Document preparation apparatus, document preparation method, and document preparation program

Abstract

To save labor and improve efficiency in automated document preparation.SOLUTION: A document preparation apparatus includes: a template setting unit which sets a template for preparing a document; a template accumulation unit which accumulates the set template; a template designation unit which designates a template for use in preparing a document, from the accumulated templates; a necessary information setting unit which sets necessary information which is information necessary for preparing a document using the designated template; a prompt generation unit which generates a prompt for causing a large language model to prepare the document using the designated template and the set necessary information; and a document preparation unit which inputs the generated prompt to the large language model to prepare the document.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a document creation device for improving the efficiency and labor-saving of document creation. [Background technology]

[0002] Regardless of the industry, numerous documents are created daily in the course of business operations, and there is a constant demand for more efficient and labor-saving document creation. Therefore, technologies for automatically creating prescribed documents have been developed. For example, Patent Document 1 discloses a technology for automatically creating invoices based on order data and progress status.

[0003] In recent years, attempts have been made to create documents using automatic text generation by applying large-scale language models (LLMs) such as ChatGPT (Non-Patent Document 1) to plug-ins in word processing software such as Word. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2020-194603 [Non-patent literature]

[0005] [Non-Patent Document 1] https: / / openai.com / chatgpt Summary of the Invention [Problem to be solved by the invention]

[0006] The technology disclosed in Patent Document 1 simply calculates fees based on order data and progress status, and then creates an invoice by converting the calculation results into a predefined format. Therefore, this technology cannot flexibly handle the creation of documents in a wide variety of formats.

[0007] Furthermore, with the technology disclosed in Non-Patent Document 1, it is not easy to generate prompts to input into the LLM in order to create documents according to individual formats and content. Because there are various document creation assumptions and considerations, such as in the case of the proposal document described above, it is not easy to generate prompts to instruct the LLM to create documents that fully reflect these assumptions. Furthermore, when creating a large number of different documents, it is necessary to generate such difficult prompts each time, which places a heavy burden on the user. As such, the conventional technology cannot meet the demands for labor-saving and efficiency in the automatic creation of documents. [Means for solving the problem]

[0008] Therefore, in order to solve the above-mentioned problems, the present invention provides the following document creation device etc. That is, a document creation device is provided that has a template setting unit that sets a template for document creation, a template storage unit that stores the set template, a template designation unit that designates a template to be used for document creation from the stored templates, a necessary information setting unit that sets necessary information that is information necessary for creating a document using the designated template, a prompt generation unit that generates a prompt for causing a large-scale language model to create the document using the designated template and the set necessary information, and a document creation unit that inputs the generated prompt into the large-scale language model to create the document.

[0009] In addition to the above features, the document creation device further includes a reference information setting unit that sets reference information, which is information to be referenced when creating the document, and the prompt generation unit further generates the prompt using the set reference information.

[0010] In addition to the above features, the present invention provides a document creation device further comprising a reformatting unit that reformats the created document.

[0011] In addition to the above features, the present invention provides a document creation device further comprising a verification unit that verifies the created document.

[0012] In addition to the above features, the present invention also provides a document creation device further comprising a prompt storage unit that stores the generated prompts.

[0013] Also provided is a document creation method executed by a computer in a document creation device, the document creation method comprising: a template setting step of setting a template for document creation; a template storage step of storing the set template; a template designation step of designating a template to be used for document creation from the stored templates; a necessary information setting step of setting necessary information that is information necessary for creating a document using the designated template; a prompt generation step of generating a prompt to cause a large-scale language model to create the document using the designated template and the set necessary information; and a document creation step of inputting the generated prompt into the large-scale language model to create the document.

[0014] The present invention also provides a document creation program for causing a computer to execute the following steps in a document creation device: a template setting step for setting a template for document creation; a template storage step for storing the set template; a template designation step for designating a template to be used for document creation from the stored templates; a necessary information setting step for setting necessary information that is information necessary for creating a document using the designated template; a prompt generation step for generating a prompt for creating the document in a large-scale language model using the designated template and the set necessary information; and a document creation step for inputting the generated prompt into the large-scale language model to create the document. [Effects of the Invention]

[0015] According to the present invention, desired documents can be efficiently created by using templates. [Brief explanation of the drawings]

[0016] [Figure 1] FIG. 1 is a block diagram showing an example of a functional configuration of a document creation device according to an embodiment of the present invention; [Figure 2] Conceptual diagram showing an example of the documents to be prepared and the required information [Figure 3] FIG. 1 is a conceptual diagram showing an example of the hardware configuration for realizing a document creation device according to an embodiment of the present invention. [Figure 4] FIG. 1 is a flow chart showing an example of a process flow of a document creation device according to an embodiment of the present invention; DETAILED DESCRIPTION OF THE INVENTION

[0017] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. However, the present invention should not be limited to these embodiments and can be embodied in various forms without departing from the spirit and scope of the present invention.

[0018] <Example> <Summary> For example, when creating documents that must be written according to specified guidelines, such as building confirmation applications submitted when constructing a building, using a large-scale language model (LLM; hereafter abbreviated as LLM), setting up and accumulating templates for each type of document can contribute to efficient document creation.

[0019] The following describes the functions and processing flow of the document creation device, as well as the hardware details. The functional blocks of the device described below can be realized as a combination of hardware and software. Specifically, if a computer is used, these include hardware components such as a CPU (central processing unit), main memory, bus, or secondary storage device (such as a hard disk drive, nonvolatile memory, or storage media such as CDs or DVDs, and drives for reading these media), input devices used to input information, printers, display devices, and other external peripheral devices, as well as interfaces for these external peripheral devices, communication interfaces, driver programs and other application programs for controlling these hardware, and user interface applications. The CPU processes data entered from input devices and other interfaces and stored in memory or on a hard disk in accordance with a program loaded in main memory, processing and storing the data, and generating commands to control the above hardware and software. Alternatively, the functional blocks of the device can be realized by dedicated hardware.

[0020] Furthermore, this invention can be realized not only as an apparatus but also as a method. Furthermore, a part of such an invention can be configured as software. Furthermore, a program used to cause a computer to execute such software, and a recording medium on which the program is fixed, are naturally included within the technical scope of this invention (the same applies throughout this specification).

[0021] <Functional configuration> Fig. 1 is a block diagram showing an example of the functional configuration of a document creation device according to this embodiment. As shown in Fig. 1, the document creation device 100 includes a template setting unit 101, a template storage unit 102, a template designation unit 103, a necessary information setting unit 104, a reference information setting unit 105, a prompt generation unit 106, a document creation unit 107, a reformatting unit 108, a verification unit 109, and a prompt storage unit 110.

[0022] The document creation device 100 is a device that creates documents. This device is embodied, for example, as a function extension (plug-in) for a document creation application such as Word, but may also be realized in a web application. An example of a document that can be created is a document that has a certain amount of information that must be written. For example, a building confirmation application form that must be submitted when constructing a building.

[0023] This device is expected to be particularly effective when creating numerous business documents in companies, offices, etc. In other words, business documents, such as the building confirmation application mentioned above, are created according to a set format and style, so the items and instructions to be written can be generalized. By creating templates for these items and instructions, it is possible to reduce the effort required to write similar, standardized information each time a document is created.

[0024] <Template settings section> The template setting unit 101 has a function of setting templates for document creation. A template is set according to the type of document, etc., and includes the document layout (what to place where and how), the items to be written that are to be placed according to the layout, and instructions for writing them. The items to be written are determined according to the type and purpose of the document, and for example, a template for a building confirmation application has items to be written such as the names of the architect and contractor, the building address, the zoning, the site area, the building area, and the floor area.

[0025] The guidelines for writing are relatively specific and indicate the main points and methods for writing. They often indicate the main points and methods for each item included in the document. For example, the items to be written in the case of a building application and their instructions are set out as follows: "1. Owner": "Enter the owner's information." "2. Agent": "The agent is usually the designer." "3. Designer": "Enter the designer who will be the main designer for this property. If the design is being done by multiple design offices, enter the designers involved in the design in the "Other Designers" section." For ease of explanation, the items to be described and the description guidelines are listed side by side, but the description guidelines for one document may also define the items to be described and the description guidelines for each item.

[0026] In setting the template, the layout may be set in advance, or the layout may be set more specifically as required information in the required information setting section described later. As an example of the latter, in the case of a document whose layout differs depending on the local government, the layout used by the local government for the case to be created is set as required information. The template settings, including the layout, items to be written, and writing instructions, may be configured to be performed in advance by the device or an application related to the device, or may be configured so that the user can perform them themselves.

[0027] <Template storage section> The template storage unit 102 has the function of storing set templates. By storing set templates with a unique ID and name assigned to each, the user can specify any template in the template designation unit described below. Note that the above-described template setting and template storage can also be configured so that various document templates are set and stored in advance when the user uses this document creation device.

[0028] <Template specification section> The template designation unit 103 has a function of designating a template to be used for document creation from stored templates. If the layout of the designated template has not been set, the layout is set in the necessary information setting unit as described above. If the layout of the designated template has already been set, designating one of the templates will designate the layout, items to be entered, and instructions for filling out the document to be created. Template designation may be configured to be performed hierarchically. For example, for a building confirmation application, by subdividing templates for the application subject (building, elevator, building equipment other than elevator, etc.) and the local government to which the application will be submitted, the user can easily designate the template to be used for the document to be created.

[0029] <Required information setting section> The required information setting unit 104 has a function of setting the required information, which is information required to create a document using the specified template. This required information is specific to the document to be created. For example, when creating a building confirmation application, this information includes architectural drawings and wiring diagrams of the actual building that is the subject of the application. Furthermore, if the layout of the document has not been set, a predetermined layout according to the application destination, etc. is set as the required information.

[0030] Figure 2 is a conceptual diagram showing an example of documents to be created and the necessary information. As shown in the figure, when creating a "construction plan summary" 201 for a building confirmation application, for example, a "neighborhood floor plan" 202 and a "layout plan" 203 must be included as description items. These descriptions are based on a "map" 204 showing the location of the object of the application and a "layout plan" 205 of that object, and these "maps" and "layout plans" correspond to the necessary information.

[0031] The necessary information may be set by extracting specific descriptions as described above, or by setting the location of the relevant information (address, path, etc.).

[0032] <Reference information setting section> The reference information setting unit 105 has a function of setting reference information that is information to be referred to when creating the document.

[0033] Reference information is the content of documents that are related or similar to the content of the document to be created. For example, when creating a building confirmation application, it is the content of a building confirmation application submitted when a similar building was constructed. Using past cases as reference information can serve as a reference for creating the current document. Reference information may be, for example, information such as data, files, documents, etc. held by this device, information stored in an external database, or information available on the web.

[0034] Furthermore, reference information can be set not only by specifying specific documents, etc., as described above, but also in an abstract manner. For example, it can be set in the form of "The contents of the building confirmation application submitted for a building similar to this project will be used as reference information." Furthermore, more detailed conditions can be added. For example, reference information can be set to include conditions such as "building confirmation application for an elevator" or "project to which application is made in XX City, XX Prefecture."

[0035] Also, a manual for creating documents may be set as reference information. For example, when creating a "building confirmation application," "how to write a building confirmation application" may be set as reference information.

[0036] The necessary information and reference information are both used for document creation, so the user does not need to recognize them as separate entities, and they may be configured to be set as an integrated entity.

[0037] Furthermore, a preferred embodiment of a database for storing various information and documents that can serve as reference information is a device that not only digitizes documents by scanning them, but also structures and stores and manages the content of each item contained in the document. For example, the content of each item in a document can be entered into a spreadsheet with rows and columns and stored. To structure the content of a document in this way, the data can be manually entered from paper documents into a spreadsheet, or scanned and digitized image data or PDFs can be converted into text using OCR (Optical Character Recognition) and then stored in a database. Such a document DB may be included in the document creation device, or it may be an external device that can exchange information with the document creation device.

[0038] The database may also be constructed by using the LLM to extract and structure information for individual items contained in documents of various formats. For example, to extract the "drawing number" contained in a document, a prompt such as "Please extract the value of 'Drawing Number'. 'Drawing Number' is a file ID and consists only of alphanumeric characters and hyphens. It conforms to the following regular expression: [AZ]-\d{5}" can be entered into the LLM as a prompt, and processing can be performed on numerous documents that have been scanned and digitized as image data or converted to text using OCR from PDFs. The "drawing number" information can be obtained as a response, allowing the construction of a document database. Note that the specific means for extracting and structuring information to construct a document database are not particularly limited. This can be achieved using any appropriate technology.

[0039] Building a database such as the one described above is a labor-intensive process, but the inventor of the present application has invented a data extraction technology using the LLM described above that reduces the labor required and makes the process more efficient, and has already filed a patent application for this technology (application number: PCT / JP2023 / 034488).

[0040] <Prompt generation part> The prompt generation unit 106 has a function of generating a prompt for creating the document in a large-scale language model using the specified template and the set required information. The prompt may also be generated using set reference information. For example, a sentence such as "Please create a 'construction plan outline' for the building confirmation application for the water supply and drainage system of XX. Search for information on water supply and drainage systems of projects similar to this project and refer to the information obtained as a result" may be generated. The prompt may be generated by the user inputting such a set of sentences, or the prompt may be generated by the user individually inputting the project name, the document to be created, the template to be used, the required information, and the reference information.

[0041] It is also possible to configure the LLM to add further prompts to instruct the LLM on more detailed conditions and settings for the document to be created. For example, prompts can be added regarding the maximum number of characters in the document to be created and conditions regarding the output format (text only, text with tables, including numerical data, etc.). It is also possible to configure the LLM to add prompts regarding whether to reformat or validate the created document, and if so, how to perform the reformat or validation. Reformatting and validation will be described later.

[0042] <Prompt accumulation section> The prompt storage unit 110 has a function of storing the generated prompts. Storing the prompts is convenient for the user because it allows them to create documents again or to modify and use them when creating documents for other projects. Furthermore, the prompts may be stored in association with a specified template and set required information and reference information.

[0043] <Document Creation Department> The document creation unit 107 has a function of inputting the generated prompt into the large-scale language model to create the document. Creating a document using an LLM may be performed, for example, as a function extension for a document creation application such as Word, or may be provided as a web application. Furthermore, it is not always necessary to create an entire document, but rather, it may also be necessary to create one or more items included in the document. It is also possible to create a document with multiple variations based on the instruction of a single prompt, and present these multiple documents for the user to select. It is also possible to configure the created document to be editable by the user.

[0044] It is also preferable to configure the system to show which parts of the set reference information were referenced in addition to showing the created document. This will show the path and basis for the created result, which will be useful when the user judges the merits of the created result.

[0045] <Reshaping section> The reformatting unit 108 has a function of reformatting the created document. Reformatting means reformatting the document created by the document creation unit according to a specified sample, format, etc. Reformatting may be performed using the LLM by including the instruction in a prompt, or by inputting a separate script that instructs reformatting.

[0046] <Verification Department> The verification unit 109 has a function of verifying the created document. Verification means checking whether the created document complies with predetermined conditions and specifications. Furthermore, it outputs the verified verification results as necessary and reflects the verification results in the re-creation of the document. For example, it performs verification such as verifying whether the document is similar to a specified negative sample, verifying whether the created document contains specific wording (e.g., the wording "building restrictions within setback"), and confirming the validity of the created document by comparing it with a specified sample.

[0047] If the verification result is not positive, for example, if the document resembles a specified negative sample, does not contain specific wording, or is otherwise inappropriate, it is preferable to display such a result as a message. Furthermore, by inputting instructions for recreating the document along with the message into the LLM and having the LLM perform the creation process, it is possible to create a more desirable document that reflects the verification result.

[0048] The verification by the verification unit may be performed before or after the reshaping by the reshaping unit described above. Also, like the reshaping, the verification may be performed using the LLM by including the instruction in the prompt, or by inputting a separate script that instructs the verification.

[0049] <Hardware configuration> 3 is a conceptual diagram showing an example of the configuration of hardware that realizes a document production device according to an embodiment. As shown in the figure, document production device 300 includes CPU 301 that performs various arithmetic processing, RAM 302 that is a volatile storage medium, storage 303 that is a non-volatile storage medium such as a flash memory or HDD, communication interface 304, and input / output interface 305. RAM 302 reads programs that perform various arithmetic processing to cause CPU 301 to execute the programs, and also provides a work area for the programs. In addition, RAM 302 is assigned multiple addresses, and programs executed by CPU 301 can exchange data and perform processing by identifying and accessing the addresses (the same applies throughout this specification).

[0050] Here, the functions of template setting unit 101, template specification unit 103, necessary information setting unit 104, reference information setting unit 105, prompt generation unit 106, document creation unit 107, reshaping unit 108, and verification unit 109 of document creation device 100 in Fig. 1 are mainly realized by CPU 301 and RAM 302 in Fig. 3. Furthermore, the functions of template storage unit 102 and prompt storage unit 110 are mainly realized by storage 303 in Fig. 3. Furthermore, when the document DB is external or when document creation unit 105 uses an external LLM, each function is realized by mutual exchange of signals and information via communication interface 304 and input / output interface 305.

[0051] <Processing flow> FIG. 4 is a flow diagram illustrating a simplified example of the processing flow of a document creation device. First, description items and description guidelines are set as a template for document creation (S401: template setting step). Then, the set template is stored (S402: template setting step). Then, a template to be used for document creation is specified from the stored templates (S403: template specification step). Then, necessary information, which is information required to create a document using the specified template, is set (S404: necessary information setting step). Then, reference information, which is information referenced for document creation (S405: reference information setting step). Then, a prompt is generated to cause a large-scale language model to create the document using the specified template, the set necessary information, and the set reference information (S406: prompt generation step). Then, the generated prompt is input into the large-scale language model to create a document (S407: document creation step). Then, the created document is reformatted (S408: reformat step). Then, the created document is verified (S409: verification step). The verification step may be performed before the reshaping step. Furthermore, although not shown, the method may include a prompt storage step for storing the prompts generated in the prompt generation step. Furthermore, the method may not include the reference information setting step.

[0052] <Effects> According to the document creation device of this embodiment, a desired document can be created efficiently by using a template. [Explanation of symbols]

[0053] 100: Document creation device 101: Template setting section 102: Template storage unit 103: Template specification section 104: Required information setting section 105: Reference information setting section 106: Prompt generation unit 107: Document Preparation Department 108:Reshaping department 109: Verification Department 110: Prompt storage unit 301:CPU 302:RAM 303: Storage 304: Communication interface 305: Input / output interface

Claims

1. a template setting section for setting a template for document creation; a template storage unit that stores the set template; a template designation unit that designates a template to be used for document creation from the stored templates; a necessary information setting unit that sets necessary information that is information required to create a document using the specified template; a prompt generation unit that generates a prompt for causing a large-scale language model to create the document using the specified template and the set necessary information; a document creation unit that inputs the generated prompts into the large-scale language model to create the document; A document creation device having the following.

2. a reference information setting unit that sets reference information that is information to be referenced when creating the document; The document creation device according to claim 1 , wherein the prompt generation unit further generates the prompt using the set reference information.

3. 3. The document creation device according to claim 1, further comprising a reformatting unit that reformats the created document.

4. 3. The document creation device according to claim 1, further comprising a verification unit that verifies the created document.

5. 3. The document creation device according to claim 1, further comprising a prompt storage unit that stores the generated prompts.

6. A document creation method executed by a computer in a document creation device, comprising: A template setting step to set up a template for document creation; a template storage step of storing the set template; a template designation step of designating a template to be used for document creation from the stored templates; a necessary information setting step of setting necessary information that is information necessary for creating a document using the specified template; a prompt generation step of generating a prompt for generating the document into a large-scale language model using the specified template and the set necessary information; a document creation step of inputting the generated prompts into the large-scale language model to create the document; A document preparation method having the following.

7. In the document creation device, A template setting step to set up a template for document creation; a template storage step of storing the set template; a template designation step of designating a template to be used for document creation from the stored templates; a necessary information setting step of setting necessary information that is information necessary for creating a document using the specified template; a prompt generation step of generating a prompt for generating the document into a large-scale language model using the specified template and the set necessary information; a document creation step of inputting the generated prompts into the large-scale language model to create the document; A document creation program that causes a computer to execute the following.

Images