AI-based automated patent specification writing system

Abstract

Efficiently draft higher quality patent specifications. An information processing device according to one embodiment performs a process based on source information relating to the invention. A decision to determine the division units when drafting a specification for a patent application for the invention means for generating a plurality of blocks by referring to the source information and providing the description of the invention for each division unit to the generating AI; an input means for inputting instructions for creating a sentence; and an output means for outputting the draft text based on a combination of the outputs.

Description

[Technical Field]

[0001] The present invention relates to an information processing device, an information processing method, and a program, and in particular to an invention Based on the source information, AI generates a draft specification for filing a patent application for the invention. This technology uses Generative Artificial Intelligence (Generative Artificial Intelligence). [Background technology]

[0002] In recent years, there has been progress in artificial intelligence (AI) technology, especially in generative AI, represented by large-scale language models (LLMs). The results have been remarkable, and it is expected to be used in a variety of fields, including writing, translation, and summarizing. In the field of practical application, this generative AI technology is being applied to the specification, which is part of patent application documents. The specification must clearly and in detail describe the subject matter of the invention. Since the creation of such a system requires a high level of expertise and a great deal of effort, generative AI is expected to be effective. There is a strong desire for this to be achieved.

[0003] Against this background, there have been many technologies to support the preparation of patent application documents. For example, Patent Document 1 proposes a document processing system equipped with conversational AI and / or generative AI. Using the creation device, the claims, the first draft of the specification, and the prior art documents are The input information is the difference between the invention described in the claims and the invention described in the prior art. This technology is mainly used for the We focus on supporting the creation of parts related to determining the novelty and inventive step of patents.

[0004] In addition, in Patent Document 2, the creator of the electronic file of the first specification states that Identify multiple related past patent applications and extract the second specifications from the electronic files. The present invention discloses a specification creation support device that extracts sentences containing multiple keywords selected by an author. This technology extracts highly relevant descriptions from past documents to improve the accuracy of specifications. The purpose is to reduce the effort required for creation. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Publication No. 2024-172808 [Patent Document 2] Japanese Patent Application Publication No. 2023-6493 Summary of the Invention [Problem to be solved by the invention]

[0006] These prior art techniques are useful in that they ease the burden on drafters in certain aspects of drafting specifications. However, when attempting to use generative AI to create a draft of the entire specification all at once, If this happens, the generated text may be redundant, the core of the invention may not be explained sufficiently, or It is difficult to consistently obtain high-quality specifications due to the lack of logical consistency. In particular, the specification was divided into the following sections: [Title of the Invention], [Technical Field], [Background Art], [Invention] Summary of the Invention] ([Problem to be solved by the invention], [Means for solving the problem], [Invention [Effects of the Invention], [Brief Description of Drawings], [Form for Carrying Out the Invention], etc. The content and level of detail required for each item are different, so it is not possible to provide a single guideline. It has been difficult to get AI to create a high-quality specification. By dividing it into parts and giving instructions to the generation AI for each part individually, more detailed and high-quality Although it is suggested that it is possible to obtain high-level descriptions, how to determine the division units and There was no established method for efficiently drafting specifications.

[0007] The present invention has been made in consideration of the above-mentioned problems, and is a method for generating a patent application using a generation AI. When drafting specifications, information that can help you draft higher quality specifications efficiently The present invention aims to provide a processing device, an information processing method, and a program. By dividing the content of the specification into appropriate units and giving instructions to the generation AI for each division, the entire specification can be generated. Maintaining the overall structure of the document, it encourages detailed and accurate descriptions of each part, resulting in improved quality of the specification. and aims to improve efficiency. [Means for solving the problem]

[0008] One aspect of the present disclosure is to file a patent application for an invention based on source information relating to the invention. A determination means for determining division units when preparing a draft specification for the patent application, and a generating AI for the patent application. inputting an instruction to create a description of the invention for each division unit by referring to the source information; and an input means for generating the above-mentioned output based on a combination of outputs obtained from the generating AI for each division unit. and an output unit for outputting a draft text.

[0009] The source information includes attribute information indicating attributes of the patent application, and the determining means The division unit may be determined according to attribute information.

[0010] The attribute information may include the technical field of the invention.

[0011] The attribute information may include the applicant of the patent application.

[0012] The attribute information may include the urgency of the patent application.

[0013] The determining means determines one option from among a plurality of defined options as the division unit. It may be determined.

[0014] The source information includes drawings relating to the invention, and the plurality of options are based on the drawings. It may also include an option to use it as a division unit.

[0015] The drawings relating to the invention include a plurality of draft drawings in the application documents of the invention, The option (2) may include an option of using each of the plurality of draft drawings as a division unit.

[0016] One of the plurality of draft drawings includes a plurality of components, and the plurality of options However, regarding the one draft drawing, the option of dividing each of the plurality of components into units is included. Good too.

[0017] The source information may include a transcript of a meeting between the inventor of the invention and a patent attorney. good.

[0018] This information processing device allows the inventor of the invention and the patent attorney to read the draft of the specification together. The company has a means to transcribe the meeting and revise the draft text using the generation AI. You may do so.

[0019] This information processing device generates a machine learning model using information about the division units and the quality of the draft text. the determining means uses the machine learning model to determine the classification. The division unit may be determined.

[0020] The source information includes a drawing list, and a plurality of templates are recorded for the drawing list. The input means includes a storage means for storing a template selected based on the source information. The source information to which the drawing list of the source is added is referred to, and the invention is An instruction to create a description may be entered.

[0021] A corresponding technical field is defined for each of the plurality of templates, and The means generates a template drawing list corresponding to the technical field identified based on the source information. and creating a description of the invention for each division unit by referring to the source information to which the item has been added. Alternatively, the user may input an instruction to cause the device to perform the above steps.

[0022] The input means includes a list of drawings of the template selected based on the source information; By referring to the source information, drawings that should be added to the drawing list are provided to the generation AI. Alternatively, the user may input instructions to prompt the user.

[0023] Another aspect of the present disclosure is to provide a method for obtaining a patent for an invention based on source information relating to the invention. A step of determining division units when preparing a draft specification for application, and a step of generating AI and instructing the source information to create a description of the invention for each division unit. and inputting the output from the generating AI for each of the division units. and outputting the draft text.

[0024] Yet another aspect of the present disclosure is to provide a method for causing a computer to perform the following steps based on source information related to the invention: A process for determining the division unit when drafting a specification for a patent application for the invention and referring to the source information, the description of the invention is generated for each division unit. a step of inputting an instruction to generate a statement, and and outputting the draft text based on a combination of the outputs. provide. [Effects of the Invention]

[0025] According to one aspect of the present invention, when creating a draft specification for a patent application using a generative AI, By dividing the content of the invention into appropriate units and giving instructions to the generating AI for each division, High-quality draft texts can be created efficiently. [Brief explanation of the drawings]

[0026] [Figure 1] FIG. 1 is a diagram showing an overview of a specification preparation support system 1000. [Figure 2] FIG. 2 is a diagram illustrating an example of the functional configuration of a specification preparation support application 1. [Figure 3] FIG. 2 is a diagram illustrating an example of the hardware configuration of the server 10. [Figure 4] FIG. 2 is a diagram illustrating an example of the hardware configuration of a user terminal 30. [Figure 5] 10 is a flowchart showing an outline of the operation of the specification preparation support application 1. [Figure 6] FIG. 10 is a diagram illustrating source information. [Figure 7] FIG. 10 is a diagram illustrating a drawing list. [Figure 8] 3 is a sequence chart illustrating the operation of the specification preparation support system 1000. [Figure 9] FIG. 1 is a diagram illustrating a hierarchical structure of specification components. [Figure 10] FIG. 10 is a diagram illustrating an example of a division pattern. [Figure 11] FIG. 10 is a diagram illustrating an example of a source information input screen. [Figure 12] FIG. 10 is a diagram illustrating an example of a division unit setting screen. [Figure 13]A diagram showing examples of application to other fields. DETAILED DESCRIPTION OF THE INVENTION

[0027] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In the following description, the same reference numerals denote the same components, and the description thereof will be omitted where appropriate.

[0028] 1. Configuration FIG. 1 is a diagram showing an outline of the overall configuration of a specification preparation support system 1000 according to this embodiment. The specification writing support system 1000 calculates the invention specification based on the source information of the invention. We support the efficient and high-quality drafting of patent application specifications for inventions. In particular, it utilizes artificial intelligence (AI) to create specifications. This reduces the burden on users in the creation process while enabling the creation of logical and detailed specification drafts. This system is characterized by the fact that it generates patents based on invention information and related materials entered by the user. Divide the detailed text into multiple appropriate "division units" and have the generative AI create sentences for each unit. The architecture is designed to generate the final draft specification by integrating the resulting output. are.

[0029] As shown in FIG. 1, the specification preparation support system 1000 includes, as its main components, a server. The system has a server 10, a generating AI 20, and a user terminal 30. These components are The network 90 is connected to each other so that they can communicate with each other. Internet, Intranet, LAN (Local Area Network), WAN (Wide Area Network) This network can be any type of communication network, whether wired or wireless, such as a wireless or wired or wired or wireless network. Through cooperation via the network90, each component plays its own role, and the entire system Achieve harmonious movement of the body.

[0030] The user terminal 30 is an interface for a user to use the specification preparation support system 1000. The user terminal 30 is a client device that provides a user interface. Personal computers (PCs), tablets, smartphones, and other devices with information processing and communication functions The user terminal 30 may be any device having a communication function. "Specification Creation Support Application" that provides a user interface and client-side processing functions "Pre-1" is installed or can be accessed via a web browser Through this specification writing support application 1, users can easily access source information related to inventions. Entering information, issuing instructions for processing, and checking and editing the generated specification draft. App 1 provides user-friendly operation and allows you to operate the device intuitively without having to go through complicated settings. It is desirable that the system be designed to be usable.

[0031] The server 10 performs most of the back-end processing in the specification preparation support system 1000. The server 10 is typically an information processing device that executes one or more processors, memory, and It consists of a computer system equipped with a The server 10 receives the source information transmitted from the user terminal 30 and analyzes it. The server 10 performs a particularly important function: Based on the source information, the "unit of division" for processing the specification by the generation AI 20 is determined. The division units may be, for example, chapters of the specification (background art, problems, solutions, effects, etc.), figures, etc. Depending on the content and characteristics of the invention, or user specifications, such as by surface or by component in the drawing Furthermore, the server 10 sets the generated AI 20 for each determined division unit. It is responsible for generating specific instructions (prompts) for the user and sending them to the Generation AI 20. In addition, the sentences of each division unit obtained from the generation AI 20 are combined to create a consistent overall sentence. The server 10 also performs the process of formatting the specification into a draft.

[0032] Generative AI20 uses natural language processing technology, especially large-scale language models (LLMs), to Generate natural-sounding sentences based on given prompts The generation AI 20 is a system having the function of generating the data for each division unit transmitted from the server 10. Based on the instructions and relevant source information (or a summary thereof), a description of the invention corresponding to each divisional unit is prepared. The AI ​​20 generates the document. The AI ​​20 exists outside the specification creation support system 1000. It may be provided as an independent AI service, in which case the server 10 It cooperates with the generated AI 20 via PI (Application Programming Interface) etc. Alternatively, some or all of the functions of the generation AI 20 may be incorporated into the server 10 . The generation AI 20 preferably has learned the unique writing style and logical structure of patent specifications. This is expected to result in the generation of higher quality descriptions.

[0033] These components, namely, the server 10, the generating AI 20, the user terminal 30, and the user The specification preparation support application 1 that functions on the terminal 30 cooperates with the As a result, the specification writing support system 1000 of this embodiment can generate a specification based on the source information of the invention. It provides a set of functions to efficiently generate high-quality draft specifications. Interact with the system through the specification creation support application 1 and set division units according to the characteristics of the invention. By participating in the adjustment of instructions to the generation AI 20, we can create specifications that are more in line with the intention. This can be achieved.

[0034] As shown in FIG. 2, the specification preparation support application 1 includes a storage unit 11 as a main functional configuration. , receiving means 12, determining means 13, input means 14, acquiring means 15, combining means 16, adjusting The device has a control means 19, an output means 17, and a control means 19. These means cooperate to: It implements a series of processes to efficiently generate high-quality draft specifications from source information.

[0035] The storage means 11 stores various information necessary for the operation of the specification preparation support application 1. The information stored in the stage 11 includes, for example, "source information" input by the user and the generated A This includes the "partial specification" generated by I20 for each division unit. As mentioned above, the specification, including invention explanation materials, meeting records, prior art documents, and drawing data, The partial specification is a set of information that is the basis for creating the specification. The storage means 1 stores the individual explanatory text fragments generated by the generation AI 20 for each division unit. 1 may be, for example, a RAM (Random Access Memory) of the user terminal 30 or the server 10, Hard disk drives (HDDs), solid-state drives (SSDs), and other storage devices This is achieved by:

[0036] The receiving means 12 is an interface device that receives various inputs and instructions from the user. Specifically, the user can use the GUI (Graphical User Interface) of the specification preparation support application 1. Input source information via the interface, instructions to start creating a statement, and settings related to division units The receiving means 12 handles the setting and selection of the specification, and instructions for correcting the generated draft specification. , converting the input information into an appropriate format and passing it on to the control means 19 and other related functional means. For example, if a user uploads a file of invention documents or a drawing file, The receiving means 12 receives these files and stores them in the storage means 11 as source information. do.

[0037] The determining means 13 is one of the characteristic functions of this embodiment, and determines the received source information. Based on the information, the "division unit" for generating the draft specification by the generation AI 20 is determined. The decision on the division unit is extremely important as it significantly affects the overall quality of the specification and the efficiency of its production. The determining means 13 determines the content of the source information (e.g., the technical field of the invention, the complexity of the invention, the number of drawings, etc.) number and type of patent applications), patent application attribute information (e.g., type of applicant, urgency of application, target country, etc.), ), or based on user-defined splitting policies and templates, For example, if the invention is in the field of machinery and has many components, the drawings It is possible to set division units for each element, or even for each major component in the drawing. On the other hand, in software-related inventions where the process flow is the main focus, the main It may be appropriate to divide it into steps or modules. Select the most suitable split pattern for the current project from the multiple split patterns defined. We also use a machine learning model to estimate the optimal division strategy based on past success stories. It can be used.

[0038] The input means 14 inputs the division unit determined by the determination means 13. Generate prompts for the AI ​​20 to create a description of the invention corresponding to the The input means 14 has a function of inputting the instruction to the generation AI 20. The main points of content to be included, relevant parts of source information to be referenced, expected style and level of detail, The input means 14 can include these instructions. The generation AI 20 then formats the data into a format that can be interpreted (for example, a specific API format) and sends it. The quality of this instruction determines the quality of the partial specification obtained from the generated AI 20. The input means 14 is designed to generate instructions that are as specific and clear as possible.

[0039] The acquisition means 15 acquires each division data generated by the generation AI 20 based on an instruction from the input means 14. The explanatory text (partial specification) corresponding to the unit is obtained from the generated AI20. The response is usually returned as text data. The partial details are sequentially received and temporarily stored in the storage means 11. Once the document has been acquired, the process moves on to the next step, the combining process.

[0040] The combining means 16 combines the plurality of parts acquired by the acquiring means 15 and stored in the storage means 11. The sub-specifications are combined in a prescribed order to form the outline of a single, continuous draft specification. This joining order is usually consistent with the standard specification structure (e.g., background art, problem, solution, effect). The combining means 13 follows the logical order of the division units (e.g., the result, the embodiment, etc.) and the logical order of the division units determined by the combining means 13. 16 should not only include simple text concatenation, but also appropriate headings and divisions at the beginning and end of each sub-specification. It is also possible to add a cut.

[0041] The adjusting means 17 further adjusts the draft specification combined by the combining means 16. The adjustment process includes, for example, checking and correcting consistency of style and terminology. Correct, optimize conjunctions between paragraphs and sections, detect logical inconsistencies and redundant expressions, and provide correction instructions , improving overall readability, etc. The adjustment means 17 automatically performs these adjustments. Rule-based processing, rewriting by giving instructions to the regeneration AI 20, or A semi-automatic process can be implemented that presents correction suggestions to the user and prompts them to make a selection.

[0042] The output means 18 is connected by the connecting means 16 and adjusted by the adjusting means 17 as necessary. The final draft specification after the adjustment is output to the user. The display on the display of the user terminal 30, the editable document file (for example, Word format or It can be saved in various formats according to the user's needs, such as text format, or printed. The user checks the output draft specification and performs final review and correction.

[0043] The control unit 19 controls the overall operation of the specification preparation support application 1. 9, in response to a user instruction or data input from the receiving means 12, a determining means 13, an input means Each functional means such as a stage 14, an acquisition means 15, a combination means 16, an adjustment means 17, and an output means 18 In addition, the server 10 and the generation AI In this way, the control means 19 controls the specification preparation support application 1 to communicate with the system. It plays an orchestration role to ensure the smooth functioning of the system.

[0044] As shown in FIG. 3, the server 10 includes the basic hardware components of a computer. Specifically, the device includes a CPU (Central Processing Unit) 101, a memory 10 2, storage 103, and communication device 104, and these components are connected via a bus 107. This type of configuration is similar to that of a typical server computer or workstation. This is widely used in stations, but to achieve the object of this embodiment, Therefore, the performance and capacity of each component are selected appropriately.

[0045] The CPU 101 is a central processing unit that controls the overall operation of the server 10. The CPU 101 is responsible for various information processing in the Operating system (OS) and various programs (e.g., For example, the server side component of the specification preparation support application 1, the determining means and the input means described later, a module for realizing functions such as a stage, a coupling means, and an adjusting means into the memory 102; By executing these, each function of the server 10 is realized. For example, the CPU 101 The process analyzes the source information received from the user terminal 30 and determines the division unit. A process of generating instructions (prompts) to the generation AI 20 based on the division units that have been determined; Receives responses (partial specifications) from 20 and performs processes such as combining and adjusting. 01 is a single-core processor, but it can also be a multi-core processor with multiple cores. It is also possible to have a configuration in which multiple CPUs are installed.

[0046] The memory 102 temporarily stores the programs executed by the CPU 101 and data to be processed. The memory 102 is a main storage device for storing data. When the CPU 101 executes a program, it uses volatile memory such as The necessary program code and data are read from the storage 103 into the memory 102, and In this embodiment, the source information during processing, the determined Information on division units, instructions to the generation AI 20, partial specifications received from the generation AI 20, combination The draft specification data and other data are stored in the memory 102 as needed, and are processed by the CPU 101. The capacity of the memory 102 is determined taking into consideration the amount of data to be processed and the responsiveness of the system. It is desirable that sufficient funds be secured.

[0047] The storage 103 is an auxiliary storage device for long-term storage of programs and data. The storage 103 may include, for example, a hard disk drive (HDD), a solid state drive (SDD), Storage 1 03 includes the operating system of the server 10 and the specification preparation support system 1000. A set of application programs to realize this, persistence of source information received from users storage area for generated draft specifications, and rules for determining division units. In particular, it stores a large amount of source information and past processing history. When handling specification data, the storage 103 must have sufficient storage capacity and high-speed read / write performance. It is required to do so.

[0048] The communication device 104 is a device that allows the server 10 to communicate with other devices (e.g., users) via the network 90. This is an interface for communicating with the terminal 30 and the generating AI 20. is a network interface card such as an Ethernet card. It is configured by a NIC (Network Interface Card) and uses TCP / IP (Transmission Control Protocol / Interface) The communication device 10 transmits and receives data according to a predetermined communication protocol such as the IEEE 802.11a Protocol. The server 10 receives source information and processing requests from the user terminal 30 via the The draft specification is transmitted to the user terminal 30. If provided as a software program, instructions may be sent to and from the generation AI 20 via the communication device 104. The results are received.

[0049] The bus 107 includes a CPU 101, a memory 102, a storage 103, and a communication device 104. A common interface is provided to connect the various components in the server 10 and transmit data and control signals. The components cooperate with each other via the bus 107, and the entire server 10 For example, the CPU 101 communicates with the memory 102 and storage device 104 via the bus 107. The communication device 104 reads commands and data from the memory 103 and writes the processing results. The received data is transferred to the memory 102 and the CPU 101 via the bus 107, and the data to be transmitted is Data is similarly sent to communication device 104 via bus 107 .

[0050] The server 10 having the above-described hardware configuration is configured such that the CPU 101 executes the By executing the program loaded in the Various functions for this purpose, namely, analyzing source information, determining division units, and generating instructions for the generation AI This allows for efficient execution of a series of processes such as retrieving, combining, and adjusting partial specifications. Such a configuration of the server 10 is merely an example, and may vary depending on the scale of the system and the required performance. Optionally, a more decentralized or cloud-based architecture can be adopted. is.

[0051] As shown in FIG. 4, the user terminal 30 may be a general personal computer or a high-performance It has the standard hardware components that a mobile device has. Specifically, it has a CPU ( Central Processing Unit 301, memory 302, storage 303, communication device 304 , input device 305, and output device 306, and these components are connected via bus 307 This configuration allows data and control signals to be exchanged between the The user terminal 30 executes the client-side functions of the specification preparation support application 1, This makes it possible to smoothly provide an interface and communicate with the server 10.

[0052] The CPU 301 is a processing unit that controls the overall operation of the user terminal 30 . The CPU 301 executes an operating system (OS) stored in the storage 303. and the client application for the specification preparation support application 1, as well as various other applications. The application software is read into the memory 302 and executed. Drawing faces, handling user input, sending and receiving data, processing data locally, etc. Various processes required are executed on the user terminal 30. For example, when a user writes a specification, When the creation support application 1 is started, the CPU 301 executes the program and inputs source information. The output device 306 displays the input screen, the processing status display screen, etc., and the user inputs the input data from the input device 305. Accepts user operations.

[0053] The memory 302 stores programs and data used by the CPU 301 when it executes processing. RAM (Random Access Memory) is a main memory device that stores data temporarily and allows high-speed reading and writing. When the specification preparation support application 1 is executed, the program code Part of the source information entered by the user, temporary data, and statements received from the server 10 The draft data and the like are expanded on the memory 302. The CPU 301 executes the data on the memory 302. The capacity and speed of the memory 302 depend on the response of the user terminal 30. This contributes to ease of use and a comfortable operating feel.

[0054] The storage 303 stores an operating system, various application programs, This is an auxiliary storage device that permanently stores user data, etc. Storage 303 includes: HDD (Hard Disk Drive), SSD (Solid State Drive), or eMMC (embedded M In the user terminal 30, a nonvolatile memory such as a USB flash drive or a USB flash memory card is used. The program itself, setting files, and files created or downloaded by the user The downloaded source information files (invention explanation materials, drawing files, etc.) received from the server 10 The saved draft specification file and the like are stored in storage 303.

[0055] The communication device 304 is a device for allowing the user terminal 30 to communicate with the server 10 via the network 90. In some cases, it communicates directly with the generating AI 20 (e.g., a cloud-based AI service). The communication device 304 is an interface for connecting to a wired LAN adapter or a wireless LAN ( Wi-Fi (registered trademark) adapter or mobile communication module (e.g., 5G, LTE) The user terminal 30 uses the communication device 304 to transmit source information. Upload to the server 10, or download the processing results (draft specification, etc.) from the server 10. Or download it.

[0056] The input device 305 is used by the user to input information and instructions to the user terminal 30. Typical input devices 305 include a keyboard, a mouse, a touch panel, The user operates these input devices 305 to , input source information, edit text, and perform various operations on the specification creation support app1. Select an operation command. If voice input is supported, a microphone is also used as an input device. 305, and is equipped with a scanner and camera for digitizing drawings and handwritten notes. In some cases, this may be the case.

[0057] The output device 306 is a device that allows the user terminal 30 to present processing results and information to the user. The most common output device 306 is a liquid crystal display (LCD) or an organic electroluminescent (EL) display. It is a display device such as a display screen, and can display the GUI of the specification preparation support application 1 and the specification draft being prepared. Sentences, various messages, etc. are displayed on the screen. Voice notifications and reading functions are also available. If provided, the output device 306 also includes a speaker, and if necessary, a printer may be connected. It is also possible to output to paper media.

[0058] The bus 307 is connected to a CPU 301, a memory 302, a storage 303, a communication device 304, an input / output device 306, and a data transfer device 308. Each hardware component in the user terminal 30, such as the input device 305 and the output device 306 It is a common electrical path that interconnects devices and transmits data, addresses, and control signals. These components cooperate via a bus 307 to function as the user terminal 30. Fulfill this.

[0059] The specification preparation support application 1 is executed on the user terminal 30 having the above-described hardware configuration. By operating the invention, the user can efficiently input information about the invention and access the server 10 and the The specification creation process, which is carried out in conjunction with AI20, can be intuitively operated and managed, and the results can be displayed The specific model and specifications of the user terminal 30 can be Although this system may be used in a variety of ways depending on the situation and user requirements, the system described in this embodiment may be used in a variety of ways. The basic requirement is that the system has sufficient processing power, input / output functions, and communication functions to realize the functions required.

[0060] 2. Operation Next, the specification preparation support system 1000 according to this embodiment, particularly the specification preparation support application 1 (including when its main functions are executed on the server 10) is a draft specification for a patent application. The overall process flow for creating the Figure 5 shows the process from the user's instruction to start the process to the output of the final draft specification. This process flow is shown in a schematic diagram of a series of steps from accepting user input to , and cooperates with the generation AI 20 as necessary to automatically or semi-automatically carry out each stage of specification creation. The purpose is to have them carry out the

[0061] First, the user starts the specification preparation support application 1 on the user terminal 30 and Once you start entering source information for the subject invention, the system will begin the process of retrieving the source information. As mentioned above, source information is information necessary for understanding the invention and writing the specification. It contains all the necessary information, including the inventor's description, technical notes, draft drawings, relevant literature, and This includes records of meetings between inventors and patent attorneys (transcription data and audio data), etc. The acceptance means 12 of the specification preparation support application 1 displays this information on the user interface. The image data is received via the interface and stored in the storage means 11. OCR processing to extract text from audio data, and speech recognition processing to generate text information from audio data. Any of these may be performed.

[0062] FIG. 6 is a diagram illustrating an example of source information. As shown in FIG. 6, source information 50 includes Separately, invention explanation document 51 (invention proposal), meeting audio 52 (transcription), prior art These may include documents 53 and reference specifications 54. These are merely examples, and Any other information that is useful for understanding the content of the invention and preparing an appropriate specification may be used. Any information can be used as source information. Accepted as input from the user or retrieved from an external database for further processing, This is referenced when determining the division unit and generating instructions (prompts) to the generation AI 20.

[0063] The invention description document 51 (invention proposal) is prepared by the inventor himself or a team including the inventor. It is a document that describes the most basic details of the invention. The problem that the invention aims to solve, the means to solve the problem (the structure of the invention), the effect of the invention As a result, it includes specific modes (embodiments) for carrying out the invention, necessary drawings, etc. The invention description materials are extremely important as primary information for understanding the essence of the invention. It is the main reference source for creating the outline of the book. The format varies depending on the company or research institution. There are a variety of formats, from standardized documents to more free-form notes and presentation materials. The specification preparation support application 1 can handle these various document file formats (e.g. For example, DOCX, PDF, PPTX, etc.) and analyze its contents.

[0064] The audio of the meeting 52 (transcription) is, for example, a conversation between an inventor and a patent attorney about the contents of an invention. It is a record of the discussion, question and answer session, or brainstorming session that took place during the meeting. The "origin" refers to the content, background, technical details, inventor's intention, and rights of the invention that is the subject of a patent application. Oral consultations regarding the patent application strategy or feedback on the draft specification This refers to the content of communication recorded as text data. For example, communication may take place between an inventor and a patent attorney (or a person in charge of filing an application, etc.). Hearings on inventions, technical review meetings, readings of draft specifications, and other meetings, discussions, This includes questions and answers, presentations, etc. This "transcription" is typically The audio recordings of these communications were processed using speech recognition to generate However, it also includes manual transcriptions and manually corrected speech recognition results.

[0065] Meeting Audio 52 provides the background and nuances of the invention that cannot be understood from the invention description alone. It contains a wealth of information on the inventor's intentions, technical points, and even hints on patent strategies. In particular, oral explanations and discussions may contain important information that is important for grasping the essence of an invention. They often contain keywords and logical developments, and these are converted into text and added to the source information. This allows the generation AI 20 to understand the invention more deeply and create a detailed specification. The system uses the voice recording itself as input and processes it internally to convert it into text. It may be configured to generate a transcribed text, or a pre-transcribed text file. This transcription data may be used to input information that reflects the "real voice" of the invention. It serves as information to complement the invention description materials.

[0066] Prior art document 53 evaluates the novelty and inventive step of the present invention and describes the background art section. This refers to existing patent publications, academic papers, technical reports, etc. that are used as references when developing an invention. This includes related literature discovered through preliminary research and literature cited during the patent office examination process. These references (bibliographical information, full text, drawings, etc.) are used as source information. By entering information into the system, the positioning of the invention can be clarified and the differences from the prior art can be effectively identified. In particular, the present invention can provide the generation AI 20 with information to describe the problems of the prior art. In clarifying the issue of how to overcome this, it is necessary to analyze the prior art documents appropriately. The system extracts relevant parts from these documents and prompts them. It can be used to generate

[0067] Reference specification 54 is a specification of a past patent application in a technical field similar to that of the present invention, or a specification of the same The specifications of the applicant's previous applications, etc., serve as reference for the expression, structure, and granularity of the description of the specification It is a document, especially when it is used to describe preferred phrases in a particular technical field or to provide the examiner with information about the invention. These reference specifications are useful for learning the style of expression that can be used to accurately convey the By providing the generated AI 20 as source information, the generated AI 20 may be used in accordance with the practice in the art or in the application. It increases the likelihood of generating high-quality sentences that match the human intent. It is also used as a reference for specific expressions, such as how to describe names and functions.

[0068] This diverse source information is managed in an integrated manner by the specification preparation support application1, and each Depending on the characteristics of the information source, it is used appropriately at each stage of the specification drafting process. The outline and technical features of the invention are extracted from the invention explanation materials and meeting records, and the background technology and problem definition are Prior art documents are used as references, and the overall structure and style of expression are based on similar patterns from the reference specifications. The system simply stores this source information. rather than semantic analysis (e.g., entity extraction, relation extraction, topic modeling, etc.) ) to evaluate the importance and relevance of each piece of information, determine the division unit, and improve the accuracy of prompt generation. The more abundant and high-quality source information is provided, the more useful this system will be. This allows for more accurate segmentation and more effective prompt generation by the AI ​​20. This will result in a significant improvement in the quality of the final output specification draft.

[0069] Referring again to Figure 5, the system then generates the claimed A draft of the scope (claim) is created (step S12). The basic format is to manually create or edit using the support functions of the detailed writing support app 1. Analyzes the content of source information, extracts the key points of the invention, and automatically generates a claim template. This automatic generation function may be realized by utilizing the generation AI 20. The draft claims will serve as an important guide for the subsequent drafting of the specification.

[0070] The system then provides information about the drawings to aid in understanding the invention and to accompany the specification. Specifically, first, a drawing list is created (step S13). A list of the titles and brief descriptions of the drawings that are considered necessary to explain the present invention. This drawing list can be created manually by the user or by adding drawings and inventions from the source information. Based on the main components of the document, the system proposes a format for the user to approve or modify. It is possible.

[0071] As shown in Figure 7, the drawing list is typically in a tabular or similar data structure. Each entry has at least one entry for each drawing. It includes information such as "drawing number," "title," and "elements shown." It is a number that uniquely identifies each drawing, and is consecutive, for example, "Drawing 1" and "Drawing 2." The drawings are numbered as shown in Figure 1, which will be used when referencing the drawings in the text of the specification.

[0072] "Title" is a title that briefly explains what each drawing shows. For example, In the example of Figure 7, the title of "Figure 1" is "System Configuration Diagram" and the title of "Figure 2" is "Machine The title of the drawing is written to indicate the subject or type of the drawing, such as "Noh Composition Diagram." The information in the drawing list not only helps the user to get an overview of the entire drawing list and understand the role of each drawing. The system determines the division unit and generates the description of each drawing for the generation AI 20. This is an important clue when creating prompts to help users. The title "System Configuration Diagram" suggests that the diagram shows the relationships between the components of the entire system. It can determine that the user is experiencing a problem and prompt the user to generate an explanation accordingly.

[0073] "Illustrated elements" refers to major components or reference numerals specifically depicted in each drawing, or or a list of the main concepts or processing steps that the drawing is trying to explain. In the example of FIG. 7, the information processing system 1000, the server 10, the generation A The list includes "IP 20, user terminal 30, specification creation support application 1, and network 90." This means that Figure 1 shows the overall system configuration including these elements. Similarly, with regard to "FIG. 2," "storage means 11, control means 12, input means 13, output means 14” is a hardware component of the server 10 in FIG. , memory 102, storage 103, and communication IF 104 are the user terminals in FIG. The last 30 hardware components are CPU 301, memory 302, storage 303 , a communication IF 304, an input device 305, and an output device 306 are shown as illustrated elements. Furthermore, regarding the flowchart in Figure 5, the main step "Sort" Obtaining the patent information (S101), creating a claim plan (S102), creating a drawing list (S10 3)" and examples of source information in "Figure 6" include "invention explanation materials, transcriptions, etc. The "Prior Art Documents" are listed as the primary source of information. When you set the division unit to "per drawing" or "per main element in the drawing", It is essential to specifically identify the content to be explained in each division unit. When generating instructions, including these element names as keywords will help you to create more accurate instructions. We can expect to generate clear text.

[0074] The creation of the drawing list (step S13) can be performed in several ways. The user can enter the drawing number, title, etc. of each drawing through the interface of the specification creation support application 1. In this case, the system Second, the system can use source information (especially The system analyzes the drawings (including the drawings included in the invention description and existing drawing files) and automatically creates a list of drawings. Another method is to generate candidate solutions and present them to the user for correction or approval. Figure captions and reference numbers in the source information or references to drawings in the invention description Using the location and other clues, the drawing number, title, and illustrated elements are estimated. (created in step S12) and explain the components. It is also effective to make a list of the drawings required to

[0075] The drawing list thus created is stored in the storage means 11 and is referenced in subsequent processing. In particular, the drawing list plays an important role in step S15, which determines the division unit. For example, the determining means 13 may determine each drawing listed in the drawing list as one division unit. Alternatively, you can use a strategy of dividing each drawing into parts. It is also possible to set the division unit at a finer granularity based on the list of For example, not only the unit "Fig. 1 System Configuration Diagram" but also "Explanation of Server 10 in Fig. 1" , "Description of generated AI20 in Figure 1" It is also possible to generate division units, which allow detailed and comprehensive descriptions of each element. It is possible to prompt the generation AI 20. In addition, the creation of the partial specification in step S17 Sometimes, the input means 14 refers to the drawing list and inputs the drawing information for each division unit (e.g., a specific drawing). A prompt including the corresponding title and information of the illustrated element is generated and sent to the generation AI 20. This allows the generation AI 20 to clearly determine which part of which drawing an explanation should be generated for. It is expected that the system will be able to accurately understand and generate accurate sentences.

[0076] Referring again to Figure 5, once the drawing list is finalized, the actual drawings are then drawn based on the list. The user creates or prepares a surface (step S14). This is generally done using Computer-Aided Design (CADA) software, but The system has limited functionality to automatically generate simple block diagrams and flowcharts. Alternatively, existing drawing files can be imported into the system and managed.

[0077] Once the above preparatory steps (S11 to S14) are complete, it is time to draft the specification. The process starts. First, the determining means 13 of the specification preparation support application 1 determines the source information, the preparation The AI ​​20 then processes the specification by taking into consideration the proposed claims, drawing list, etc. The division unit is determined based on, for example, the "Background Art" "," "Problems that the invention aims to solve," "Means for solving the problems," and "Effects of the invention" The basic components of the specification, the explanation of the embodiments corresponding to each claim, and each The determining means 13 is to determine the technical field of the invention, the application strategy, etc. The division unit is determined with the optimal granularity and order according to the user's preferences. This is done based on pre-defined templates and rules, or machine learning models are used to It is dynamic and based on knowledge learned from previous cases.

[0078] Once the division unit is determined, the system selects the next division unit to be processed from the list of the determined division units. The target portion to be processed is identified (step S16). The input means 14 of the specification creation support application 1 inputs the relevant part of the specification ( Generate and send prompts to create a partial specification. generates a description of the target part based on this instruction and related source information, and sends it to the system. The acquisition means 15 acquires this response (partial specification) and stores it in the storage means 11. (Step S17).

[0079] When the process of step S17 is completed, the system performs the following for all the division units that have been determined. It is determined whether the preparation of the partial specification has been completed (step S18). If there are any positions remaining (NO in step S18), the process returns to step S16, and the next target position is In this way, all the divided units are identified and the same processing (S16, S17) is repeated. Partial specifications will be prepared for each position in turn.

[0080] When it is determined that the preparation of partial specifications for all divisional units has been completed (step S18 If the answer is YES, the combining means 16 of the specification preparation support application 1 is stored in the storage means 11. Multiple partial specifications may be combined in a predetermined logical order to form a single continuous draft specification ( In this combining process, the text is simply combined into a combined specification (step S19). In addition to concatenating the data, headings and formatting are also added as necessary.

[0081] Next, the adjustment means 17 performs the final adjustment process on the combined draft specification (step This adjustment process includes, for example, checking the overall logical structure, ensuring consistency of terminology, etc. These include ensuring accuracy, refining expressions, correcting typos, etc. The adjustment means 17 automatically performs these processes. This can be done dynamically, or the user can be prompted with suggested corrections and make adjustments interactively. It is also effective to use the generated AI20 again to rewrite and proofread the entire combined specification. is.

[0082] The draft specification that has been finalized is output to the user by the output means 18 (the Step S21). The output is displayed on the screen of the user terminal 30, and is saved in an editable file format (e.g. This can be done in a variety of formats, such as saving in a file format (e.g. DOCX, TXT, etc.) or printing. The user reviews the output draft specification and manually finalizes it if necessary. By going through the above-mentioned series of processing flows, the specification preparation support system according to this embodiment The system 1000 supports the efficient drafting of high-quality specifications.

[0083] The overall flowchart in Figure 5, the specific example of source information in Figure 6, and The creation of the drawing list in 7 is performed when the specification drafting support system 1000 generates a draft specification. This section shows the preparation stage and basic processing flow of the sequence chart in Figure 8. By referring to the specification creation support application 1 (or the server 10 that performs the actual processing), More specific details about the information exchange and processing collaboration with AI20 Figure 8 shows the process from the user inputting source information to the final statement output. ,Main processing steps and interactions between the specification creation support app 1 and the generation AI 20 The sections are shown in chronological order.

[0084] First, the user sends a request for information about the invention to the specification preparation support application 1 via the user terminal 30. In response to this, the specification creation support application 1 This step is the same as step S11 (source) in FIG. The received source information is stored in the storage means 11 and is used for subsequent The source information is used in the processing of the invention explanation materials, meetings, etc., as shown in Figure 6. This includes transcriptions, prior art documents, reference specifications, etc.

[0085] Next, the specification creation support application 1 determines whether specific information can be described using only the received source information. is insufficient to fully understand the invention, or that additional information is necessary to further understand the invention. If so, the generation AI 20 can be requested to generate additional source information (step Additional source information is source information that is not prepared by the user. ,information used as source information in later processing, e.g., proposed claims, drawing list The generation AI 20 generates the required information based on this request. The requested additional source information is generated (step S103), and this is input to the specification creation support application 1. The specification preparation support application 1 receives this additional source information and responds to the existing source information. This step (S102, S103) is optional and is used in conjunction with the initial source. It may be omitted if the information is sufficient.

[0086] Next, the determination means 13 of the specification preparation support application 1 determines whether or not the collected and organized source information is sufficient. At least some of the information (including the initial source information and any additional source information generated as needed) (Step S10) 4). This process corresponds to step S15 in Figure 5. The division unit is the logical structure of the specification. (e.g., background art, problems, solutions, effects, each embodiment, explanations of each drawing, etc.) The information on the determined division unit is referenced in subsequent processing.

[0087] Before explaining how to determine the division unit, we will explain the hierarchical structure of the components of the specification. As shown in Figure 1, the components of a specification can be captured at multiple hierarchical levels. The first level (L1) corresponds to the full text of the patent application. It includes all documents necessary for filing an application, such as the patent application, specification, claims, drawings, and abstract. The main focus of this embodiment is to support the creation of specifications, but the system as a whole In this case, it is possible to consider overall progress management and consistency checks at this first level. is.

[0088] In the next second layer (L2), the "full text" of the first layer is divided into more specific document units. For example, "application", "specification", "claims", and "drawings" fall under this hierarchy. In this embodiment, the focus is on the creation of the "specification." This division allows for the application of different processing flows and templates depending on the characteristics of each document. For example, in the process of writing a specification and the process of writing claims, The instructions to the generation AI 20 may be different.

[0089] Furthermore, at the third level (L3), the "specification" is divided according to its standard components. In the example of Figure 9, there are sections such as "Introduction" (introductory part), "Configuration", "Operation", and "Modification". The "intro" usually includes the title of the invention, Technical field], [Background technology], [Problem to be solved by the invention], [Method for solving the problem] "Configuration" and "operation" are typically used to describe the process of implementing an invention. In the [Mode for carrying out the invention], the specific structure of the invention and how it functions are explained. "Modifications" refer to variations and applications other than the basic embodiments. This division at the third level is intended to organize the logical flow of the entire specification and to This is extremely effective in clarifying what should be done. can adopt these standard items as basic division units.

[0090] The fourth level (L4) is a level where each block in the third level is further subdivided. For example, ,The "introduction" section of the third level is as shown in ,Figure 9. It consists of [Name of Invention], [Technical Field], The invention may be divided into more specific sections of the specification, such as "Background Art." The "operation" part is based on the drawing list (see Figure 7) and is used to select individual "drawings" (e.g., "Drawings"). 1, 2, ..., 5). What does it show? What are the components depicted in the drawing? How are they related? Each drawing is treated as an independent division unit. The ratio is particularly high in technical fields that make extensive use of drawings (e.g., machinery, electricity, physics, etc.) This is suitable for prompting the generation AI 20 to provide a detailed and comprehensive description of each drawing.

[0091] The most detailed layer is the fifth layer (L5). For example, if a division unit called "Figure 1" is set at the fourth level, In this case, the fifth layer is composed of the individual components shown in FIG. 1 (e.g., "storage means," "input The elements shown in the functional configuration diagram of Figure 2, such as "means" and "output means", and the hardware shown in Figures 3 and 4, The description of each element (such as the elements shown in the hardware configuration diagram) is treated as an independent division unit. Alternatively, in the description of a specific drawing, it is possible to include a description such as "Outline of the drawing" or " List of main components and explanation of each component's function, and It is also possible to divide it into more detailed points, such as "explanation of the relationship" and "description of the relationship." By setting the division unit at a granular level, the Generative AI20 focuses on a very narrow range of topics. This increases the likelihood of obtaining more detailed, accurate, and in-depth descriptions. circle.

[0092] The determination means 13 of the specification preparation support application 1 (see FIG. 2) determines whether or not the first layer (L1) of the specification preparation support application 1 is a The division unit is determined by taking into consideration the hierarchical structure from the first layer to the fifth layer (L5). Through the interface of the Creation Support App 1, you can select at what hierarchical level or what You can specify how to set the division unit by combination. For example, if the item is highly urgent, So, the approach is to quickly create a big picture with a rough division at the third level, and then fill in the details later. On the other hand, for very complex inventions or cases where an exhaustive detailed description is required, In this case, we divide it into smaller parts at the fourth or fifth level and carefully write sentences for each part. In addition, the determining means 13 determines the content of the source information (see FIG. 6) and ,The attribute information of the patent application (technical field, applicant's intention, etc.) is analyzed, and within this hierarchical structure, For example, the system may have a function to automatically propose an optimal splitting strategy based on the source information. If there are a large number of drawings or detailed sectioning of the invention description, In this way, the division of the specification into more detailed levels is recommended. The elements are hierarchically analyzed, and division units are defined at an appropriate granularity. Providing accurate instructions to the applicant 20 will contribute to the efficient drafting of high-quality specifications in this embodiment. This is the key to success.

[0093] FIG. 10 is a diagram illustrating an example of a division pattern. As shown in FIG. 10, each division pattern is as follows: Each pattern is identified by a unique "pattern ID." The main components of the book are the "introduction," "structure," "action," and "variations." 9. For each block (block corresponding to the third layer L3 in FIG. 9), which layer level (L1 to L3 in FIG. 9) L5) or what combination of divisions is defined. There are.

[0094] For example, pattern ID "001" in Figure 10 shows an example of the simplest division pattern. In this pattern, the "introduction", "composition" and "variations" are particularly hierarchical. No level is specified, which means that for example the user can create these parts manually or Or it could mean that the system defaults to relatively large units. The "operation" part is written as "L1". This is the first layer in Figure 9, that is, It suggests capturing information about "actions" at the "full text" level. For example, turn "001" does not use the generation AI for the "action" part, but only for the other parts. It is intended for special cases such as those targeting a very general level of instruction. A more concrete interpretation is that the other items are manual or coarse divisions. On the other hand, for some reason, the generation of "actions" requires a strong awareness of the overall context. Alternatively, the notation "L1" may be used to refer to a specific specification item as a whole. This may be a placeholder meaning that the data will be processed collectively by the generation AI.

[0095] Next, if we look at pattern ID "002", we can see that it has "Intro", "Structure", "Movement", All of the "Modifications" are designated as "L3." This is the third layer in FIG. The main logical blocks of the specification (the entire "introduction", the entire "structure", the entire "operation", the entire "transformation" This pattern means that each of the "whole" examples is treated as a separate unit. It is suitable for those who want to draft the entire specification in a standard and quick manner. For the sentence, the generation AI 20 will generate sentences that cover a relatively wide range of content. This level of division makes it easy for users to grasp the overall structure with minimal operations, and It may be good for checking the view and direction.

[0096] Pattern ID "003" is an example of a more detailed division. " is at the "L3" level just like pattern "002", but "Configuration" and "Operation" The part is designated as "L4". "L4" corresponds to the fourth layer in Figure 9. For example, divide "structure" and "operation" into individual "drawing" explanation units based on the drawing list. In other words, in this pattern, the whole "intro" and the whole "variation" are Each is treated as a single unit, but the "composition" is explained in Figure 1, Figure 2, etc. Similarly, the "actions" are divided into units such as the action in Figure 1, the action in Figure 2, etc. Then, the generation AI 20 creates a partial specification for each of the drawings. You can expect to get a more focused and detailed explanation of the invention, especially for inventions with many drawings, This is an effective division pattern when the contents shown in each drawing can be clearly distinguished.

[0097] Furthermore, pattern ID "004" is an example of a pattern that performs the most detailed division. "Introduction" and "Variations" are "L3", "Configuration" is "L4", but "Operation" is In this case, it is written as "L4 / L5". This notation "L4 / L5" is used for each drawing ( L4), and further divide it into major components within the drawing (L5) as needed. For example, within the unit (L4) of operation explanation in Figure 1, The prompt is generated by dividing the action into units (L5) such as the action of a specific element A and the action of element B. , and have the generation AI 20 create sentences for each of them. It is possible to obtain a very detailed and comprehensive description of the division of roles among the components. Such detailed divisions allow for the identification of the "structure" and "application" which are important for clarifying the core of the invention and the scope of the patent. This is particularly effective in describing "actions," allowing the generation AI20 to generate sentences with more specific and in-depth content. It helps to achieve this.

[0098] The determination means 13 of the specification preparation support application 1 determines whether or not the division pattern ( The system stores multiple items of information (such as the information stored in the table in Figure 10), and users can select the characteristics of the project. (e.g., technical field, complexity of the invention, urgency of filing, required quality level of specification, etc.) Depending on the situation, the most suitable one can be selected from these patterns. The determining means 13 determines the attribute information of the patent application (for example, if the technical field is "machine", the pattern "003" is recommended, and if it is "Software", the "Operation" part of pattern "004" is recommended. It automatically suggests appropriate splitting patterns based on the Based on these predefined patterns, the user can specify It is also possible to customize the hierarchical level of division for each item. For example, While selecting the "003" track, set the "Intro" section to a more detailed "L4" level (for example, For example, you can flexibly set it up by dividing it into individual items such as [Background Technology] or [Issues]. By doing so, it is possible to determine the division unit that better matches the user's intention. By equipping the system with the concept of components and multiple division patterns based on them, This will enable us to support the creation of high-quality specifications while responding to a variety of needs.

[0099] Referring again to FIG. 5, once the division units are determined, the specification creation support application 1 Prepare prompts (instructions) for the generation AI 20 to create a partial specification for each position. The prompt includes the main points of the content to be explained in the division unit, references, etc. The specific parts of the source information that should be included, the expected writing style, length, and keywords to include. The design of this prompt directly affects the quality of the generated partial specification. Therefore, it is desirable to be as specific and clear as possible.

[0100] Next, the input means 14 of the specification preparation support application 1 inputs the prepared prompt and the Generate AI20 by combining relevant source information (or excerpts or summaries) from the prompt. The generation AI 20 then inputs the specified division number (step S106). Generate an explanatory text (partial specification) about the unit and return it to the specification creation support application 1. The acquisition means 15 acquires the generated partial specification (step S107). The process from step S105 to step S107 is repeated for all the division units that have been determined. This is executed (corresponding to "repeated" in FIG. 8). This corresponds to the processing of 18 (loopback in case of NO).

[0101] When the partial specifications for all division units have been obtained, the result of the specification preparation support application 1 will be displayed. The combining means 16 combines the acquired multiple partial specifications in a predetermined order to create one continuous specification. A combined specification draft (combined specification) is formed (step S108). Equivalent to 19.

[0102] Thereafter, as an option, the specification preparation support application 1 may Then, the generation AI 20 can be requested to check and adjust the data again (step S109). The check requests include, for example, verifying logical consistency, eliminating redundancies in expressions, and unifying terminology. Correcting typos or reviewing from a specific perspective (e.g., suggesting inventive step) The generation AI 20 analyzes and checks the combined specification based on this request. The results (such as suggested amendments and findings) are sent to the specification preparation support application 1 (step S110 This step (S109, S110) is the same as step S20 (adjusting the combined specification) in FIG. ) The user can review the results of this check and approve or further Modifications can also be made.

[0103] Finally, the draft specification, after all necessary checks and adjustments have been completed, is uploaded to the specification writing support app 1. The output means 18 outputs the result to the user terminal 30 (step S111). This corresponds to step S21 in 5. The user reviews this final draft and, if necessary, The document is then manually edited and completed as a formal specification for a patent application. The specification writing support system 1000 of this embodiment includes a specification writing support application 1 and a generation AI Through close collaboration with the 20, we are able to efficiently advance each stage of specification preparation and produce high-quality drafts. Support the development of

[0104] In operation of the specification preparation support system 1000 according to this embodiment, the user operates the user terminal. Users can interact with the system via a specification creation support app1 running on the device, and input the necessary information. FIG. 11 shows the “Source” screen that the specification preparation support application 1 presents to the user. This screen is a diagram showing an example of an "information input screen." This screen is used when a user starts a new specification writing project. It is used when starting a project or adding or modifying information in an existing project. The information entered through the screen is processed in step S11 (source) in the flowchart of FIG. This is the basis for the process of (obtaining information) and is also the basis for the decision in step S15 (determining division units). It is also used as a basis for determining the determination means 13.

[0105] From here on, we will show you example screens of the specification creation support app 1.

[0106] 11 is a diagram illustrating an example of a source information input screen. Source information input screen shown in FIG. The area is roughly divided into three areas: "Project information input area," "Source file input area," and "Attribute specification area." This screen is composed of three main areas: a fixed area, a display area, and a display unit. These areas allow users to input information intuitively and efficiently. It is logically arranged and has an appropriate input interface (text boxes, file selection, etc.) These areas provide the necessary Enter or specify the information, and then click the "OK" button (or "Save," "Start," or whatever you want to do). By pressing the button with the name corresponding to the input, the input information will be registered in the system and the specification creation process will begin. Initiate or progress a process.

[0107] The "Case Information Input Area" is where you enter basic identification information for the patent case to which the specification you are creating relates. In the example in Figure 11, there are input fields for "Project Name" and "Management Number". In the "Name of the Invention" field, enter the name of the invention, for example, "XX System." This allows users and systems to easily The "Management Number" can be used to identify cases within the office or The unique identification code and serial number used within the company are entered. This is useful for managing the specifications and related data that are created and for later searching and referencing. This information is important primarily from a project management perspective.

[0108] The "Source File Input Area" is where you enter the specific documents (source information) required to create the specification. ) to the system. There are multiple input fields to accommodate various information sources. In the example in Figure 11, Four types of software: "Inventor's documents," "Transcription records," "Prior art documents," and "Patent reference publications" Each input field contains an example of a typical file name (e.g., "invention_spec.docx", "meeting_transcript.txt", etc.) as placeholders This will give the user an idea of ​​what type of file to input. Click the "Choose File..." button to the right of each item and select your local Select the appropriate file from storage or a shared folder on the network and upload it. Loading is possible with the ability to specify multiple files at once and drag-and-drop. These source files can be uploaded to the server. The contents are stored in the storage means 11 of the server 10, and are analyzed and used as basic data for creating specifications. will be done.

[0109] The "attribute specification area" is an area for specifying various attribute information related to the patent application. The attribute information specified here is determined by the determination means 13 of the specification preparation support application 1. In the example in Figure 11, "Technology field" and "Customer" are important factors in determining the division unit. Attributes such as "urgency" and "generation AI selection" are displayed in a drop-down list format. "Technical field" indicates the technical category to which the invention belongs (e.g., mechanical, electrical, chemical, software). Select the field (Software, Bio, etc.) that you want the system to use to create a technical specification. The customer can consider the configuration and preferred division pattern (see Figure 10). Select the organization to which the applicant or inventor belongs (e.g., Company A, University B, etc.). If there is a specification creation policy or past application trends, a division strategy and expression style can be developed accordingly. "Urgency" is useful for determining deadlines and priorities (e.g., normal, urgent, top priority, etc.). If the urgency is high, create a draft quickly in coarser divisions. A split pattern (for example, pattern ID "002" in Figure 10) is likely to be selected, and time is wasted. If there is room for improvement, we can pursue quality in more detailed divisions (for example, the pattern in Figure 10). Turn ID "004" is likely to be selected. In "Generation AI Selection", the system cooperates The type and model of the AI20 to be generated (e.g., Megigi, ChatXYZ, ClaudePro, etc.) or a specific Depending on the characteristics of the generation AI, the optimal prompt format and split unit Since the granularity of the data may differ, providing this option will help generate higher quality results. Adjustments can be made to obtain

[0110] After entering and selecting the necessary information in each of these fields, the user clicks the "OK" button. This allows the entered project information, source files, and attribute information to be The control means 19 of the specification preparation support application 1 controls the specification preparation support application 1 based on the information as shown in FIG. In particular, the specification creation process starts. The attribute information thus obtained is passed to the determining means 13, and a division pattern table such as that shown in FIG. The optimal division unit is determined by referring to the data or through inference using a machine learning model (see step 5 in Figure 5). In this way, the source information input screen is used for the user and the system. This is an important interface between the two, and accurate information input here will affect the subsequent automated specification creation process. This will have a major impact on the quality and efficiency of the process. By doing so, the system will be able to maximize its capabilities and produce specification drafts that meet the user's expectations. contributes to the generation of

[0111] 12 is a diagram illustrating an example of a division unit setting screen. This division unit setting screen may be used to set, for example, The division unit determined in step S15 is used when the user checks, corrects, or edits the division unit. The "Division Unit Setting Screen" shown in Figure 12 is mainly displayed in a tree format. A list of "Title Names" and an area to display or set "Related Information" associated with each division unit This screen allows users to intuitively grasp the overall structure of the specification and the granularity of each part. This screen allows the user to control the system provided Check the proposed division unit, and adjust the division depth (hierarchical level) according to your intention and the characteristics of the project. ) and add, delete, or change the order of specific division units.

[0112] The "Division Unit Name" column displays each element that makes up the specification in a hierarchical manner. Example of Figure 12 In the example, the highest level division units are "Introduction (L3)", "Structure", and "Action". These correspond to the main logical blocks of the third layer (L3) explained in Figure 9. In this example, "(Standard)" is shown as the related information for "(L3)". For example, in the intro section of pattern IDs "002" and "003" in Figure 10, This suggests that it is a standard division unit that is processed collectively at a hierarchical level. To the left of the title there is a "+" or "-" icon to expand or collapse the tree. The user can operate these to check the lower hierarchical structure. Pressing it sets the lower level (finer) division unit, and pressing the "-" icon sets the upper level The (coarser) division unit of the level is set.

[0113] For example, the division unit "Configuration" has subordinates such as "Figure 1," "Figure 2," and "Figure 3." It also includes division units (corresponding to the fourth layer L4 in FIG. 9) corresponding to individual drawings. The division units in "Figure 1" are divided into sub-units such as "server," "user terminal," and "generation AI." In addition, the more detailed division units (fifth layer L in Fig. 9) corresponding to the main components shown in Fig. 1 5) as an option. The user can Clicking on the icon will display more detailed division levels and allow you to select them if necessary. Similarly, the division unit of "action" can be divided into "Figure 4" ", "Figure 5", "Figure 6", etc. For example, "Figure 4" is further divided into "Step Step S101, Step S102, etc., which correspond to each step in the flowchart. It may include detailed division units.

[0114] The "Related Information" column contains supplementary information for each division unit and the information used to process that unit. In the example in Figure 12, the section "Figure 1" is displayed. The file name "Figure 1.pdf" is shown as related information for each division. When the AI20 is made to create an explanation about the division unit of this "Figure 1", the system This indicates that the contents of the drawing file "Fig.1.pdf" will be used as the primary reference. For example, "Figure 2" is associated with "Figure 2.pdf", "Figure 3" is associated with "Figure 3.pdf", etc. By checking this related information column, you can understand what information is used in the processing of each division unit. The system also allows users to edit this related information field and Source information files to be referenced (e.g., specific invention description chapters or specific prior art It may be possible to provide a function that allows explicit specification of the source of the document.

[0115] On this division unit setting screen, the user first selects the overall image of the division units proposed by the system. This suggestion is based on the technical field and urgency specified on the source information input screen in Figure 11, The selected division pattern (see Figure 10) or the structure of the source information itself (e.g., The determination means 13 automatically determines the number of drawings based on the chapter structure of the explanatory material, etc. If the user thinks that the suggestion is appropriate, he or she can simply press the "OK" button to proceed to the next step. The process can proceed (step S16 onwards in FIG. 5).

[0116] If the user wants to make changes to the proposed division unit, they can perform various operations on this screen. For example, if you want to change the hierarchical level of a specific division unit, you can change the target unit by Select it and use the context menu or dedicated operation buttons to move to a higher level (coarse division) ) or can be expanded to a lower level (fine division). If a specific division unit (for example, the explanation of "Server" under "Figure 1") is deemed unnecessary, It is also possible to remove a unit from the list. Conversely, it is possible to add a new unit that was not suggested by the system. It is also possible to add divisional units (for example, descriptions of embodiments corresponding to particular claims). Furthermore, if you feel that the order of division units is inappropriate for the logical flow of the specification, It is also possible to change the order of units by dragging and dropping. The "Remove" button is a function that discards the changes made by the user and returns to the initial state proposed by the system. to provide.

[0117] In this way, the division unit setting screen allows users to create a division structure that can be considered a "blueprint" for creating specifications. This screen provides a powerful tool for visually checking and finely controlling the structure of the system. By appropriately setting the division unit through prompts, resulting in accurate and comprehensive understanding of the invention. This increases the likelihood of generating a logical, high-quality partial specification that reflects the Combining these high-quality partial specifications will lead to an improvement in the overall quality of the final specification draft. This process of user confirmation and adjustment of division units is a combination of fully automated and human expert judgment. It plays an important role in finding the optimum balance between

[0118] 3. Working Example This specification was prepared using a method based on the present disclosure. The Google Gemini 2.5 Pro was used. The draft was created in May 2025. As a result, approximately 40 A patent specification (this paper) of ,000 characters was obtained. On the other hand, the same source information and the same generation A If you use I20 and have students write the entire statement in one prompt without dividing it up, it will cost about 10,000 yen. The patent specification obtained in this single prompt was simply Not only is the volume small, but some drawings are missing explanations and the illustrated elements are not explained. In this way, the method based on the present invention was used to divide the product into certain units, and there were problems with the quality. By dividing the time and effort into two parts, we were able to obtain a patent specification that was better in both quality and quantity. Finally, it is checked and corrected by a human (patent attorney).

[0119] 4. Variations Although the embodiments of the present invention have been described in detail above, the present invention is not limited to the above embodiments. The following are some examples of modifications that can be made without departing from the spirit of the invention. These modifications may be used alone or in combination with at least some of the other modifications. A combination of these may also be employed.

[0120] 4-1.System Configuration In the above embodiment, the specification preparation support system 1000 includes the server 10, the generation AI 20, and user terminal 30. The system configuration is not limited to this. For example, the functions of the generation AI 20 may be completely integrated into the server 10. Conversely, most of the main processing functions of the specification preparation support application 1 are user-defined. It can also be configured as a standalone application that runs on the terminal 30. In the latter case, the user terminal 30 directly communicates with an external AI generation service. In addition, the functions of the server 10 may be distributed to multiple servers in a distributed system configuration, or a cluster system. It is also possible to use a cloud computing environment. For example, the type of source information and Depending on the volume, we can load balance the processing or use a microservice architecture that specializes in specific processing. Furthermore, a single specific model can be used as the generation AI 20. In addition to using the above, depending on the nature of the division unit to be processed and the required style of text, It is also effective to use multiple different generative AI models. The model is highly capable of comprehensively collecting and summarizing information, and the configuration and operation of the embodiment are explained. By using a model that is good at generating logical and precise sentences, In this case, the specification preparation support application 1 (or the server 10) can improve the quality of the specification. ) has the ability to select the appropriate generative AI model and distribute instructions.

[0121] 4-2. Various Data In the above embodiment, the source information includes invention explanation materials, meeting audio (transcription), Prior art documents, reference specifications, etc. are listed as examples (see Figure 6), but some of these may be omitted. In addition, various other data can be used as source information. Related experimental data (numerical data, graphs, images, etc.), simulation results, product Specifications, marketing materials, and competitor product information also help deepen understanding of your invention and make it more persuasive. This can be the source information for creating a powerful specification. It supports these various data formats, analyzes the content as needed, and extracts text information. It is desirable to have a function to check the relationship between the drawings and the documents. ) and division unit settings (see Figure 12) so that more detailed metadata can be added. For example, for each drawing, it is possible to determine which claim is closely related to that drawing. or to explain which technical feature each division unit The expected length of the text, the level of expertise, and the priority of specific source information to be referenced are all factors that influence the ranking. By being able to set the priority in detail, more detailed control is possible. In addition, users can access specifications they have previously prepared, as well as exemplary specifications that are highly regarded in a particular field. These are registered in the system as "reference style data," and the generation AI 20 uses them as reference. It is also possible to consider a function to adjust writing style and expression.

[0122] 4-3. Processing content and order The processing flow shown in Figures 5 and 8 is an example, and the order and content of each step can be flexibly changed. Some steps may be omitted. For example, the drafting of claims (see Figure 5) Preparation steps such as creating a drawing list (S12), creating a drawing list (S13), and creating drawings (S14) are It is not necessary to do it in this order, and you can change the order depending on your work style and the situation of the project. After the division unit is determined (S15), Users can manually enter or edit summaries, keywords, and main points for each division. It is also effective to provide a step for issuing instructions (prompts) to the generation AI 20. This will make the partial specifications more specific, and we can expect to improve the quality of the partial specifications that are generated. A more interactive approach to the process of combining specifications (S19) and reconciling specifications (S20) For example, after the system presents a proposed combined specification, the user point out a specific part and instruct the AI ​​20 to regenerate or correct only that part. It is possible to provide an interface that allows multiple representations to be presented and selected. In addition, the main part (e.g., the introductory part and the main embodiment) may be completed before all the partial specifications are completed. First, combine and adjust only the remaining (unprocessed) parts, receive user reviews, and then apply that feedback to the remaining parts. An iterative and step-by-step process of creation that reflects division-level processing is also effective.

[0123] 4-4. Splitting according to attribute information The source information includes attribute information of the patent application (technical field, applicant, urgency, etc.), and the determination means 13 may determine the division unit according to these attribute information. This configuration is more suitable for the case. For example, if the technical field is "software," In this case, each step in the flowchart is used as a division unit, and if the urgency level is "urgent," the pattern in Figure 10 is used. One possible control would be to select a coarse division such as the segment "002".

[0124] 4-5.Division unit template The determining means 13 determines the plurality of defined options (i.e., templates). Alternatively, one option from among a plurality of options (e.g., a group) may be determined as a division unit. This improves operability and serves as the basis for automatic suggestions by the system.

[0125] 4-6.Determining division units based on drawings The source information may include drawings relating to the invention, and the division unit may be determined based on the drawings ( For example, by drawing or by component within a drawing). This organization is especially important when drawings play a key role. It is effective in generating detailed and accurate descriptions of inventions that fulfill multiple roles. Each of the draft drawings can be used as a division unit, or each of multiple components in one draft drawing can be used as a division unit. By using this method, the generation AI 20 can focus on each part and generate high-quality descriptions.

[0126] 4-7.Source Information The source information may include a transcript of a meeting between the inventor and the patent attorney. This is not limited to meetings between inventors and patent attorneys, but also between inventors and between inventors and companies' intellectual property rights. It can be any type of participant, such as a club member or a group of engineers. It can be anything, including explanations, questions and answers, and brainstorming. This structure is useful for gaining a deeper understanding of the background and nuances of the invention and reflecting this in the specification. This is beneficial (see Figure 6).

[0127] The specification writing support app 1 allows inventors and patent attorneys to read and understand the generated draft specification. The function to revise the draft text using the transcription and generation AI of the meeting is also available. This configuration may streamline the review process and facilitate collaborative specification review. This correction method is used to improve the quality of the text. "The explanation for Figure YY is insufficient," etc., and gives specific correction instructions to the generation AI 20. It is possible.

[0128] 4-8. Use of machine learning Information about how the division units are determined and the quality of the resulting draft text (e.g., The machine learning model is trained using the evaluation and correction history by the decision means 13, and the trained A model may be used to determine the division unit. This configuration becomes more accurate as the system gains experience. In order to realize a self-improving system that can automatically propose the most optimal division unit, This learning method is highly effective because it accumulates user feedback and past success stories. The splitting strategy is continuously optimized. User feedback can be, for example, The subjective evaluation score and the (objective) time spent checking the draft specification were the minimum. Either one can be used.

[0129] In the above embodiment and the modifications, the generated AI 20 is mainly Although the above explanation has focused on the use of AI (especially The use of machine learning models is not limited to this. Applying AI technology to various other stages to streamline the process and reduce the burden on users. It is possible that...

[0130] As a first example, when a user inputs or enters the "Technical Field" on the source information input screen (see Figure 11), When selecting or selecting a technology, or when the system infers the technology field from the content of the source information, AI automatically recommends or selects the most suitable "Drawing List Template" for your technical field This is a feature that can be used to create templates for the drawing list. In this case, the system uses a large amount of past patent specification data. and its technical field (e.g., patent classification information such as IPC, FI, and F-terms) The machine learning model is trained on the drawings (those that were actually used) and the configuration of the drawings that were actually used. Based on the technical field information (or estimated technical field information) of the newly entered case, This trained model is used to analyze drawing lists that are considered to be typical and effective in the technical field. The user can create a template for the drawing (including recommendations for drawing types, order, and what elements should be illustrated on each drawing). Based on this AI recommendation, users can more quickly and accurately select the drawing configuration. You can start fighting.

[0131] As a second example, a user can create source information (see Figure 6, e.g., invention explanation materials or meeting notes) When a patent application is entered into the system, the AI ​​analyzes the content of the source information and determines whether the invention is A function that automatically predicts the "technical field" that a user is likely to belong to and presents it to the user is conceivable. For example, estimation in this technical field involves identifying characteristic keywords from text data in the source information. It extracts codes and technical terms and then uses them based on their frequency of appearance and relevance. Classification using a classification model (e.g., a supervised learning model with patent classification as the correct label) The estimated technical field is then checked by the user and added to the drawing list template. Selection of a division pattern, which will be described later, or optimization of instructions to the generation AI 20, etc. This allows users to easily specify the technical field when they are unsure which field to use, or when they are dealing with multiple technical fields. Even in the case of inventions that require proper initialization, it is possible to support appropriate initialization.

[0132] These AI-powered features incorporate user feedback and continuously learn, improving their accuracy. For example, AI can recommend drawing list templates and If a user makes corrections or changes to the estimated technology field, the details of the operation are stored as learning data. This allows the system to be used as it is and then retrained using machine learning models. This will make them smarter and enable them to provide support that is more tailored to individual cases and user preferences. This kind of multifaceted use of AI will not just automate document creation, but will also improve the efficiency of specification creation. It also supports users in the upstream processes, such as organizing information and formulating strategies at the initial stage. It provides powerful support and has the potential to dramatically improve the quality and efficiency of specification preparation overall. This is the case where the "learning means" as described in claim 12 is included, and the determining means is The idea of ​​"determining the division unit using the model" is expanded to aspects other than determining the division unit. It can be seen as an extension.

[0133] 4-9. Drawing list template In the above embodiment, the creation of the drawing list (step S13 in FIG. 5) is made more efficient. In order to facilitate appropriate drawing organization according to technical field, a drawing list template function is provided. In this modification, the specification preparation support application 1 or the server The storage means 11 of the 10 stores in advance templates of drawing lists corresponding to a plurality of technical fields. Each template is designed to represent the typical drawing structure required in a particular technical field. Composition (guideline for drawing numbers, suggested drawing titles, list of major elements to be shown on each drawing) This defines the following:

[0134] For example, among software-related inventions, there is a technology called "server / client system." For the technical field, Figure 1 "System Overview", Figure 2 "Functional Configuration", Figure 3 "Server Hardware Figure 4: Client hardware configuration diagram, Figure 5: Main processing sequence Figure 6: Database configuration example, Figure 7: Input screen example, Figure 8: Output screen example A template is provided that defines a series of drawing configurations such as For the technical field of "Internet-based applications," Figure 1 "System Overview (Other Systems)" (If it includes cooperation with other devices, the outline diagram of the cooperation is included)”, Figure 2 “Main function configuration diagram”, Figure 3 “Terminal hardware Figure 4: Flowchart of main processing; Figure 5: Main database or data Templates such as Figure 6 "Example of structure", Figure 7 "Example of main input screen", and Figure 8 "Example of main output screen" Furthermore, for inventions that involve physical structures such as machine parts, Figure 1 "Diagram showing the mounting position of parts in the entire device", Figure 2 "Overall perspective view of a single part", Figure 3 "Exploded perspective view of the part (if it consists of multiple components)", Figure 4 "Main cross-sectional view of the part (internal A drawing template such as "When showing a part structure" is effective.

[0135] When the user specifies the technical field of the invention on the source information input screen (see Figure 11), The decision means 13 of the development support application 1 (or a dedicated template selection means) selects the designated technique. A template of a drawing list corresponding to the technical field is automatically read out from the storage means 11. Using this template as a model, an initial draft of the drawing list (see Figure 7) is generated and The user can review this initial draft and add, delete, or modify drawings to suit their invention. You can modify or specify the titles and illustrated elements of each drawing. The system further analyzes the content of the source information (e.g., keywords in the invention description and drawing plans). Based on the analysis and the selected template, more specific drawing list candidates are automatically generated. This template feature allows users, especially inexperienced inventors and patent attorneys, to Even if the drawings are not in accordance with the general guidelines, it is easy to understand the structure of the drawings that are generally recommended in the technical field, and It is possible to efficiently create an appropriate drawing list, which results in improved understanding of the specification. It also contributes to ease of interpretation and clarification of the scope of rights.

[0136] Here, the system automatically selects or users can select templates for the drawing list according to the technical field. We have explained the function of presenting the results to the user. To support the construction of drawings according to the content, we use a generation AI (or a dedicated analysis and proposal AI) ) to proactively suggest drawings to users that should be added or drawing configurations that should be changed. In this variant, the system may provide a source initially entered by the user. Information (see Figure 6, e.g., invention description and proposed claims) and selected or initially generated drawings The generated AI 20 is then sent to the list template (as described in the previous variant). Enter.

[0137] The generation AI 20 comprehensively analyzes this input information and makes the content of the invention clearer and more effective. I thought it would be a good idea to add it to the current drawing list template to explain it more clearly. Identify drawings that need to be revised or drawings that need to be changed, and provide the reasons and specifics. For example, AI can analyze the invention description in the source information and present it to the user along with suggestions for visual content. I read the documentation and found that even though the explanation of a particular component is detailed, I still don't understand the If it is determined that the template does not include a drawing specifically showing the element, It is recommended to add an exploded perspective view or a cross-sectional view to show the detailed structure of the The technical characteristics of the components will become clearer. Alternatively, explain the principles of operation of specific functional features recited in the proposed claim. If you determine that your current drawing list is lacking flowcharts and block diagrams for If so, this will be pointed out and appropriate figures will be added.

[0138] Furthermore, the generation AI 20 may generate a drawing draft (e.g., a sketch handwritten by an inventor) in the source information. We analyze images (such as sketches and rough outlines) and use the results to create more appropriate expressions and drawings for patent drawings. It is also possible to propose a method of illustration. For example, for a handwritten block diagram, Standard drafting practices are used for block placement, connecting lines, and reference symbols. In addition, the same components are depicted in different ways on multiple drawings. If there are any discrepancies in the reference codes, they will be detected and consistency will be achieved. It is also effective to propose amendments to the process.

[0139] In this way, the AI ​​can analyze the contents of the source information and the current drawing list (or its template) Drawings that are considered useful for comparing and analyzing the invention and clarifying the scope of the patent right. By proactively suggesting additions and modifications, users can identify things in the drawing that they themselves may not have noticed. It is especially useful for inexperienced inventors and patent attorneys, as it makes it easier to discover what is lacking or what needs to be improved. Your objective and specific suggestions will be of great help in preparing a quality drawing set. The user can then refer to the AI's suggestions to revise and expand the drawing list, and then decide whether to proceed based on the results. This function proceeds to the actual drawing creation (step S14 in FIG. 5). This is performed in the step S13) or before the step S15) of determining the division unit. This is effective, and the drawing information itself that the generating AI 20 refers to when creating a partial specification is This will contribute to improving the quality of the final specification as a whole. Gar.

[0140] 4-10. Application UI In the above embodiment, the specification preparation support application 1 provides a user interface. The source information input screen (Fig. 11) and the division unit setting screen (Fig. 12) are shown as examples. The user interface has been redesigned to allow for more intuitive and flexible control of system operations. In particular, it is important to decide which process to perform among the series of processes for preparing a specification (see Figure 5). The user can decide which processing to perform and which processing results to check. Allows users to specify more precisely what corrections or automation levels they want. An interface like this could be envisaged.

[0141] For example, the specification preparation support application 1 may include the main processing steps (e.g., ,S12: Create a claim proposal, S13: Create a drawing list, S14: Create drawings, S15 : Determine the division unit, S17: Prepare a partial specification for the target part, S20: Prepare a combined specification The system should display a list of steps (such as adjustments) and provide a check box next to each step. The user selects the check box corresponding to the process that the generated AI 20 wants to execute. and check the boxes for the processes you want to perform yourself or that are not supported by this system. This allows users to choose their skills, the nature of the job, and the time they have available. For example, if you are confident in creating claims, However, if you want to leave the creation of the drawing list to AI, or conversely, if you want to describe the basic structure yourself and then It can meet a variety of needs, such as when you want to ask AI to only flesh out the details or refine the expression after the sentence. It becomes possible.

[0142] Furthermore, even when the generation AI 20 executes a process, the user can freely select the processing result. You can select whether or not to insert a confirmation and approval step by using additional checkboxes and option settings. It is also effective to be able to specify the process. For example, the process of "Prepare a partial specification (S17)" If you select to have the generation AI 20 perform the process, the user will be able to confirm the process after each partial specification is generated. If this option is on, The system presents each partial specification to the user as it is generated, and the user can After checking the content and giving approval (or instructions for correction), the process moves to the next division unit. If this option is off, partial specifications are generated for all division units consecutively, and The former is a method for finely controlling the output of the generation AI. This is effective when you want to correct the process going in the wrong direction early on, and the latter is a big problem. This is suitable for cases where you want to process a large amount of data quickly. Other AI-based processes such as determining division units (S15) and adjusting combined specifications (S20) It can be similarly provided.

[0143] Such UI variations allow users to control the level of automation of the system at their discretion. By making it possible to control the process, users can choose the process they want, rather than forcing them to conform to a uniform process flow. Enables flexible system usage according to the diverse work styles and expertise of users and the requirements of each project. As a result, users can use generative AI technology with greater confidence, and can use their own expertise and AI By effectively combining the capabilities of I, we maximize both the quality of the final specification and the efficiency of its creation. This flexible control interface is expected to increase the user's initiative. This is an important factor in maximizing the benefits of AI-assisted assistance while respecting gender. .

[0144] 4-11.Other These modifications are intended to accommodate a wider variety of user needs and interests while maintaining the basic concept of the present invention. This is an extension to accommodate the operating environment and is included in the technical scope of the present invention.

[0145] The UI and data described in the embodiments are merely examples. The part may be replaced with another form of UI or data. The database described above is a single database that has been partially merged with the multiple databases described above. It may also be operated as a database.

[0146] In the embodiments, what is described as a single software element (e.g., a generating AI) Depending on the detailed functions, multiple software elements (e.g., multiple generation AIs) are used. In addition, in the embodiment, software such as the specification preparation support application 1 is executed. Some of the processing described as being performed by the user may also be performed by the user.

[0147] The functional configuration and hardware of the specification preparation support system 1000 described in the embodiment The configuration is merely an example. may be omitted, or new functional or hardware elements may be added. Furthermore, the correspondence between the functional elements and the hardware elements described in the embodiments is merely This is an example. The functions required for the specification preparation support system 1000 can be implemented. If there is, each functional element may be implemented in any hardware. At least some of the functions described as being possessed by the server 10 are implemented in the user terminal 30. The server 10 may be a physical server or a virtual server (a so-called (including cloud).

[0148] The program (program product) executed by the processor such as the CPU 101 is a computer program. Distributed in a form recorded on a readable non-transitory recording medium (such as a DVD-ROM) Alternatively, the software may be provided in a downloadable format from a server on a network. .

[0149] The scope of application of the technology of this disclosure is not limited to the drafting of patent specifications. For example, legal documents , technical documents, academic documents, business documents, educational documents, and / or content Legal documents include, for example, contracts, pleadings, briefs, and Technical documents include technical reports, research papers, product manuals, instruction manuals, and and software specifications. Business documents include business plans, proposals, and research reports. Educational documents include textbooks and educational materials. Content includes news articles and blog posts. In these application fields, the "source information indicating the attributes of the patent application" of this disclosure is also The idea is that the division unit is determined according to the attribute information, and the decision means determines the division unit according to the attribute information. For example, the optimal division unit and document are determined according to the "attributes" such as the type of document, the target audience, and the urgency. A mechanism can be introduced in which the system suggests configuration templates.

[0150] FIG. 13 shows examples of application to fields other than patent specifications. The technology of the present disclosure can be applied by defining source information and division units. [Explanation of symbols]

[0151] 1...specification statement preparation support application, 10...server, 11...storage means, 12...reception means, 13 ...Determining means, 14...Input means, 15...Acquiring means, 16...Combining means, 17...Adjusting means, 18 ...output means, 19...control means, 20...generation AI, 30...user terminal, 90...network 101...CPU, 102...memory, 103...storage, 104...communication device, 107... Bus, 301...CPU, 302...memory, 303...storage, 304...communication device, 30 5...input device, 306...output device, 307...bus, 1000...specification preparation support system

Claims

1. a determining means for determining a division unit for preparing a draft specification for a patent application for the invention based on source information relating to the invention; an input means for inputting an instruction to a generation AI to create a description of the invention for each division unit by referring to the source information; an output means for outputting the draft sentence based on a combination of outputs obtained from the generation AI for each division unit; and the source information includes attribute information indicating attributes of the patent application; The determining means determines the division unit in accordance with the attribute information. Information processing device.

2. The attribute information includes the technical field of the invention. The information processing device according to claim 1 .

3. The attribute information includes the applicant of the patent application. The information processing device according to claim 1 .

4. The attribute information includes the urgency of the patent application. The information processing device according to claim 1 .

5. A determination means for determining a division unit when preparing a draft specification for a patent application for the invention based on source information related to the invention; an input means for inputting an instruction to a generation AI to create a description of the invention for each division unit by referring to the source information; an output means for outputting the draft sentence based on a combination of outputs obtained from the generation AI for each division unit; and the determining means determines one option from among a plurality of defined options as the division unit; the source information includes drawings related to the invention; The plurality of options include an option of dividing the unit based on the drawing. Information processing device.

6. The drawings relating to the invention include a plurality of draft drawings in the application documents of the invention; The plurality of options include an option in which each of the plurality of draft drawings is a division unit. The information processing device according to claim 5 .

7. One draft drawing among the plurality of draft drawings includes a plurality of components; The plurality of options include an option in which each of a plurality of components is a division unit for the one draft drawing. The information processing device according to claim 6 .

8. A determination means for determining a division unit when preparing a draft specification for a patent application for the invention based on source information related to the invention; an input means for inputting an instruction to a generation AI to create a description of the invention for each division unit by referring to the source information; an output means for outputting the draft sentence based on a combination of outputs obtained from the generation AI for each of the division units; A correction means for correcting the draft text of the specification by transcribing the meeting in which the inventor of the invention and the patent attorney read the draft text together and using the generated AI; An information processing device having the above.

9. A determination means for determining a division unit when preparing a draft specification for a patent application for the invention based on source information related to the invention; an input means for inputting an instruction to a generation AI to create a description of the invention for each division unit by referring to the source information; an output means for outputting the draft sentence based on a combination of outputs obtained from the generation AI for each division unit; and the source information includes a drawing list; a storage means for storing a plurality of templates for the drawing list; The input means refers to the source information to which a drawing list of the template selected based on the source information has been added, and inputs an instruction to create a description of the invention for each division unit. Information processing device.

10. a corresponding technical field is defined for each of the plurality of templates; The input means refers to the source information to which a list of drawings of a template corresponding to the technical field identified based on the source information has been added, and inputs an instruction to create a description of the invention for each division unit. The information processing device according to claim 9 .

11. The input means inputs an instruction to make the generation AI suggest drawings that should be added to the drawing list by referring to the source information and the drawing list of the template selected based on the source information. The information processing device according to claim 9 .

12. A computer comprising: determining a division unit for drafting a specification for filing a patent application for the invention based on source information relating to the invention; inputting an instruction to a generation AI to create a description of the invention for each division unit by referring to the source information; outputting the draft sentence based on a combination of outputs obtained from the generation AI for each of the division units; and the source information includes attribute information indicating attributes of the patent application; In the determining step, the division unit is determined according to the attribute information. Information processing methods.

13. A computer comprising: determining a division unit for drafting a specification for filing a patent application for the invention based on source information relating to the invention; inputting an instruction to a generation AI to create a description of the invention for each division unit by referring to the source information; outputting the draft sentence based on a combination of outputs obtained from the generation AI for each of the division units; and In the determining step, one option is determined as the division unit from among a plurality of defined options; the source information includes drawings related to the invention; The plurality of options include an option of dividing the unit based on the drawing. Information processing methods.

14. A computer comprising: determining a division unit for drafting a specification for filing a patent application for the invention based on source information relating to the invention; inputting an instruction to a generation AI to create a description of the invention for each division unit by referring to the source information; outputting the draft sentence based on a combination of outputs obtained from the generation AI for each of the division units; A step of transcribing a meeting in which the inventor of the invention and the patent attorney read through the draft of the specification and revising the draft using the generation AI; An information processing method comprising:

15. A computer comprising: determining a division unit for drafting a specification for filing a patent application for the invention based on source information relating to the invention; inputting an instruction to a generation AI to create a description of the invention for each division unit by referring to the source information; outputting the draft sentence based on a combination of outputs obtained from the generation AI for each of the division units; and the source information includes a drawing list; storing a plurality of templates for the drawing list; In the input step, an instruction to create a description of the invention for each division unit is input by referring to the source information to which the drawing list of the template selected based on the source information has been added. Information processing methods.

16. On the computer, determining a division unit for drafting a specification for filing a patent application for the invention based on source information relating to the invention; inputting an instruction to a generation AI to create a description of the invention for each division unit by referring to the source information; outputting the draft sentence based on a combination of outputs obtained from the generation AI for each of the division units; A program for executing the source information includes attribute information indicating attributes of the patent application; In the determining step, the division unit is determined according to the attribute information. program.

17. A computer comprising: determining a division unit for drafting a specification for filing a patent application for the invention based on source information relating to the invention; inputting an instruction to a generation AI to create a description of the invention for each division unit by referring to the source information; outputting the draft sentence based on a combination of outputs obtained from the generation AI for each of the division units; A program for executing In the determining step, one option is determined as the division unit from among a plurality of defined options; the source information includes drawings related to the invention; The plurality of options include an option of dividing the unit based on the drawing. program.

18. A computer comprising: determining a division unit for drafting a specification for filing a patent application for the invention based on source information relating to the invention; inputting an instruction to a generation AI to create a description of the invention for each division unit by referring to the source information; outputting the draft sentence based on a combination of outputs obtained from the generation AI for each of the division units; A step of transcribing a meeting in which the inventor of the invention and the patent attorney read through the draft of the specification and revising the draft using the generation AI; A program to execute.

19. A computer comprising: determining a division unit for drafting a specification for filing a patent application for the invention based on source information relating to the invention; inputting an instruction to a generation AI to create a description of the invention for each division unit by referring to the source information; outputting the draft sentence based on a combination of outputs obtained from the generation AI for each of the division units; A program for executing the source information includes a drawing list; storing a plurality of templates for the drawing list; In the input step, an instruction to create a description of the invention for each division unit is input by referring to the source information to which the drawing list of the template selected based on the source information has been added. program.

Images