Development document creation device, development document creation method, and development document creation program

Abstract

The present invention provides a development document creation device, a development document creation method, and a development document creation program capable of creating a document specialized in system development. A development document creation device according to the present invention comprises: a development information acquisition unit that acquires development information; a first recording unit that records a prompt of a large-scale language model as template information in accordance with a development task; a task acquisition unit that acquires the development task from a user; a question generation unit that generates a first question on the basis of the template information corresponding to the acquired task; an answer acquisition unit that acquires a first user answer, to the first question, from the user; an answer generation unit that generates an answer as a solution on the basis of the first user answer; and a creation unit that creates a development document on the basis of the development information and the first user answer, to the first question, that has been acquired from the user.

Description

Development Document Creation Device, Development Document Creation Method, and Development Document Creation Program

[0001] The present invention relates to a development document creation device, a development document creation method, and a development document creation program, and more particularly to a development document creation device, a development document creation method, and a development document creation program capable of assisting in the creation of development documents using a large language model.

[0002] Conventionally, there is a technology that enables the easy execution of the order receiving and placing activities through sufficient support in the entire order receiving and placing process. For example, in the order receiving and placing support device disclosed in Patent Document 1, based on the order receiving and placing related information in the order receiving and placing database, it probabilistically predicts how likely the words and sentences given in the prompt are to occur as natural language, and learning and adjustment are performed to generate order receiving and placing support information. Thus, it has a large language model unit having a large language model specialized for the creation of order receiving and placing support information, an information reception unit that receives the user's response information in the conversation with the user, an initial setting prompt generation unit that instructs as a prerequisite for the generation of order receiving and placing support information in the large language model, and an order receiving and placing prompt generation unit that causes the large language model to generate order receiving and placing support information suitable for the order receiving and placing stage based on the prerequisite conditions.

[0003] However, the order receiving and placing support device disclosed in Patent Document 1 focused on conducting a conversation with the user in accordance with the order receiving and placing stage from the acceptance of the user's issues to the placement of an order with the vendor using a large language model specialized for the creation of order receiving and placing support information, but did not pay attention to the creation of documents specialized for system development, such as in-house planning documents, environment definition documents, requirement definition documents, design documents, and plan documents. Therefore, the emergence of a technology that enables the creation of documents specialized for development, such as in-house planning documents, environment definition documents, requirement definition documents, design documents, and plan documents, in system development has been desired.

[0004] Japanese Patent No. 7526415

[0005] Therefore, the present invention has been made in view of the above points, and provides a development document creation device, a development document creation method, and a development document creation program that enable the creation of development-specific documents such as in-house project proposals, environment definition documents, requirements definition documents, design documents, and plans in system development.

[0006] In other words, the development document creation device according to the first embodiment is a development document creation device that supports the creation of development documents using a large-scale language model, and is characterized by comprising: a development information acquisition unit that acquires development information relating to development to be described in the development document; a first recording unit that records prompts of a large-scale language model as template information according to the development task; a task acquisition unit that acquires development tasks from a user; a question generation unit that generates a first question using a large-scale language model based on the template information corresponding to the acquired task; an answer acquisition unit that acquires a first user answer to the first question from a user; an answer generation unit that generates an answer which is a solution based on the first user answer acquired by the answer acquisition unit; and a creation unit that creates a development document using a large-scale language model based on the development information and the first user answer to the first question acquired from the user.

[0007] A second embodiment is a development document creation device of the first embodiment, wherein the development document includes an in-house project proposal, an environment definition document, a requirements definition document, a design document, and a plan document, and the creation unit may create an in-house project proposal, create a current business flow, a list of issues, and an environment definition document based on the in-house project proposal, create a requirements definition document based on the current business flow, a list of issues, and an environment definition document, and create a design document and a plan document based on the created requirements definition document.

[0008] In the development document creation apparatus of the third embodiment, the first recording unit may record the introduction procedures and maintenance know-how of the corporate resource plan, and the creation unit may create a development document as a deliverable that corresponds to the recorded introduction procedures and maintenance know-how of the corporate resource plan.

[0009] In the development document creation device of the fourth embodiment, the creation unit may create a development document based on a first prompt, which is a pre-templated prompt for the introduction of a corporate resource plan, and a second prompt, which is based on dialogue with the user.

[0010] The development document creation device of the fifth embodiment may further include a second recording unit that stores prompts used to create development documents, created by the creation unit; a shared display unit configured to allow other users in the same team to view and execute projects handled by one user; and a reflection unit that reflects prompts when creating development documents other than the development document.

[0011] In the sixth embodiment of the development document creation apparatus, the question generation unit may further generate a second question that complements the first question based on the first user response from the user.

[0012] In the development document creation apparatus according to the seventh embodiment, the response acquisition unit may acquire a second user response to a second question from the user, and the creation unit may create a development document based on the development information, the first response, and the second response.

[0013] In the eighth embodiment of the development document creation device, the first prompt may include the conversation history between the user and the AI ​​necessary for creating the development document.

[0014] In the development document creation apparatus of the ninth embodiment, the answer generation unit may generate answers corresponding to the task category.

[0015] In the development document creation apparatus of the tenth embodiment, the response generation unit may generate a response according to the current date.

[0016] The eleventh embodiment of the development document creation method is a method for supporting the creation of a development document using a large-scale language model, characterized in that it causes a computer to execute: a development information acquisition step of acquiring development information relating to development to be described in the development document; a recording step of recording prompts of a large-scale language model as template information according to the development task; a task acquisition step of acquiring development tasks from a user; a question generation step of generating a first question using a large-scale language model based on template information corresponding to the acquired task; an answer acquisition step of acquiring a first user answer to the first question from a user; an answer generation step of generating an answer which is a solution based on the first user answer acquired by the answer acquisition unit; and a creation step of creating a development document using a large-scale language model based on the development information and the first user answer to the first question acquired from the user.

[0017] The development document creation program according to the twelfth embodiment is a program that supports the creation of development documents using a large-scale language model, and is characterized in that it implements the following functions on a computer: a development information acquisition function that acquires development information related to development to be described in the development document; a recording function that records prompts of a large-scale language model as template information according to the development task; a task acquisition function that acquires development tasks from a user; a question generation function that generates a first question using a large-scale language model based on template information according to the acquired task; an answer acquisition function that acquires a first user answer to the first question from a user; an answer generation function that generates an answer which is a solution based on the first user answer acquired by the answer acquisition unit; and a creation function that creates a development document using a large-scale language model based on the development information and the first user answer to the first question acquired from the user.

[0018] The development document creation device according to the present invention is a development document creation device that supports the creation of development documents using a large-scale language model, and is characterized by comprising: a development information acquisition unit that acquires development information related to development to be described in the development document; a first recording unit that records prompts of a large-scale language model as template information according to the development task; a task acquisition unit that acquires development tasks from a user; a question generation unit that generates a first question using a large-scale language model based on the template information corresponding to the acquired task; an answer acquisition unit that acquires a first user answer to the first question from a user; an answer generation unit that generates an answer which is a solution based on the first user answer acquired by the answer acquisition unit; and a creation unit that creates a development document using a large-scale language model based on the development information and the first user answer to the first question acquired from the user, thereby enabling the creation of development-specific documents such as in-house project proposals, environment definition documents, requirements definition documents, design documents, and plans in system development.

[0019] Furthermore, the development document creation method and development document creation program according to the present invention, like the development document creation apparatus according to the present invention, enable the creation of development-specific documents such as in-house project proposals, environment definition documents, requirements definition documents, design documents, and plans for system development.

[0020] Figure 1 is a diagram illustrating the schematic configuration of a development document creation system according to an embodiment of the present invention. Figure 2 is a block diagram showing the functional configuration of a development document creation device according to an embodiment of the present invention. Figure 3 is a diagram showing the specifications of a development document creation device according to an embodiment of the present invention. Figure 4 is a diagram showing the screen transitions of a development document creation device according to an embodiment of the present invention. Figure 5 is a diagram showing the chat screen of a development document creation device according to an embodiment of the present invention. Figure 6 is a diagram showing a screen image of a development document creation device according to an embodiment of the present invention. Figure 7 is a diagram showing the templates of a development document creation device according to an embodiment of the present invention. Figure 8 is a diagram showing the project templates of a development document creation device according to an embodiment of the present invention. Figure 9 is a diagram showing the process templates of a development document creation device according to an embodiment of the present invention. Figure 10 is an example of a flowchart of a development document creation method according to an embodiment of the present invention.

[0021] Hereinafter, an embodiment of the present invention will be described with reference to the figures. Note that the figures are examples, and the present invention is not limited to what is shown in the figures. For example, the illustrated development document creation device (server), large-scale language model server, number of user terminals, dataset (table), flowchart, and display screen examples are examples, and the present invention is not limited to these.

[0022] <System Configuration> Figure 1 shows an example of the configuration of a development document creation system according to one embodiment of the present invention. The development document creation system 600 includes at least a server (development document creation device) 100, a user terminal 200, and a large-scale language model server 300.

[0023] Server 100 functions as a platform that provides development document creation support services and outputs the created development documents to user terminals 200 connected via network 500. Here, "development documents" refer to development-specific documents in system development, and may include in-house project proposals, environment definition documents, requirements definition documents, design documents, and plan documents. As will be described in detail later, Server 100 creates development documents while exchanging information with user terminals 200. At this time, Server 100 creates development documents using the functions of the large-scale language model provided by the large-scale language model server 300 in the cloud.

[0024] Large language models (LLMs) are a type of generative AI (Artificial Intelligence) specialized in natural language processing (NLP). They are language models that achieve advanced natural language generation (NLG) by being constructed through deep learning of vast amounts of text data. A language model, in this context, refers to a mechanism that understands patterns in natural language used by humans and predicts the word that will follow a given word by analyzing the probability of word occurrence. Examples of large language models include ChatGPT (Chat Generative Pre-trained Transformer) (registered trademark), PaLM2 (registered trademark), LaMDA2 (registered trademark), etc. The words and sentences used as instructions to input into these large-scale language models are generally referred to as prompts. Note that the language models used are not limited to large-scale language models; small-scale language models (SLMs) may also be used. Small-scale language models are lightweight language models that excel at processing specific tasks. For user file input (file attachments), the system supports text formats (txt, html, xml, csv), Office files (Excel, PowerPoint, Word), PDFs, and images (jpg, png). These files are converted to text format and saved as conversation history linked to the user's statements. For images in particular, the system acquires two types of information: "OCR result information" and "information converted from the image to text format," and stores them within the conversation history. This allows the AI ​​to provide more accurate responses even to images containing text.

[0025] Server 100 can be any information processing device capable of realizing the functions described in each embodiment. Server 100 includes, for example, a server device, a computer (e.g., a desktop, laptop, tablet, etc.), or a communication platform. In Figure 1, Server 100 and the large-scale language model server 300 are shown as separate entities, but this is not limited to them. That is, each function described as being provided by Server 100 may be realized by multiple servers or by a single server. Also, the functions provided by the large-scale language model server 300 may be realized as some of the functions of Server 100. Server 100 may be, for example, a distributed server system that operates cooperatively by communicating over a network, or a so-called cloud server. That is, Server 100 is not limited to a physical server, but may also include a virtual server created by software. Furthermore, Server 100 can be any information processing device capable of realizing the functions described in each embodiment, and may include, for example, a server device, a computer (not limited to, but examples include a desktop, laptop, tablet, etc.), a communication platform, etc.

[0026] The user terminal 200 is a communication terminal used by users of the development document creation support system 600. It has an application installed for using the development document creation support service and is capable of sending and receiving various types of information with the server 100. However, installation of the application on the user terminal 200 is not mandatory. Users may access a web page provided by the server 100 for using the development document creation support service via a web browser or the like from the user terminal 200 and send and receive various types of information with the server 100.

[0027] In Figure 1, a notebook personal computer is shown as the user terminal 200, but any terminal capable of performing the functions described below may be used as the user terminal 200. For example, the user terminal 200 may be a computer (e.g., a tablet device, a desktop personal computer), a smartphone, etc.

[0028] The development documentation support system 600 may further include a database server (not shown) that stores various types of information (data) used by the development documentation support system 600. The data used by the development documentation support system 600 may include, for example, prompts to be input into a large-scale language model or user information.

[0029] Network 500 may include wireless networks and wired networks, and may be, for example, wireless LANs (WLANs), wide area networks (WANs), ISDNs (integrated service digital networks), wireless LANs, CDMA (code division multiple access), LTE (long term evolution), LTE-Advanced, 4th generation communication (4G), 5th generation communication (5G), and 6th generation communication (6G) and later mobile communication systems. Network 500 is not limited to these examples, and may also include, for example, a Public Switched Telephone Network (PSTN), Bluetooth (registered trademark), optical fiber lines, satellite communication networks, etc. Network 500 may also be a combination of these.

[0030] Figure 2 is a functional block diagram of the development document creation device (server) 100 of the embodiment. The development document creation device 100 of this embodiment includes a development information acquisition unit 10, a first recording unit 20, a task acquisition unit 30, a question generation unit 40, an answer acquisition unit 50, an answer generation unit 60, a creation unit 70, a second recording unit 80, a shared display unit 90, and a reflection unit 110. These are functional units realized by the CPU executing an information processing program deployed on RAM. The development document creation device 100 is further equipped with storage units such as ROM, RAM, SSD, HDD, and various interfaces (not shown).

[0031] The development information acquisition unit 10 acquires development information related to development described in development documents. Development documents include, for example, in-house project proposals, environment definition documents, requirements definition documents, design documents, and plans. Development information includes information such as enterprise resource planning (ERP) implementation know-how, ERP maintenance know-how, standard documents, customer ASIS business documents, and security masks. A security mask is a technology that protects data confidentiality by hiding, rewriting, or changing specific information while maintaining the original data structure and meaning, using data registered in a master of mask keywords.

[0032] The first recording unit 20 records prompts for the large-scale language model as template information according to the development task. Development tasks include, for example, creating requirements definition documents, creating a list of new functions, creating basic design documents, creating detailed design documents, and performing programming. By providing the development information acquisition unit 10 and the first recording unit 20, the development document creation device 100 can create development documents by statistically processing them at the word level, corresponding to specialized know-how. This makes it possible to perform tasks that are not dependent on the user's ERP expertise or AI skills.

[0033] The task acquisition unit 30 acquires development tasks from the user. As mentioned above, development tasks include, for example, creating a requirements definition document, creating a list of new features, creating a basic design document, creating a detailed design document, and performing programming. The user selects the task they want to perform from a template. The task acquisition unit 30 acquires tasks related to the template selected by the user. Since the work prompts for ERP implementation are pre-templated, when a task is selected from a template, the AI ​​immediately asks questions and the user can start working.

[0034] The question generation unit 40 generates a first question using a large-scale language model based on template information corresponding to the acquired task. The question generation unit 40 further generates a second question that complements the first question based on the first user response from the user. The first question is, for example, "Please tell us about the problems your company is facing," "What problems are there with your current systems and operations?", and "If you know the cause, please tell us as well." The first user response is, for example, "Regarding revenue management by department and segment, it takes a long time to collect revenue. I think the reason is that there is no system in place," "We do not know the cost of individual products. This is because we do not allocate common costs," "We do not know the yield rate," "The approval process is paper-based," "We still have legacy systems," and "The maintenance deadline for the ERP we are using is approaching." The second set of questions might include, for example, "You mentioned that revenue collection takes a long time. What specific processes are used to collect revenue? Which departments or segments are particularly problematic?", "You mentioned that common costs are not being allocated. How are costs currently calculated? What criteria do you plan to establish for allocating common costs?", "You mentioned that yield rates are not being tracked. How is production data currently managed? How is the data necessary for calculating yield rates collected?", "You mentioned that approval processes are paper-based. What kind of approval processes exist? How are the departments and personnel involved in approvals involved?", and "You mentioned that legacy systems remain. Which specific systems are still in use, and for what tasks are they used? Which parts of these legacy systems are problematic?". The question generation unit 40 then continues to generate questions, delving deeper with the third, fourth, and subsequent questions.

[0035] The response acquisition unit 50 acquires the first user response to the first question from the user. The response acquisition unit 50 acquires the second user response to the second question from the user. Examples of the second user response include: "It involves email correspondence and meetings. In total, it takes about 20 hours per month. The problem is with Department A.", "Costs are calculated using the XX system, but I don't know the details (I can't) without checking with the person in charge of the package. Confirmation is necessary.", "We use the YY system for production management. Data collection is done manually.", "There are many subordinate-to-superior approval processes. However, there are also flows that require final approval from the Information Systems Department or Management Department.", "We still have an inventory management system. It's an independent system and doesn't have sufficient integration with other existing systems, which is causing problems.", "The support period ends in September 2025, and once it expires, the vendor will no longer provide support, which is a problem."

[0036] The response generation unit 60 generates a solution response based on the first user response acquired by the response acquisition unit. It generates a solution response even if the user gives an irrelevant response.

[0037] The response generation unit 60 then generates responses for the third response, the fourth response, and so on. These responses from the third response onward represent the "ideal state." Examples of the "ideal state" include: "Utilize the revenue management information of software XX to build an automated revenue collection process," "Use the cost management function of software XX to set the allocation criteria for common costs," "Utilize the production management function of software XX to automatically collect the data necessary for yield calculation," "Use the workflow function of software XX to build an electronic approval process," "Plan a full migration to software XX and begin integration," and "Proceed with the migration to software XX as quickly as possible to avoid maintenance risks."

[0038] The response generation unit 60 generates responses corresponding to the task category. To stabilize the AI's behavior, a mechanism is included to recognize the genre of the current task. Specifically, it extends the "category" found in general systems and links it to system prompts for the AI. This allows the AI ​​to correctly recognize the topic and behave appropriately. If a category is set, the AI ​​is automatically instructed to act as a specialist in that category. Multiple categories can be set. Based on the category and the automatically given system prompt, the AI ​​will act as a specialist in software XX. Without this specification, it will start talking about general systems.

[0039] The answer generation unit 60 generates an answer according to the current date. Since AI typically recognizes the knowledge cutoff date as the current date, the program is programmed to correctly recognize the current date by providing it with current date information. Even if a user comments without providing date information, the AI ​​correctly understands the date, so the entry date is entered correctly.

[0040] The response generation unit 60 incorporates a mechanism that uses the RAG concept to provide responses while taking into account search results from search resources. RAG stands for Retrievable-Augmented Generation, a framework that improves output accuracy, which is difficult to control with prompts alone, by combining LLM (Large-Scale Language Model) text generation with the retrieval of reliable external information. By allowing the AI ​​to select the search target, it solves problems such as decreased response accuracy due to unnecessary information and exceeding the number of tokens (speed reduction, errors) due to unnecessary information. However, this alone can lead to the problem that the AI ​​will not perform the search that you want it to perform. Therefore, the development document creation device 100 allows you to clearly specify which searches you want to be performed and which parts you want to leave to the AI's judgment.

[0041] One way to retrieve more important data from a graph database is that, generally, searching for paths in a graph database without specifying anything will yield a large number of paths. However, because AI has a limit on the number of tokens it can use, it is necessary to retrieve only the important paths. The development document creation device 100 incorporates attribute information related to importance in the edge property information, and has a mechanism to refine the searched information (i.e., retrieve only the information with high importance). When a question requiring a search of the graph database is asked, the AI ​​determines that "a search of the graph database is necessary," and the search process to the graph database is executed internally, and an answer that takes into account the search results of the graph database is output. The relationships between intermediate tables can also be clearly answered. This is something that cannot be answered with simple text search or vector search. For example, when searching for a path between A and C for things related to A→B→C, such as table relationships, conventional text search or vector search had the problem of difficulty in obtaining the data corresponding to B. The development document creation device 100 addresses this problem by storing data in the graph database and performing a RAG (Random Aggregation) from there, thereby providing the intended answer. This makes it easy to search for relationships between tables included in packaged software such as ERP systems (which are pre-bundled), and enables AI to write the correct queries.

[0042] The creation unit 70 creates development documents using a large-scale language model based on development information and the first user's response to the first question obtained from the user. The creation unit 70 creates an in-house project proposal, creates a current business flow, a list of issues, and an environment definition document based on the in-house project proposal, creates a requirements definition document based on the current business flow, a list of issues, and an environment definition document, and creates a design document and a plan document based on the created requirements definition document. The creation unit 70 produces development documents as deliverables that correspond to the recorded Enterprise Resource Planning (ERP) implementation procedures and ERP maintenance know-how. The creation unit 70 creates development documents based on a first prompt (background prompt) which is a pre-templated template of the ERP implementation work prompts, and a second prompt (front prompt) which is based on dialogue with the user. The creation unit 70 creates development documents based on development information and the first and second responses. The first prompt (background prompt) includes the conversation history between the user and the AI ​​necessary for creating the development document. The conversation is a dialogue between the user and the AI, and there may be multiple users. In other words, examples include statements from user A, statements from the AI, statements from user B, and statements from the AI.

[0043] The second recording unit 80 stores the prompts created by the creation unit 70 that were used to create the development documents. The prompts created by the creation unit 70 that were used to create the development documents are the prompts from the previous process, such as the ASIS function list, the new system requirements definition document, the new system non-functional requirements, the issue list, the organization / person in charge list, and the authority configuration list. The AI ​​can process these previous process prompts and the resulting development documents statistically at the word level to create the documents. For example, if the current process is a task to create a basic design document, the AI ​​will refer to the requirements definition prompt, the latest requirements definition document, the issue list, etc., and start the basic design task for the add-on function to create the basic design document.

[0044] The shared display section 90 is configured such that projects assigned to one user can be viewed and executed by other users within the same team. On the home screen, not only the projects the user is working on but also the projects of other members can be displayed. This enables, for example, "a veteran employee takes over and executes the work of a new employee midway" or "a manager checks the work status of each member." When the filter narrowed down to logged-in users is removed, projects of users other than oneself are also displayed and can be viewed and executed. Since it is equipped with a team member sharing mode, work among members can be shared. Also, all information within the project can be referred to and utilized by AI.

[0045] The reflection section 110 reflects prompts when creating other development documents different from the development document. For example, if the previous process project is the creation of a requirements definition document, the AI can create prompts such as the AS-IS function list, new system requirements definition document, new system non-functional requirements, issue list, organization / responsible person list, authority configuration list, etc., and the requirements definition document as the resulting development document by statistically processing them at the word level. And, for example, if another development document to be created in the current process is a basic design document, it refers to the requirements definition prompt, the latest requirements definition document, the issue list, etc., starts the basic design task of the add-on function, and creates the basic design document.

[0046] Figure 3 shows the specifications of a development document creation device 100 according to an embodiment of the present invention. As introduction process templates 1, a requirements definition template 11, a function list template 12, an additional development basic design template 13, and an additional development program template 14 are provided. The worker 2 selects the template suitable for the task they wish to perform from these templates. Meanwhile, the development document creation device 100 acquires Enterprise Resource Planning (ERP) implementation knowledge and system prompts 3. The Enterprise Resource Planning (ERP) implementation knowledge and system prompts 3 include ERP implementation know-how 31, ERP standards 32, and document templates 33. The development document creation device 100 acquires these ERP implementation know-how 31, ERP standards 32, and document templates 33, as well as the latest web information 34. Then, upon consultation or request from the worker 2, it prompts for necessary information or performs tasks. It also navigates the worker 2 and provides deliverables 4 to the worker 2. Deliverables 4 include a requirements definition document 41, a design document 42, and source code 43. Source code is a string of characters (text or text file) written in a programming language that represents a computer program.

[0047] Figure 4 shows the screen transitions of the development document creation device 100 according to an embodiment of the present invention. The left side shows the execution environment 5, and the right side shows the server setting environment 6. When logging into the execution environment 5, the project list screen 51 is displayed. On the project list screen 51, if worker 2 selects the execution screen 52, the execution screen 52 is displayed, and if worker 2 selects the settings screen 53, the settings screen 53 is displayed. On the settings screen 53, mask keyword registration 54 is possible. Mask keywords allow you to register keywords that are hidden so that other users cannot understand them. For example, if personal information such as "Yamada Hanako" is mistakenly entered in the chat field, this is a master function to automatically convert it to a different string "XX XX" before sending it to the AI. In this case, the pair of "Yamada Hanako" and "XX XX" is registered in advance in the mask keyword master.

[0048] When operator 2 logs in to the server setting environment 6, the management screen 61 is displayed. On the management screen 61, operator 2 can perform various registrations. The various registrations include project template registration 62, process registration 63, task registration 64, category registration 65, publication destination registration 66, and license registration 67. In the project template registration 62, templates for each project can be registered. In the process registration 63, each process can be registered. In the task registration 64, each task can be registered. In the category registration 65, each category can be registered. Regarding the publication destination registration 66 and the license registration 67, first, publication destinations (publication destination A, publication destination B,...) and licenses (license X, license Y,...) are registered, and then each license is assigned to a publication destination.

[0049] Referring to Figure 5, the chat screen of the development document creation device according to an embodiment of the present invention will be described. Figure 5 is a diagram showing the chat screen of the development document creation device according to an embodiment of the present invention. The chat screen shown in Figure 5 represents the process of a hearing to create a list of issues. The top section shows questions from the AI, and the middle section shows answers from the user. Questions from the AI ​​are well-structured, such as, as shown in the top section of Figure 5, "Please tell us about the problems your company is facing," "What are the problems with your current system and operations?", and "If you know the causes, please also tell us." The user's answers to these questions are well-structured, such as, as shown in the middle section of Figure 5, "We are unable to manage the profit and loss for each construction project," "The current system, which was developed from scratch, can calculate profit and loss by organization, but it does not have a function to manage it by project, so we do not know the profit and loss for each project," and "In the new system, we would like to use the standard functions to aggregate sales and costs by project and manage the profit and loss for each project." The AI's question at the top and the user's answer in the middle do not necessarily correspond one-to-one. The user can also respond with a request not asked by the AI, such as, "In the new system, I would like to use standard functions to aggregate sales and costs by project and manage profit and loss for each project." The AI ​​is designed to respond to what the user wants to write, taking that into account, and to propose the "ideal state." The bottom section shows that the user has entered the first answer to the AI's question, "We are not currently managing profit and loss by construction project," and is about to enter the second answer as a message in the comments section. The user can also upload a file when answering. If the user is finished with this second answer, they press the "Complete" button in the lower right to proceed to the next task. If they want to enter a third answer, they press the "Continue" button in the lower right to enter the answer.

[0050] Figure 6 shows a screen image of the development document creation device 100 according to an embodiment of the present invention. Figure 6 is an image of the project list screen 7. First, if you check "Show only projects I have created", only projects you have created will be displayed. The first column from the left displays the ID 71, the second column from the left displays the project name 72, the third column from the left displays the template 73, the fourth column from the left displays the target category 74, the fifth column from the left displays the creator 75, and the sixth column from the left displays the creation date 76. These items can be paginated and sorted. Clicking the "+" button displays the new project creation screen 8 in a pop-up window. After changing the selection of the template 81, pressing the confirmation button 82 displays the related processes. The first column from the left displays the sequence 83, the second column from the left displays the process ID 84, and the third column from the left displays the project name 85.

[0051] Figure 7 shows a template for a development document creation device 100 according to an embodiment of the present invention. Process templates 91-1 to 91-3 are linked under project template 91, and each process template 91-1 to 91-3 contains a template for each task. For example, as shown in Figure 7, process template 91-1 contains tasks 1-1 to 1-3 as templates for task 1. Process template 91-2 contains tasks 2-1 and 2-2 as templates for task 2.

[0052] Figure 8 shows a project template for a development document creation device 100 according to an embodiment of the present invention. In Figure 8, project template 91 is shown. In this case, the screen for creating the issue list with ID 13 is displayed. The license and category are displayed. In the central table, the vertical axis is number, and the horizontal axis from left to right is order, process ID, process name, and process category. For example, as shown in Figure 8, item 1 is the creation of the issue list with process ID 31, and the process category is software XX. Item 2 is the consideration of a solution to the issue with process ID 32, and the process category is software XX. Item 3 is the output of the issue list file with process ID 37, and no process category is assigned. The order of these can be changed using the upward and downward arrows. For example, clicking on the process name "Create Issue List" displays the process template shown in Figure 9. Note that the screen shown in Figure 8 is not a screen used by the user, but a screen used by the administrator for template settings.

[0053] Figure 9 shows a process template for a development document creation device 100 according to an embodiment of the present invention. In Figure 9, process template 91 is shown. In this case, process ID 31, "Creation of Issue List," is displayed. The category is Software XX. In the central table, the vertical axis is number, and the horizontal axis, from left to right, is order, task ID, task name, and task category. For example, as shown in Figure 8, task 1 is task ID 86, issue confirmation, and the task category is Software XX. Task 2 is task ID 87, issue list creation, and the task category is Software XX. Task 3 is task ID 88, issue list summary, and the task category is Software XX. The order of these can be changed using the upward and downward arrows. Note that the screen shown in Figure 9 is not a screen used by the user, but a screen used by the administrator for template settings.

[0054] Referring to Figure 10, the development document creation program according to the embodiment will be described together with the development document creation method. Figure 10 is an example of a flowchart of the development document creation method according to the embodiment. The development document creation method is executed by the CPU of the development document creation device 100 based on the development document creation program. The development document creation program includes a development information acquisition step S10, a recording step S20, a task acquisition step S30, a question generation step S40, an answer acquisition step S50, an answer generation step S60, and a creation step S70, etc. The development document creation program causes the CPU of the development document creation device 100 to implement the development information acquisition function, recording function, task acquisition function, question generation function, answer acquisition function, answer generation function, and creation function, etc. These functions are executed in the order shown in the flowchart of Figure 9, but the order can be changed as appropriate. Note that each function overlaps with the description of the various functional parts of the development document creation device 100 described above, so a detailed explanation will be omitted.

[0055] The development information acquisition function acquires development information related to development to be described in the development document (Step S10: Development Information Acquisition Step).

[0056] The recording function records prompts from a large-scale language model as template information according to the development task (step S20: recording step).

[0057] The task acquisition function acquires development tasks from the user (Step S30: Task Acquisition Step).

[0058] The question generation function generates a first question using a large-scale language model based on template information corresponding to the acquired task (Step S40: Question generation step).

[0059] The response acquisition function acquires the first user response to the first question from the user (Step S50: Response Acquisition Step).

[0060] The answer generation function generates a solution answer based on the first user answer obtained by the answer acquisition function (step S60: answer generation step). Note that the steps from question generation step S40 to answer generation step S60 are not performed only once, but are usually repeated 1 to N times. If an answer is generated in answer generation step S60, the process returns to question generation step S40, a second question is generated, the second user answer is obtained in answer acquisition step S50, and the answer is generated in answer generation step S60. From the second time onward, the AI's answer and the next question may be output simultaneously. The user decides whether to proceed from answer generation step S60 to the next creation step S70 or return to question generation step S40.

[0061] The creation function creates a development document using a large-scale language model based on the development information and the first user's response to the first question obtained from the user (Step S70: Creation Step).

[0062] According to each aspect of this disclosure described above, it becomes possible to create development-specific documents such as in-house project proposals, environment definition documents, requirements definition documents, design documents, and plans for system development.

[0063] The development documentation creation program for the embodiment can be implemented on the CPU using a computer language such as Arduino® IDE.

[0064] [Functions and Circuitry] Next, the functions and circuitry of the development document creation device 100 described above will be explained. Each functional part of the CPU of the development document creation device 100 may be implemented as a function of a computer's arithmetic processing unit, etc. That is, the CPU may be implemented as a development information acquisition function, recording function, task acquisition function, question generation function, answer acquisition function, answer generation function, and creation function, respectively, by a computer's arithmetic processing unit, etc. The development document creation program can enable a computer to implement each of the above-mentioned functions. The development document creation program may be recorded on a computer-readable non-temporary storage medium, such as memory, a solid-state drive, a hard disk drive, or an optical disc. The storage medium may be rephrased as, for example, a non-temporary computer-readable medium that stores the development document creation program. The development document creation program may also be transmitted online. Furthermore, the computer's arithmetic processing unit, etc. described above may be composed of, for example, an integrated circuit. That is, the CPU of the development document creation device 100 may be implemented as a development information acquisition circuit, recording circuit, task acquisition circuit, question generation circuit, answer acquisition circuit, answer generation circuit, and creation circuit that constitute a computer's arithmetic processing unit, etc.

[0065] Furthermore, the present invention is not limited to the development document creation apparatus 100, development document creation method, and development document creation program according to the above-described embodiment, and can be implemented by various other modifications or applications without departing from the gist of the present invention as described in the claims. Also, although the word "information" is used in the above-described embodiment, the word "information" can be replaced with "data," and the word "data" can be replaced with "information."

[0066] In this embodiment, the device is referred to as a development document creation device 100, but it is not limited to development documents; it can also be used to create, program, and review documents other than development documents, and its applications are not particularly limited.

[0067] [Aspects and Effects of this Embodiment] Next, an aspect of this embodiment and the effects of each aspect will be described. Note that the aspects described below are examples as of the time of filing, and this embodiment is not limited to the aspects described below. That is, this embodiment is not limited to the aspects described below, and may be realized by appropriately combining the parts described above. Also, a lower-level aspect may be referenced in any of the higher-level aspects. Furthermore, the effects of this embodiment described below are examples, and the effects of each aspect are not limited to those described below. Also, each aspect may have at least one of the effects described below, for example.

[0068] (Aspect 1) One aspect of the development document creation device is a development document creation device that supports the creation of development documents using a large-scale language model, comprising: a development information acquisition unit that acquires development information related to development to be described in the development document; a first recording unit that records prompts of the large-scale language model as template information according to the development task; a task acquisition unit that acquires development tasks from the user; a question generation unit that generates a first question using the large-scale language model based on the template information corresponding to the acquired task; an answer acquisition unit that acquires a first user answer to the first question from the user; an answer generation unit that generates an answer which is a solution based on the first user answer acquired by the answer acquisition unit; and a creation unit that creates a development document using the large-scale language model based on the development information and the first user answer to the first question acquired from the user. As a result, the development document creation device can create development-specific documents such as in-house project proposals, environment definition documents, requirements definition documents, design documents, and plans for system development. The same applies to the development document creation method and the development document creation program.

[0069] (Aspect 2) In one aspect of the development document creation device, the development document includes an in-house project proposal, an environment definition document, a requirements definition document, a design document, and a plan document. The creation unit may create an in-house project proposal, create a current business flow, a list of issues, and an environment definition document based on the in-house project proposal, create a requirements definition document based on the current business flow, a list of issues, and an environment definition document, and create a design document and a plan document based on the created requirements definition document. This enables the development document creation device to create development-specific documents such as an in-house project proposal, an environment definition document, a requirements definition document, a design document, and a plan document for system development.

[0070] (Aspect 3) In one aspect of the development document creation device, the first recording unit may record the introduction procedures and maintenance know-how of the corporate resource plan, and the creation unit may create a development document as a deliverable that corresponds to the recorded introduction procedures and maintenance know-how of the corporate resource plan. This enables the development document creation device to perform tasks that are not dependent on the user's ERP expertise or AI skills.

[0071] (Aspect 4) In one embodiment of the development document creation device, the creation unit may create a development document based on a first prompt, which is a pre-templated prompt for the introduction of the corporate resource plan, and a second prompt, which is based on dialogue with the user. This enables the development document creation device to perform tasks that are not dependent on the user's ERP expertise or AI skills.

[0072] (Aspect 5) A development document creation device in one aspect may further include a second recording unit that stores prompts used to create development documents, created by a creation unit; a shared display unit configured to allow other users in the same team to view and execute projects handled by one user; and a reflection unit that reflects prompts when creating development documents other than the development document. This enables the development document creation device to perform tasks that are not dependent on the user's ERP expertise or AI skills.

[0073] (Aspect 6) In one aspect of the development document creation device, the question generation unit may further generate a second question that complements the first question based on the first user response from the user. This enables the development document creation device to more accurately create development-specific documents such as in-house project proposals, environment definition documents, requirements definition documents, design documents, and plans for system development.

[0074] (Aspect 7) In one embodiment of the development document creation device, the response acquisition unit may acquire a second user response to a second question from the user, and the creation unit may create a development document based on the development information, the first response and the second response. This enables the development document creation device to more accurately create development-specific documents such as in-house project proposals, environment definition documents, requirements definition documents, design documents and plans for system development.

[0075] (Aspect 8) In one embodiment of the development document creation device, the first prompt may include the conversation history between the user and the AI ​​necessary for creating the development document. This enables the development document creation device to more accurately create development-specific documents such as in-house project proposals, environment definition documents, requirements definition documents, design documents, and plans for system development.

[0076] (Aspect 9) In one embodiment of the development document creation device, the answer generation unit may generate answers corresponding to the task category. This enables the development document creation device to more accurately create development-specific documents such as in-house project proposals, environment definition documents, requirements definition documents, design documents, and plans for system development.

[0077] (Aspect 10) In one embodiment of the development document creation device, the response generation unit may generate responses according to the current date. This enables the development document creation device to more accurately create development-specific documents such as in-house project proposals, environment definition documents, requirements definition documents, design documents, and plans for system development.

[0078] 1. Implementation Process Template 2. Worker 3. ERP Implementation Knowledge and System Prompts 4. Navigation and Deliverables 5. Execution Environment 6. Server Configuration Environment 7. Project List Screen 8. New Project Creation Screen 10. Development Information Acquisition Unit 11. Requirements Definition Document Template 12. Function List Template 13. Additional Development Basic Design Document 14. Additional Development Program 20. First Recording Unit 30. Task Acquisition Unit 31. ERP Implementation Know-how 32. ERP Standards 33. Document Templates 34. Latest Web Information 40. Question Generation Unit 50. Answer Acquisition Unit 51. Project List Screen 52. Execution Screen 53. Settings Screen 54. Mask Keyword Registration 60. Answer Generation Unit 61. Management Screen 62. Project Template Registration 63. Process Registration 64. Task Registration (Prompt) 65. Category Registration 66. Publication Destination Registration 67. License Registration 70. Creation Unit 71. ID 72. Name 73. Template 74 Target Category 75 Creator 76 Creation Date 80 Second Record Section 81 Template 82 Confirmation 83 Sequence 84 Process ID 85 Project Name 90 Shared Display Section 91 Project Template 91-1 Process Template 1 91-2 Process Template 2 91-3 Process Template 3 100 Development Document Creation Device 110 Reflection Section 200 User Terminal 300 Large-Scale Language Model Server 500 Network

Claims

1. A development document creation device that supports the creation of development documents using a large-scale language model, comprising: a development information acquisition unit that acquires development information relating to development to be described in the development document; a first recording unit that records prompts of the large-scale language model as template information according to the development task; a task acquisition unit that acquires the development task from a user; a question generation unit that generates a first question using the large-scale language model based on the template information corresponding to the acquired task; an answer acquisition unit that acquires a first user answer to the first question from a user; an answer generation unit that generates an answer which is a solution based on the first user answer acquired by the answer acquisition unit; and a creation unit that creates a development document using the large-scale language model based on the development information and the first user answer to the first question acquired from the user.

2. The development document creation apparatus according to claim 1, wherein the development document includes an in-house project proposal, an environment definition document, a requirements definition document, a design document, and a plan document, and the creation unit creates the in-house project proposal, creates a current business flow, a list of issues, and the environment definition document based on the in-house project proposal, creates the requirements definition document based on the current business flow, the list of issues, and the environment definition document, and creates the design document and the plan document based on the created requirements definition document.

3. The development document creation apparatus according to claim 1, wherein the first recording unit records the introduction procedures and maintenance know-how of the corporate resource plan, and the creation unit creates the development document corresponding to the recorded introduction procedures and maintenance know-how of the corporate resource plan as a deliverable.

4. The development document creation device according to claim 3, wherein the creation unit creates the development document based on a first prompt which is a pre-templated prompt for the introduction of the corporate resource plan and a second prompt which is based on dialogue with the user.

5. The development document generation apparatus according to claim 1, further comprising: a second recording unit that stores prompts used to create the development document created by the creation unit; a shared display unit configured to allow other users in the same team to view and execute a project handled by one user; and a reflection unit that reflects the prompts when creating a development document different from the development document.

6. The development document creation apparatus according to claim 1, wherein the question generation unit further generates a second question that complements the first question based on the first user response from the user.

7. The development document creation apparatus according to claim 6, wherein the response acquisition unit acquires a second user response to the second question from the user, and the creation unit creates the development document based on the development information, the first user response and the second user response.

8. The development document creation apparatus according to claim 4, wherein the first prompt includes a conversation history between the user and the AI ​​necessary for creating the development document.

9. The development document creation apparatus according to claim 7, wherein the response generation unit generates the response corresponding to the category of the task.

10. The development document creation apparatus according to claim 7, wherein the response generation unit generates the response according to the current date.

11. A development document creation method that supports the creation of development documents using a large-scale language model, the method comprising: causing a computer to execute: a development information acquisition step of acquiring development information relating to development to be described in the development document; a recording step of recording prompts of the large-scale language model as template information according to the development task; a task acquisition step of acquiring the development task from a user; a question generation step of generating a first question using the large-scale language model based on the template information corresponding to the acquired task; an answer acquisition step of acquiring a first user answer to the first question from a user; an answer generation step of generating an answer which is a solution based on the first user answer acquired in the answer acquisition step; and a creation step of creating a development document using the large-scale language model based on the development information and the first user answer to the first question acquired from the user.

12. A development document creation program that supports the creation of development documents using a large-scale language model, wherein the computer implements: a development information acquisition function that acquires development information related to the development to be described in the development document; a recording function that records prompts of the large-scale language model as template information according to the development task; a task acquisition function that acquires the development task from the user; a question generation function that generates a first question using the large-scale language model based on the template information corresponding to the acquired task; an answer acquisition function that acquires a first user answer to the first question from the user; an answer generation function that generates an answer which is a solution based on the first user answer acquired by the answer acquisition function; and a creation function that creates a development document using the large-scale language model based on the development information and the first user answer to the first question acquired from the user.

Images