigned the same "role" may hold different "positions." Multiple individuals may hold the same "position." The agent model may be a model that mimics the thinking of an individual holding a "position."

[0016] The document presentation unit 12 presents the generated document (step S12). The recipient to whom the document presentation unit 12 presents may be the "target person". The recipient to whom the document presentation unit 12 presents may be the "proposer". The recipient to whom the document presentation unit 12 presents may be the "approver".

[0017] Thus, the document creation device 10 performs a document creation method that includes creating a document and presenting the created document, using information related to the document and an agent model that mimics the thinking of a subject who has been given a role in making a proposal using the document.

[0018] The document creation device 10 described above may be realized by a computer reading a computer program recorded on a recording medium. In this case, the computer program may cause the computer to execute a document creation method that includes creating a document and presenting the created document, using information about the document and an agent model that mimics the thinking of a subject who has been given a role in making a proposal using the document. [Technical Effects]

[0019] The document creation device 10 described in this disclosure creates and proposes documents using an agent model that mimics the thinking of a "target person" who has been given a "role" related to proposals using the documents. Therefore, it can propose documents that are in line with the "target person's" thinking while reducing the effort required from the "target person". [2: Second Embodiment]

[0020] A second embodiment relating to a document creation device, a document creation method, and a recording medium will be described with reference to Figures 3 to 5. Hereinafter, the second embodiment relating to the document creation device, the document creation method, and the recording medium will be described using the document creation device 20. Note that, in the second embodiment, explanations that overlap with the description of the first embodiment described above will be omitted as appropriate. [2-1: Configuration of the document creation device 20]

[0021] The configuration of the document creation device 20 will be explained with reference to Figure 3. Figure 3 is a block diagram showing the configuration of the document creation device 20.

[0022] As shown in Figure 3, the data creation device 20 comprises an arithmetic unit 21 and a storage device 22. Furthermore, the data creation device 20 may also include a communication device 23, an input device 24, and an output device 25. However, the data creation device 20 does not have to include the communication device 23 and at least one of the input device 24 and the output device 25. The arithmetic unit 21, the storage device 22, the communication device 23, the input device 24, and the output device 25 may be connected via a data bus 26.

[0023] The arithmetic unit 21 includes at least one processor (i.e., one or more processors) as hardware. The processor may include, for example, a processor conforming to a von Neumann computer architecture. A processor conforming to a von Neumann computer architecture may include at least one of a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit). The processor may also include, for example, a processor conforming to a non-von Neumann computer architecture. A processor conforming to a non-von Neumann computer architecture may include at least one of an FPGA (Field Programmable Gate Array) and an ASIC (Application Specific Circuit).

[0024] The arithmetic unit 21 reads a computer program 221 which includes at least one of computer program code and computer program instructions. For example, the arithmetic unit 21 may read a computer program 221 stored in a storage device 22. For example, the arithmetic unit 21 may read a computer program 221 stored on a computer-readable and non-temporary recording medium using a recording medium reader (not shown) provided in the data creation device 20. The computer program 221 read from the recording medium may be stored in the storage device 22. The arithmetic unit 21 may obtain (i.e., download or read) a computer program 221 from a device (not shown) located outside the data creation device 20 via a communication device 23 (or other communication device). The downloaded computer program 221 may be stored in the storage device 22.

[0025] The arithmetic unit 21 executes the loaded computer program 221. As a result, a logical functional block for executing the document creation method that the document creation device 20 should perform is realized within the arithmetic unit 21. In other words, the arithmetic unit 21, together with the storage device 22 on which the computer program 221 is recorded (in other words, together with the storage device 22 and the computer program 221 recorded in the storage device 22), can function as a controller or computer for realizing a logical functional block for executing the processing that the document creation device 20 should perform. That is, together with at least one processor in the arithmetic unit 21, the memory (recording medium) in the storage device 22 and the computer program 221 are configured so that the document creation device 20 performs the information processing that the document creation device 20 should perform.

[0026] The computing device 21 may implement a computational model that can be constructed by machine learning by the computing device executing a computer program 221. An example of a computational model that can be constructed by machine learning is a computational model that includes a neural network (so-called artificial intelligence (AI)). In this case, the learning of the computational model may include learning the parameters of the neural network (for example, at least one of the weights and biases). The computing device 21 may execute an information processing method using the computational model. That is, the operation of executing an information processing method may include the operation of executing an information processing method using the computational model. Furthermore, the computing device 21 may implement a computational model that has been constructed by offline machine learning using training data. In addition, the computational model implemented in the computing device 21 may be updated by online machine learning on the computing device 21. Alternatively, the arithmetic unit 21 may execute the information processing method using an arithmetic model implemented in an external device (i.e., a device provided outside the data creation device 20) in addition to or instead of the arithmetic model implemented in the arithmetic unit 21.

[0027] Furthermore, the recording medium for recording the computer program 221 executed by the arithmetic unit 21 may include at least one of the following: optical discs such as CD-ROM, CD-R, CD-RW, flexible disk, MO, DVD-ROM, DVD-RAM, DVD-R, DVD+R, DVD-RW, DVD+RW, and Blu-ray (registered trademark); magnetic media such as magnetic tape; magneto-optical disks; semiconductor memory such as USB memory; and any other medium capable of storing a program. The recording medium may also include equipment capable of recording computer programs (for example, general-purpose or dedicated equipment on which the computer program 221 is implemented in an executable state in at least one form such as software and firmware). Furthermore, each process and function included in the computer program 221 may be realized by logical processing blocks implemented within the arithmetic unit 21 (i.e., the processor) when the arithmetic unit 21 executes the computer program 221, or by hardware such as a predetermined gate array (FPGA (Field Programmable Gate Array), ASIC (Application Specific Integrated Circuit)) provided by the arithmetic unit 21, or in a form in which logical processing blocks and partial hardware modules that realize some elements of the hardware are mixed.

[0028] The storage device 22 includes at least one memory capable of storing desired data. In other words, the storage device 22 includes at least one memory containing desired data. For example, the storage device 22 may store a computer program 221 executed by the arithmetic unit 21. In this case, the storage device 22 (memory) may be used as the recording medium described above for recording the computer program 221 executed by the arithmetic unit 21. The storage device 22 may temporarily store data that the arithmetic unit 21 temporarily uses when the arithmetic unit 21 is executing the computer program 221. The storage device 22 may store data that the data creation device 20 stores long-term. The storage device 22 may include at least one of RAM (Random Access Memory), ROM (Read Only Memory), hard disk drive, magneto-optical disk drive, SSD (Solid State Drive), and disk array drive. In other words, the storage device 22 may include a recording medium that is not temporary.

[0029] The communication device 23 may be capable of communicating with devices outside the document creation device 20. The communication device 23 may use either wired or wireless communication.

[0030] The input device 24 is a device capable of receiving information input to the document creation device 20 from an external source. The input device 24 may include an operating device (e.g., a keyboard, mouse, touch panel, etc.) that can be operated by the user of the document creation device 20. The input device 24 may include a recording medium reader capable of reading information recorded on a recording medium that can be attached to and detached from the document creation device 20, such as a USB (Universal Serial Bus) memory. When information is input to the document creation device 20 via the communication device 23 (in other words, when the document creation device 20 acquires information via the communication device 23), the communication device 23 may function as an input device.

[0031] The output device 25 is a device capable of outputting information to the outside of the document creation device 20. The output device 25 may output visual information such as characters and images, auditory information such as sounds, or tactile information such as vibrations. The output device 25 may include, for example, at least one of a display, speaker, printer, and vibration motor. The output device 25 may also be capable of outputting information to a recording medium that can be attached to or detached from the document creation device 20, such as a USB memory stick. When the document creation device 20 outputs information via the communication device 23, the communication device 23 may function as an output device.

[0032] Figure 3 shows an example of a logical functional block implemented within the arithmetic unit 21 to execute the document creation method. As shown in Figure 3, the arithmetic unit 21 implements a document creation unit 211, a document presentation unit 212, and an agent model creation unit 213. The "document creation unit 211" is a component corresponding to the "document creation unit 11" in the first embodiment described above, and the "document presentation unit 212" is a component corresponding to the "document presentation unit 12" in the first embodiment described above. The document creation unit 211 may include a draft information receiving unit 2111, a content creation unit 2112, a document determination unit 2113, and a format conversion unit 2114. The agent model creation unit 213 may include a metadata creation unit 2131. [2-2: Document creation method executed by the document creation device 20]

[0033] The document creation device 20 is configured as a device for creating documents. The person who uses the document creation device 20 is referred to as the user. The user may also be a "proposer" who makes a proposal using the documents. The document creation device 20 may support the proposal of documents by the "proposer". The document creation device 20 may create an agent model to support the creation of documents. The document creation device 20 may create documents using the created agent model. As described above, the agent model imitates the thinking of the target person.

[0034] The operations performed by the document creation device 20 will be explained with reference to Figures 4 and 5. Figure 4 is a flowchart showing an example of the flow of agent model creation operations performed by the document creation device 20. Figure 5 is a block diagram showing an example of the flow of the document creation method performed by the document creation device 20. [2-2-1: Agent model creation] [Step S21: Receipt of subject log data]

[0035] As shown in Figure 4, the metadata creation unit 2131 receives information about the subject (step S21). The information about the subject received by the metadata creation unit 2131 may be information that can be obtained from a human being. The information about the subject received by the metadata creation unit 2131 may be referred to as "(subject's) log data". The subject's log data may include, for example, at least one of the following: emails sent and received by the subject in the past, minutes of meetings attended by the subject, transcripts of phone calls the subject has made, transcripts of web conferences the subject has attended, the subject's employee profile, and the subject's resume. Furthermore, if the subject is a well-known person such as a company president, the subject's log data may also include interview articles about the subject that exist on the web.

[0036] The information about the subject that the metadata creation unit 2131 receives may be information entered by a human. The metadata creation unit 2131 may also accept human input via the input device 24 as information about the subject.

[0037] The information entered by a human may, for example, be text data about the subject entered by a human via the input device 24. In other words, the metadata creation unit 2131 may accept text data about the subject. Alternatively, the information entered by a human may, for example, be voice data entered by a human via the input device 24. In other words, the metadata creation unit 2131 may accept voice data about the subject.

[0038] The metadata creation unit 2131 may accept information about itself entered by the subject. For example, if the subject is an approver, the metadata creation unit 2131 may accept information about itself entered by the approver. Alternatively, the metadata creation unit 2131 may accept information about the subject entered by someone other than the subject. For example, the metadata creation unit 2131 may accept information about the approver entered by the proposer. For example, the metadata creation unit 2131 may accept information about the approver entered by the proposer on behalf of the approver. For example, the metadata creation unit 2131 may accept text data entered by the proposer such as "The approver is in charge of marketing chocolate confectionery."

[0039] The metadata creation unit 2131 may store the subject's log data 222 in the storage device 22. The metadata creation unit 2131 may also store the log data 222 in a storage device other than the storage device 22. The following describes the case where the log data 222 is stored in the storage device 22. When the metadata creation unit 2131 receives audio data, it may also record the text data of the speech recognition result of the relevant audio data as log data 222 in the storage device 22. [Step S22: Metadata creation]

[0040] The metadata creation unit 2131 creates metadata for the subject (step S22). The metadata creation unit 2131 may extract metadata for the subject from the subject's log data 222. The metadata creation unit 2131 may also create metadata by converting the subject's log data 222 into a predetermined format.

[0041] The metadata creation unit 2131 may use Large Language Models (LLMs). The metadata creation unit 2131 may input the subject's log data 222 into the LLM and have the LLM output the subject's metadata. The metadata creation unit 2131 may input the subject's log data and prompts for metadata extraction into the LLM and have the LLM output the subject's metadata. The metadata creation unit 2131 may accept a prompt such as, "The following log data exists regarding the approver: {[Approver's log data]}. Based on this, please output the person's work role, purpose, etc., in JSON format." For example, suppose there is text data as [Approver's log data] that says, "I am in charge of marketing chocolate confectionery...". In this case, the metadata creation unit 2131 may accept a prompt such as, "The approver has entered the following: {I am in charge of marketing chocolate confectionery...}. Based on this, please output the approver's work role, purpose, etc., in JSON format." The metadata creation unit 2131 may, for example, create text data as metadata for the approver, such as, "His role is department head, and he has budget authority up to 1 million yen. His past role was marketer, and his objective was to increase chocolate confectionery sales by 10%."

[0042] The metadata creation unit 2131 may create metadata for multiple subjects. For example, suppose approver A and approver B are subjects holding the same position. In this case, the metadata creation unit 2131 may create metadata for the positions of approver A and approver B. The metadata creation unit 2131 may input a prompt containing log data for multiple subjects to the LLM and have the LLM output metadata for multiple subjects.

[0043] The metadata creation unit 2131 may assign the role of interviewer to the LLM and have the LLM conduct an interview with the subject. For example, the metadata creation unit 2131 may give the LLM a prompt such as "Please ask questions so that you can understand the business role, purpose, etc. of the person you will talk to next", and let the LLM and the subject have a conversation. In this case, the metadata creation unit 2131 may receive the information elicited by the LLM from the subject. The metadata creation unit 2131 may extract the metadata of the subject from the interview results.

[0044] The metadata creation unit 2131 may store the metadata 223 of the subject in the storage device 22. The metadata creation unit 2131 may also store the metadata 223 in a storage device other than the storage device 22. Hereinafter, the case where the metadata 223 is stored in the storage device 22 will be described. [Step S23: Creation of agent model]

[0045] The agent model creation unit 213 creates an agent model of the subject (step S23). For example, the agent model creation unit 213 may input the metadata of the subject to the LLM and have the LLM output the agent model of the subject. The agent model creation unit 213 may input the log data of the subject, the metadata of the subject, and a prompt for creating an agent model to the LLM, and have the LLM output an agent model. The agent model may be a model given personality by metadata.

[0046] The agent model may be an example of the computation model described above. The agent model creation unit 213 may create an agent model by fine-tuning a pre-trained computation model. The agent model creation unit 213 may create an agent model by setting a system prompt for the pre-trained computation model. The system prompt may be text data based on metadata. The agent model creation unit 213 may set the text data "You are a user with the characteristics of {[Approver Metadata]}. From now on, please act as a budget approver" as the system prompt for the LLM. The pre-trained computation model may be an LLM. The pre-trained computation model may be a Vision Language Model (VLM).

[0047] The agent model creation unit 213 may create multiple agent models. Each of the multiple agent models may mimic the business thinking and judgment tendencies of each of the multiple subjects. The multiple subjects may include the proposer of the proposal and the approver of the proposal. The multiple subjects may include at least one of the following: a higher approver who approves the approver, a customer who implements the proposal, a consumer who consumes the proposal, a media organization, and an investor in the customer. The customer who implements the proposal may be a company. The consumer who consumes the proposal may be a purchaser who buys products offered by the company. The media may be an organization that disseminates information about products offered by the company. An investor in the customer may be a shareholder.

[0048] The agent model does not necessarily imitate a single person. The agent model may be a model that performs a role abstracted from the roles of a plurality of subjects. For example, the agent model creation unit 213 may create an agent model that imitates the integrated thinking of a plurality of persons given a specific role. The agent model creation unit 213 may, for example, create an agent model that imitates the integrated thinking of "Approver A" and "Approver B" who hold the same position. For example, assume that "Approver A" and "Approver B" are persons who hold management positions in the corresponding company. In this case, the agent model creation unit 213 may generate an agent model for the management positions in the corresponding company.

[0049] When creating an agent model for a proposer who proposes materials, the metadata creation unit 2131 may receive rules regarding the proposal of materials. As an example, the metadata creation unit 2131 may receive input of text data such as "Please make the agent model of the proposer comply with the following constraints. {[Rules of that department, laws, etc.]}." from the user. In this case, the agent model creation unit 213 may create an agent model of the proposer based on the received rules. As another example, the metadata creation unit 2131 may receive from the user a prompt such as "The proposer is engaged in creating materials for food advertisements. Please describe the rules to be observed from a legal perspective at this time." In this case, the agent model creation unit 213 may create an agent model of the proposer based on the described rules. [2-2-2: Creation of Materials]

[0050] As shown in FIG. 5, the case information reception unit 2111 receives case information regarding the proposal of the materials to be created (step S24). The case information may be proposed by the proposer. The case information may include text data. [Step S25: Creation of Materials Using the First Agent Model]

[0051] The agent model creation unit 213 may accept a system prompt specified by the user. The input device 24 may provide an interface for the user to input text data. The user may input a system prompt for the first agent model into the interface provided by the input device 24, and the agent model creation unit 213 may accept the input system prompt for the first agent model. The first agent model may be the agent model of the "proposer". For example, the user may input the following text as the system prompt for the first agent model: "You have the following characteristics: 1. Your company develops products for convenience stores in Japan. 2. As the section chief of the product development department, you are working hard to secure a budget for new product development." The system prompt may also be text quoted from metadata 223. In response to receiving the system prompt for the first agent model, the agent model creation unit 213 may set the system prompt for the first agent model in the pre-trained computation model and create the first agent model.

[0052] The content creation unit 2112 creates the document using the first agent model (step S25). Alternatively, the content creation unit 2112 can be said to create the document in collaboration with the first agent model. [Step S25-1: Creation of the document's framework]

[0053] The content creation unit 2112 may accept a first prompt for creating the basic framework of the document. The content creation unit 2112 may input the first prompt to the first agent model and have the first agent model output the basic framework of the document. For example, the content creation unit 2112 may accept a prompt such as, "Purpose of the document: {[Text for the purpose of creating the document]}. The approver is this kind of person {[Approver metadata]}. We want to adhere to these constraints {[Text for the constraints of creating the document]}. Using this information as a reference, we would like to first solidify the basic framework so that we can create about 10 slides. Please create the title and summary text." In this case, specific text data may be entered into each of [Text for the purpose of creating the document], [Approver metadata], and [Text for the constraints of creating the document]. For example, the content creation unit 2112 may output text data that includes "[Page 1: Title slide], [Page 2: Overview of the new product], [Page 3: Market research and consumer response], ...]" as the basic framework of the document. In other words, the first agent model may consider what should be included in the document. [Step S25-2: Creating details for each document]

[0054] The content creation unit 2112 may accept a second prompt for creating the details of each document. The content creation unit 2112 may input the second prompt to the first agent model and have the first agent model create the details of each document. For example, the content creation unit 2112 may accept a prompt such as, "Purpose of the document: {[Text for the purpose of creating the document]}. The approver is this kind of person {[Approver metadata]}. We want to adhere to these constraints {[Text for the constraints of creating the document]}. {[Outline of page n output in step S25-1]}. At this time, please create the details of the slides in bullet points, within 200 characters, according to this outline." Specific text data may be entered into each of the following: [Text for the purpose of creating the document], [Approver metadata], [Text for the constraints of creating the document], and [Outline of page n output in step S25-1].

[0055] Furthermore, the first agent model does not have to be the "proposer's" agent model. The first agent model may be a model without any particular characteristics. The first agent model may be, for example, a model that has not been fine-tuned.

[0056] The content creation unit 2112 may extract relevant data. The content creation unit 2112 may accept a prompt such as, "Purpose of the document: {[Text for the purpose of creating the document]}. The approver is this kind of person {[Approver metadata]}. We want to adhere to these constraints {[Text for the constraints of creating the document]}. {[Details of page n output in step S25-2]}. Please extract relevant data based on this information." Specific text data may be entered into each of the following: [Text for the purpose of creating the document], [Approver metadata], [Text for the constraints of creating the document], and [Details of page n output in step S25-2].

[0057] The content creation unit 2112 may input prompts to the LLM and have it extract relevant data. The relevant data may be figures or tables. The prompts may include information indicating one or more selected relevant data. The LLM may be capable of performing web searches and local searches.

[0058] The content creation unit 2112 may generate figures and tables. The content creation unit 2112 may receive text data such as "Purpose of the document: {[Text for the purpose of document creation]}. The approver is this kind of person {[Metadata of the approver]}. We want to adhere to these constraints {[Text for the constraints of document creation]}. {[Details of page n output in step S25-2]}. We would like to generate related figures based on this information. Please generate a prompt suitable for input to the image generation AI." and generate a prompt. The content creation unit 2112 may input the received text data into the LLM and have the LLM generate a prompt. The content creation unit 2112 may input the generated prompt into the VLM and have the VLM generate related figures and tables. The VLM may be, for example, DALLE-3. [Step S26: Document determination using the second agent model]

[0059] The agent model creation unit 213 may accept system prompts specified by the user. The input device 24 may provide an interface for the user to input text data. The user may input a system prompt for the second agent model into the interface provided by the input device 24, and the agent model creation unit 213 may accept the input system prompt for the second agent model. The second agent model may be an agent model for an "approver". In other words, the second agent model may imitate a target person other than the target person imitated by the first agent model.

[0060] The user may input the following text as the system prompt for the second agent model: "You have the following characteristics: 1. Your company develops products for convenience stores in Japan. 2. As the head of the product development department, you are in a position to approve the budget for new product development. 3. This year the budget is tight, so I only want to approve products that will sell. 4. However, if it seems likely to become popular among high school students, I will approve it whenever possible." The system prompt may also be text quoted from metadata 223. Upon receiving the system prompt for the second agent model, the agent model creation unit 213 may create the second agent model by setting the system prompt for the second agent model in the pre-trained computation model.

[0061] The document determination unit 2113 determines the document using the second agent model (step S26). Alternatively, the document determination unit 2113 may determine the document in cooperation with the second agent model. The document determination unit 2113 may output a predicted result of the approver's response. The document determination unit 2113 may provide the document to the second agent model and have the second agent model output a predicted approver's response. The document determination unit 2113 may output the determination result from the second agent model. The document determination unit 2113 may provide the document to the second agent model and have the second agent model output a determination result of whether the approver approves the document. [Step S26-1: Inquiry to the approver agent model]

[0062] The document evaluation unit 2113 may accept the prompt, "{[Text output by the content creation unit 2112]}. What do you think of this content? Please tell us whether you will approve the budget and provide feedback for improvement at that time." Note that [Text output by the content creation unit 2112] may contain specific text data. The document evaluation unit 2113 may also provide the received prompt to the approver agent to make an inquiry and output the inquiry result. The inquiry result may be an expected result of feedback from the approver. [Step S26-2: Inquiry to the senior approver agent]

[0063] The second agent model may be an agent model for a "senior approver." The document judgment unit 2113 may accept the prompt: "{[Text output by the content creation unit 2112]}. The approver has responded to this content as follows: {[Inquiry result from step S26-1]}. Please provide improvement feedback from a different perspective." Note that specific text data may be entered for [Text output by the content creation unit 2112] and [Inquiry result from step S26-1]. The document judgment unit 2113 may provide the received prompt to the senior approver agent to make an inquiry and output the inquiry result. The inquiry result may be an expected result of feedback from the senior approver.

[0064] Furthermore, the document determination unit 2113 may query the approver agent model by providing a prompt that includes the query result from step S26-2. This is expected to refine the feedback.

[0065] The second agent model may be an agent model of a "customer implementing the proposal." Note that "customer implementing the proposal" may be just one example of a target for implementing the measures. The second agent model may be an agent model of a "consumer consuming the proposal." The second agent model may be an agent model of a "media organization." The second agent model may be an agent model of a "person investing in the customer." For example, the data determination unit 2113 may conduct a survey of the agent model of a "consumer consuming the proposal."

[0066] If the second agent model does not approve the document (step S27: No), the process returns to step S25. Upon returning to step S25, the document creation unit 211 may obtain feedback from the second agent model regarding the generated document and have the first agent model revise the document based on the feedback. The first agent model may revise the document so that the proposal is approved by the "approver (and higher-level approvers)". The first agent model may revise the document so that the proposal is accepted by "customers implementing the proposal", "consumers consuming the proposal", "media organizations", "investors in the customer", etc. The document revised by the first agent model is referred to as the "revised document". The document creation unit 211 can increase the likelihood of the document being approved by reflecting feedback from agent models that mimic the target audience other than the proposer. The document creation unit 211 may repeat steps S25 and S26 until the second agent model approves the document.

[0067] The format conversion unit 2114 converts the modified document into a specified format (step S28). The format conversion unit 2114 may input the modified document and information indicating the predetermined format into the LLM and output the document converted to the LLM.

[0068] For example, the format conversion unit 2114 may convert a text-format document into a slide-format document. In this case, the format conversion unit 2114 may input the prompt "Please output {[text-format document]} in Marp format" to the LLM and have the LLM output a Marp format file. Note that specific text data may be entered in [text-format document].

[0069] For example, the format conversion unit 2114 may create a document in one text format from another text format document. In this case, the format conversion unit 2114 may input the text "{[Text format document]}. Output a talk script that matches this" into the LLM and have the LLM output a talk script.

[0070] Furthermore, the format conversion unit 2114 may convert the text-formatted material into image and video formatted material. The format conversion unit 2114 may also create a presentation manuscript that includes both text and images.

[0071] The format conversion unit 2114 may store the data 224 in the storage device 22. The format conversion unit 2114 may also store the data 224 in a storage device other than the storage device 22. The following describes the case where the data 224 is stored in the storage device 22.

[0072] The document presentation unit 212 presents documents in a specified format (step S29). The document presentation unit 212 may have the proposer agent model present the proposal. The proposer agent model may give a presentation using text. The proposer agent model may give a presentation using both text and audio. The document presentation unit 212 may have the proposer agent model speak the contents of a talk script. The document presentation unit 212 may also have the proposer agent model conduct a question and answer session. The document presentation unit 212 may present the documents as if a virtual proposer were speaking by moving the proposer's face image. The document presentation unit 212 may present a video of the slides that progresses at appropriate timings, synchronized with the speaking of the contents of the talk script. Providing video can reduce the need for proposal activities in a meeting format.

[0073] In this embodiment, the case in which the agent model is created within the document creation device 20 is described, but the document creation device 20 may also perform document creation operations using an agent model created outside of the document creation device 20. The above is merely one example of an agent model creation method, and an agent model created by a method other than the one described above may also be used. [2-4: Technical Effects]

[0074] In business decision-making processes, the goal is to determine whether or not to implement a measure. Implementing a measure involves personnel, resources, and costs, requiring approval from superiors or other approvers. Proposal activities are essential to gain approval. However, proposal activities, which involve manually creating documents and explaining them to approvers, are time-consuming and involve a lot of manual work.

[0075] The document creation device 20 described in this disclosure creates and proposes documents using at least a proposer agent model and an approver agent model, thereby reducing the effort required from proposers and approvers while proposing documents that align with their thinking. The document creation device 20 can propose documents to approvers that are suitable for their characteristics and requirements, thereby increasing the success rate of budget acquisition. Furthermore, since the document creation device 20 can act as an agent for the proposer's proposal activities (document presentation activities), it can reduce the effort required from the proposer. [3: Third Embodiment]

[0076] A third embodiment relating to a document creation device, a document creation method, and a recording medium will be described with reference to Figures 6 and 7. In the following description, the third embodiment relating to a document creation device, a document creation method, and a recording medium will be described using the document creation device 30. The third embodiment makes modifications based on the actual reactions of a human subject to the document. Hereinafter, when simply referred to as "subject," the subject may be a human. In addition, in the third embodiment, explanations that overlap with the descriptions of the first and second embodiments described above will be omitted as appropriate. In addition, in the drawings, parts common to the first and second embodiments will be denoted by the same reference numerals.

[0077] As shown in Figure 6, the arithmetic unit 21 of the document creation device 30 includes, as logical functional blocks, a document creation unit 311, a document presentation unit 212, and an agent model creation unit 213. The document creation unit 311 may also include a response acquisition unit 3115 in addition to a draft information receiving unit 2111, a content creation unit 3112, a document determination unit 3113, and a format conversion unit 2114. [3-1: Document creation method executed by the document creation device 30]

[0078] The response acquisition unit 3115 acquires information indicating the subject's reaction to the created document (step S31). Information indicating the subject's reaction is referred to as "reaction information". The response acquisition unit 3115 may acquire the document that the subject has reviewed. The subject may be someone other than the proposer. The subject may also be an approver.

[0079] The response acquisition unit 3115 may acquire at least one of the following as response information: text data, audio data, and image data. The response acquisition unit 3115 may acquire multiple types of response information. The text data may be data of comments from the subject. The text data may be data of the survey results. The audio data may include the audio of comments from the subject. The time when the subject made a comment may be associated with the section of the proposed material. The image data may be data of a still image or video. The image data may be data of the subject's face. The image data may be information that allows for the determination of the subject's facial expression.

[0080] The reaction acquisition unit 3115 extracts the subject's reactions that are useful for correcting the document (step S32). The reaction acquisition unit 3115 may also extract information that identifies the location within the document that corresponds to the subject's reaction that is useful for correcting the document. The location that corresponds to the subject's reaction that is useful for correcting the document is called the "correction location".

[0081] The reaction acquisition unit 3115 may acquire video data of the approver when the document is proposed as reaction information. The video data as reaction information may include scenes in which the approver smiles. The reaction acquisition unit 3115 may extract the approver's smile as a target reaction useful for revising the document and as a favorable reaction by the approver. The video data as reaction information may include scenes in which the approver frowns. The reaction acquisition unit 3115 may extract the approver's frown as a target reaction useful for revising the document and as an unfavorable reaction by the approver. Furthermore, the video data as reaction information may include scenes in which the approver is not concentrating on the proposal, such as looking at their mobile phone. The reaction acquisition unit 3115 may extract the approver's lack of concentration on the proposal as a target reaction useful for revising the document and as an unfavorable reaction by the approver. In addition, the reaction acquisition unit 3115 may extract the degree of favorable reactions and the degree of unfavorable reactions.

[0082] The content creation unit 3112 uses the first agent model to correct the parts that need to be corrected (step S33). The content creation unit 3112 may also receive a prompt such as "Content of the document: {[page n, line m of the document]}. The approver has given the following feedback: {[Approver Feedback]}. Please correct the content of the document to match the feedback." and input this into the first agent model. [Approver Feedback] may include information indicating whether the response is favorable or unfavorable, and the degree of the favorable or unfavorable response.

[0083] The document determination unit 2113 determines the corrected document using the second agent model (step S26). The format conversion unit 2114 converts the document to the specified format (step S28). The document presentation unit 212 presents the corrected document in the specified format (step S29).

[0084] Alternatively, in step S33, the content creation unit 3112 may receive the prompt "Content of the document: {[Page n, line m of the document]}. The approver has given the following feedback: {[Approver Feedback]}. At this time, we would like to revise the content of the document to match the feedback, do you have any questions?" and input this into the first agent model. The response acquisition unit 3115 may inquire with the approver agent about the questions output by the first agent model. The response acquisition unit 3115 may have the approver agent modify {[Approver Feedback]} in a way that is easy for the first agent model to understand. If the approver's response is an implicit response such as frowning, the response acquisition unit 3115 may receive the prompt "Guess the reason for the frowning and think of improvement feedback," and have the approver agent model, for example, make inferences. [3-2: Technical Effects]

[0085] The document creation device 30 for this disclosure makes revisions based on the actual reactions of the subject, who is a human being, to the document. Therefore, compared to cases where revisions based on the subject's actual reactions to the document are not made, it is possible to present a document that is more likely to be approved. [4: Note]

[0086] Some or all of the above embodiments may also be described as follows, but are not limited to the following: [Appendix 1] A document creation device comprising: a document creation means for creating a document using information about the document and an agent model that mimics the thinking of a subject given a role related to a proposal using the document; and a presentation means for presenting the created document. [Appendix 2] The document creation device according to Appendix 1, wherein the agent model includes a plurality of agent models, each of which mimics the business thinking and judgment tendencies of each of the plurality of subjects. [Appendix 3] The document creation device according to Appendix 2, wherein the plurality of subjects include the proposer of the proposal and the approver of the proposal, the plurality of agent models include a proposer agent model that mimics the business thinking and judgment tendencies of the proposer and an approver agent model that mimics the business thinking and judgment tendencies of the approver, and the document creation means causes the approver agent model to output feedback on the created document, and causes the proposer agent model to output the document modified based on the feedback. [Note 4] The document creation device according to Note 3, wherein the target persons include at least one of the following: a higher approver who approves the approver, a customer who implements the proposal, a consumer who consumes the proposal, a media organization, and a person who invests in the customer. [Note 5] The document creation device according to Note 1 or 2, wherein the agent model imitates the thoughts of a plurality of persons assigned specific roles. [Note 6] The document creation means converts the modified document into a predetermined format and outputs the converted document. [Note 7] The document creation device according to Note 6, wherein the document includes at least one of text, a talk script, slides, images, and videos. [Note 8] The document creation means acquires reaction information indicating the approver's reaction to the created document, extracts points for improvement in the document based on the reaction, and modifies the document.[Note 9] The document creation means is the document creation device described in Note 3, which causes the proposer agent model to modify the document until the approver agent model outputs approval feedback for the document. [Note 10] The document creation means is the document creation device described in Note 1 or 2, which creates the document using Large Language Models (LLMs). [Note 11] A computer-executed document creation method, which includes creating the document using information about the document and an agent model that mimics the thinking of a subject given a role related to a proposal using the document, and presenting the created document. [Note 12] A recording medium on which a computer program is recorded that causes a computer to execute a document creation method, which includes creating the document using information about the document and an agent model that mimics the thinking of a subject given a role related to a proposal using the document, and presenting the created document.

[0087] Furthermore, some or all of the configurations described in Appendices 2 to 10, which are subordinate to Appendice 1 above, may also be subordinate to Appendices 11 and 12, respectively, in the same manner as the subordinate relationships described in Appendices 2 to 10. Moreover, not limited to Appendices 1, 11, and 12, some or all of the configurations described as appendices may also be subordinate to various hardware, software, various recording means for recording software, or systems, without departing from the embodiments described above.

[0088] This disclosure may be modified as appropriate, insofar as it does not contradict the gist or idea of ​​the invention as can be inferred from the claims and the specification as a whole, and such modifications to the data preparation apparatus, data preparation method and program also fall within the technical concept of this disclosure.

[0089] 10, 20, 30 Document creation device 11, 211, 311 Document creation unit 12, 212 Document presentation unit 2111 Draft information reception unit 2112, 3112 Content creation unit 2113 Document judgment unit 2114 Format conversion unit 213 Agent model creation unit 2131 Metadata creation unit 3115 Response acquisition unit

Claims

1. A document creation device comprising: a document creation means for creating a document using information about the document and an agent model that mimics the thinking of a subject assigned a role related to a proposal using the document; and a presentation means for presenting the created document.

2. The data creation apparatus according to claim 1, wherein the agent model includes a plurality of agent models, and each of the plurality of agent models imitates the business thinking and judgment tendencies of each of the plurality of subjects.

3. The data creation device according to claim 2, wherein the plurality of subjects include the proposer of the proposal and the approver of the proposal, the plurality of agent models include a proposer agent model that mimics the business thinking and judgment tendencies of the proposer and an approver agent model that mimics the business thinking and judgment tendencies of the approver, and the data creation means causes the approver agent model to output feedback on the data it has created, and causes the proposer agent model to output the data that has been modified based on the feedback.

4. The data creation apparatus according to claim 3, wherein the subject persons include at least one of the following: a superior approver who approves the approver, a customer who implements the proposal, a consumer who consumes the proposal, a media organization, and a person who invests in the customer.

5. The data creation device according to claim 1 or 2, wherein the agent model imitates the thought process of a plurality of persons assigned specific roles, integrating their thoughts.

6. The document creation device according to claim 1 or 2, wherein the document creation means converts the modified document into a predetermined format and outputs the converted document.

7. The material creation apparatus according to claim 6, wherein the material includes at least one of text, talk scripts, slides, images, and videos.

8. The document creation apparatus according to claim 1 or 2, wherein the document creation means acquires reaction information indicating the reaction of the approver of the proposal to the created document, extracts points for improvement of the document based on the reaction, and modifies the document.

9. The document creation device according to claim 3, wherein the document creation means causes the proposer agent model to modify the document until the approver agent model outputs approval feedback for the document.

10. The data creation apparatus according to claim 1 or 2, wherein the data creation means creates the data using Large Language Models (LLM).

11. A computer-based method for creating a document, comprising creating the document using information about the document and an agent model that mimics the thinking of a subject assigned a role related to a proposal using the document, and presenting the created document.

12. A recording medium on which a computer program is stored that causes a computer to execute a method for creating a document, which includes creating the document using information about the document and an agent model that mimics the thinking of a subject assigned a role related to a proposal using the document, and presenting the created document.

Images