Document output method

Abstract

Medical documents such as medical certificates and referral letters, which are created on a daily basis, are automatically generated without any workload. [Solution] This invention combines OCR and a large-scale language model (LLM) to identify the document type and assign the extracted text information to the appropriate field. The text information acquired by OCR is compared with a database, and the LLM analyzes the context and places it in the appropriate field (listed on the right side of the screen). Furthermore, a PDF template is used to automatically generate medical documents in a unified format (displayed on the left side of the screen). The listed text information can be placed in the appropriate field by user operation.

Description

[Technical Field]

[0001] The present invention relates to an automatic document generation tool (document output method, document output system, document output program) for medical documents in particular, which utilizes OCR (optical character recognition) and LLM (large-scale language model). [Background technology]

[0002] In medical settings, many medical documents such as medical records, referral letters, and medical certificates are created on a daily basis. These documents not only record the details of a patient's medical treatment, but are also used to share information between medical institutions and as proof for insurance claims, so they must be created accurately and quickly.

[0003] In recent years, the spread of electronic health records (EHRs) has led to the digitization of medical information, but many paper-based documents still exist in practice. For example, if a patient brings in a referral letter from another hospital, if the document is paper, medical staff either manually enter the details or scan it and save it as a PDF. However, scanned PDFs are treated as image data and are therefore not suitable for searching or editing. In addition, patients may bring in copies of past test results or prescriptions, making the task of integrating this information into the medical system time-consuming.

[0004] In this context, OCR (optical character recognition) technology can be used to extract text data from scanned paper documents. However, OCR alone only recognizes characters, and it is difficult to automatically determine which items the content corresponds to. For example, additional processing is required to assign information such as "name" and "date of birth" in a referral letter to the appropriate fields in the electronic medical record.

[0005] Additionally, medical documents often have different formats depending on the hospital and region, making standardized processing difficult. While one hospital's medical certificate may list "Name," another may use "Name" or "Patient Name." While a human can make a uniform judgment based on context, a system needs to set uniform rules.

[0006] Patent Document 1 discloses a system that uses OCR technology to recognize characters and numbers in documents and analyzes and displays data based on their location information. Specifically, the functions of the object detection modeling unit (130') and form configuration modeling unit (150') enable the system to identify input items and match appropriate information to those items. Because location information is stored in a database, documents that do not exist in the database are not applicable. Patent Document 1 also simplifies the OCR process by comparing it with form patterns, specializing in the recognition of only numbers. [Prior art documents] [Patent documents]

[0007] [Patent Document 1] Patent No. 7444495 (specification

[0058]

[0140]

[0174]

[0176] etc.) Summary of the Invention [Problem to be solved by the invention]

[0008] However, the technology in Patent Document 1 only matches items by comparing character recognition using OCR with existing form patterns in a database, and lacks the flexibility to adapt to new document formats, making it unable to handle documents with formats not present in the database. Furthermore, a system that identifies appropriate items based on the context of text data acquired using OCR and automatically classifies and assigns them to other documents would be useful in promoting business efficiency. In particular, in the medical industry, advances in medical technology mean that formats are updated daily, necessitating minor data revisions. These linkages must be updated by non-engineers. [Means for solving the problem]

[0009] The present invention has been made in consideration of these issues, and aims to provide a document output method that identifies the document type and accurately assigns the character information contained in the document to appropriate items as a new medical document by focusing on a technology that combines text extraction using OCR and context analysis using a large-scale language model (LLM). Note that "output" in this invention has a broad meaning, including document generation, editing, final conversion to PDF, electronic output, etc.

[0010] Furthermore, the present invention has the following advantages. <Dramatically improves efficiency of document creation flow> By utilizing OCR analysis and large-scale language models, the burden of manual input and copy-paste is greatly reduced.

[0011] <Improving the interpretation performance of large-scale language models> The document type is identified, and information based on the item names (standardized items) that are often found in that document type is extracted using a large-scale language model, which makes it possible to improve the performance of the large-scale language model. <Preventing input errors and improving information accuracy> By using restriction information linked to item names and a Yes / No conversion function, document formats and expressions are standardized, reducing the occurrence of human error. <Create a new document by combining Document 1 and Document 2> The task of manually extracting necessary information from multiple documents can now be automated, improving work efficiency and convenience.

[0012] (1) A method is provided which includes a database storing document types and a list of item names corresponding to the document types, and which executes the steps of: acquiring a list of input areas from a document through optical recognition; acquiring recognized character information through analysis using optical character recognition; and automatically or manually identifying the document type of the document; after acquiring a list of item names corresponding to the identified document type from the database, the method includes a step of assigning the acquired recognized character information to the item names by a program; and a step of identifying input areas corresponding to the item names from the list of input areas, and outputs a template document.

[0013] That is, (1) is a document output method executed by a computer having a database storing document types and item names corresponding to the document types, Obtaining a list of input areas from a document through optical recognition Obtaining recognized character information by analyzing through optical character recognition. A step of automatically or manually identifying the document type of the document. A step of obtaining a list of item names corresponding to the identified document type from the database. A step of assigning the acquired recognized character information in correspondence with the item name by a program. Step of identifying an input area corresponding to an item name from the input area list Steps for outputting template documents The document output method is characterized by comprising the steps of:

[0014] This invention utilizes optical character recognition and optical recognition technologies to extract character information from paper documents and image data, identify the document type, and obtain a list of input fields through optical recognition. Because the extracted character information may contain a variety of expressions, such as "name," "first name," and "given name," the system references a pre-prepared database to obtain a list of item names corresponding to each document type. This database contains a wide variety of item names used by different medical institutions and departments, enabling flexible response without relying on simple pattern matching. This configuration enables accurate classification and allocation of input information when outputting documents, significantly reducing the effort required for manual correction. In addition, by associating optimal item names with each document type, flexible operation becomes possible without relying on conventional fixed form patterns, contributing to improved efficiency and reliability of document management.

[0015] (2) A method for automatically or manually identifying the document type of a document 1, comprising: a database storing document types and a list of item names corresponding to the document types; a step of acquiring recognized character information of the document 1 by analyzing the document 1 through optical character recognition; a step of acquiring a list of input areas through optical recognition; and a step of automatically or manually identifying the document type of the document 1; and a step of assigning the acquired recognized character information of the document 1 to the item names by a large-scale language model after the step of acquiring a list of item names corresponding to the document type of the identified document 1 from the database; a step of identifying document 2 from the database, outputting a template document, and acquiring recognized character information of document 2 by analyzing document 2 through optical character recognition, and a step of automatically or manually identifying the document type of document 2; after acquiring a list of item names corresponding to the identified document type of document 2 from the database, a step of assigning the acquired recognized character information of document 2 to the item names and values ​​associated with the item names using a large-scale language model; inputting the values ​​associated with the item names of document 2 into input fields corresponding to the item names in the template document of document 1, and outputting a new document.

[0016] That is, (2) includes, in addition to the method described in (1), a step of processing Document 1 and Document 2 and outputting a new document. A step of obtaining recognized character information of Document 1 and outputting a template document. A step of obtaining recognized character information of document 2 and identifying item names and values ​​of document 2; The method includes a step of inputting a value associated with the item name of document 2 into a corresponding input field of the template document of document 1.

[0017] In this invention, the document type is identified as in (1), and a process of extracting and assigning recognized character information using optical recognition and optical character recognition is performed, while in (2), a large-scale language model (LLM) analyzes the recognized character information extracted by OCR and finds the optimal correspondence among the obtained item names based on the context and meaning. In other words, the LLM automatically determines which item the character string obtained by OCR corresponds to, thereby understanding the meaning of the character string and accurately assigning the character string. This enables document generation that maintains consistency and accuracy even for documents that contain a mixture of expressions and formats, which was difficult with conventional systems. Furthermore, the recognized character information obtained from document 1 is organized as a template document, and the information from document 2 is integrated into the template of document 1 to generate a new document. Furthermore, the present invention can clearly associate not only item names but also their associated values ​​(specific data).

[0018] (3) At least one of recognized character information of document 2 extracted and obtained by optical character recognition, character information obtained from a data warehouse in which electronic medical record information is stored, or recognized character information extracted and obtained by optical character recognition from screen capture data of an electronic medical record. That is, (3) the character information of document 2 is at least one of information obtained by OCR recognition, information obtained from an electronic medical record data warehouse, and information obtained by analyzing screen capture data of an electronic medical record using OCR.

[0019] The present invention can acquire the character information necessary for generating a document from various sources, or from multiple sources without relying on a single source. For example, by utilizing one or more of various input sources, such as text character information extracted from paper or electronic documents or images using OCR technology, character information obtained from a data warehouse storing information such as patient profiles stored in electronic medical records, or character information extracted from screen captures of electronic medical records (and further character information obtained by OCR processing of the screen captures), the necessary character information can be acquired thoroughly and efficiently.

[0020] (4) The recognized character information is generated or selected and assigned to a PDF template that defines the configuration items and layout of the document. That is, (4) assigns the recognized character information to a PDF template that defines the layout of the item name list and input area list corresponding to the document type.

[0021] This invention uses PDF templates that predefine document components and layouts, automatically generating and selecting recognized character information obtained through data acquisition techniques such as OCR, and accurately assigning it to each item based on the template. This unifies the format of the entire document and places the necessary information in the designated locations, reducing the risk of manual layout adjustments and incorrect placement. Furthermore, processing according to templates improves document creation speed, ensures consistency and accuracy even for diverse documents such as medical documents, and ultimately produces output documents with a clear, organized layout.

[0022] (5) The value associated with the item name of document 2 is associated with the item name by a user's operation and / or correction input.

[0023] This invention employs a system that allows users to directly intervene in text information extracted by methods such as OCR, allowing users to check and adjust the text information themselves and assign it to the most appropriate item through intuitive operations such as drag and drop, and by entering corrections as needed.Even if misrecognition or inconsistencies occur that cannot be handled by automatic processing alone, the system provides an interface that allows users to easily make corrections, ensuring the accuracy and consistency of the final medical document.

[0024] (6) In cooperation with a workflow management system, the approval flow of document 2 generated based on the PDF template is managed. (7) The item name list stores one or more of the item name, value type, and value restriction information, and when a value associated with the item name is entered in the input area of ​​the document, the value type or the value restriction information is checked, and if it does not match, a warning message is displayed on the screen or the color of the text changes, and the item name list and value type or value restriction information corresponding to the document type are obtained from an electronic medical record or are matched with the electronic medical record. (8) A document output method characterized in that the output method can be set in association with the document template or the item name, and when the value type is a Yes / No selection type value, an expression format including a circle, a cross, or a check is selected when outputting the value associated with the item name of document 2. (9) The PDF template can be shared among multiple facilities or departments.

[0025] This invention integrates medical documents automatically generated based on PDF templates with a workflow management system. This not only enables the generated medical documents to be output as is, but also automates the approval process within the organization, allowing them to be efficiently circulated and reviewed by relevant parties. Users can grasp the document approval status in real time and quickly provide feedback and corrections as needed, significantly reducing the time required compared to traditional manual management. Furthermore, PDF templates are pre-defined with patient profile information (personal information such as name, date of birth, gender, address, telephone number, emergency contact, insurance information, and blood type), and patient information obtained from electronic medical records and other sources is automatically updated. This automatic update function prevents manual input errors. Furthermore, PDF templates are designed to be shared across multiple facilities and departments, eliminating the need for individual templates for each medical institution or department, allowing documents to be generated in a unified format.

[0026] (8) specifies a mechanism for flexibly setting output methods linked to document templates and item names. Specifically, different output formats can be applied depending on each item name and its value. In particular, for Yes / No selection type values, the system converts them into representation formats such as a circle (○), a cross (×), or a check mark for selection. This function enables Yes / No binary information in medical documents and reports to be output as symbols adapted to the document output format, rather than as simple text. [Effects of the Invention]

[0027] The document generation tool of this invention utilizes OCR and large-scale language models to accurately extract and analyze text information from multiple sources, such as paper and electronic medical records, and automatically assigns it to the appropriate items according to the document type.Furthermore, because it generates medical documents using predefined PDF templates, it can output medical documents with a unified layout that are easy to read. [Brief explanation of the drawings]

[0028] [Figure 1] 1 is a diagram showing a network configuration of a document output system according to an embodiment of the present invention. [Figure 2] 1 is a diagram showing a system configuration and hardware layout of a document output system according to an embodiment of the present invention. [Figure 3] FIG. 2 is a block diagram showing an example of the functional configuration of the server 20. [Figure 4] 10 is a flowchart showing the processing procedure (preprocessing) of a document generation program. [Figure 5] 10 is a flowchart showing the processing procedure (post-processing) of the document generation program. [Figure 6] 10 is an example of a user display screen. [Figure 7] FIG. 10 is a diagram for explaining the relationship between input areas and item names. [Figure 8] FIG. 10 is a diagram for explaining item names acquired through OCR processing. [Figure 9] FIG. 10 is a diagram for explaining the functions of visualizing and formatting text information. [Figure 10] A diagram showing the relationship between a letter of introduction and its OCR analysis results. [Figure 11] An explanatory diagram showing how to flexibly set the output method. BEST MODE FOR CARRYING OUT THE INVENTION

[0029] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0030] Figure 1 shows a network configuration centered on a document output system 2. User terminals 1 (e.g., PCs or tablet devices) and hospital medical systems 3 (e.g., electronic medical record systems, workflow management systems, and operation management systems) are connected via a communication network, such as the Internet, intranet, or hospital LAN, via telecommunications lines. This allows document generation requests from user terminals 1 to be sent to document output system 2, which then retrieves the necessary information. Data sequentially input and output to medical system 3 is stored in a data warehouse (DWH) 4. By centrally managing large volumes of data collected from hospital electronic medical records and other systems, DWH 4 can eliminate format inconsistencies between different systems and ensure data consistency. For example, electronic medical record systems record a large amount of data, such as patient information (personal information and profile information), medical history, test results, and prescription information. This information is typically managed in different systems and formats. Using a DWH, automated document generation is possible through real-time or batch processing. This makes it possible to integrate and consolidate such data into a single database, and the DWH4 is not just a data storage location; it also performs data modeling and index management to streamline the analysis and search of collected data, allowing the document output system 2 to quickly and accurately obtain the necessary information.

[0031] Data in the DWH can be combined with information extracted using screen captures of electronic medical records and OCR technology, allowing for centralized management of data in a variety of formats. In addition to traditional text-based data, image data and scanned paper information can also be effectively utilized. For example, by converting handwritten medical notes recorded by doctors into text using OCR technology and storing them in the DWH4, past medical treatment details can be easily searched and used in conjunction with other systems. In this way, this system automates and streamlines the document generation process, and by integrating and utilizing different data sources, it can generate more accurate and comprehensive documents.

[0032] 2 is a diagram showing the system configuration and hardware layout of a document output system, in which a server 20 and a user terminal 10 of the document output system are connected via a network (a wired / wireless network 80 such as the Internet). The computer of the server 20 includes a communication IF 22, an input / output IF 23, a memory 25, a storage 26, and a processor 29. The user terminal 10 includes a communication IF 12, an input device 13, an output device 14, a memory 15, a storage unit 16, and a processor 19. The user terminal 10 is a desktop or laptop PC, or a portable terminal such as a tablet or smartphone.

[0033] <server> The communication IF 22 is an interface for inputting and outputting signals so that the server 20 can communicate with external devices. The input / output IF 23 functions as an interface with an input device and an output device for receiving input operations from a user. The memory 25 is for temporarily storing programs and data processed by the programs, and is a volatile memory such as a DRAM. The storage 26 is a storage device for saving data, and is a flash memory, HDD, or SDD, for example. The processor 29 is hardware for executing an instruction set written in a program, and is composed of an arithmetic unit, registers, peripheral circuits, etc.

[0034] <User device> The communication IF 12 is an interface through which the user terminal 10 communicates with external devices to input and output signals. The input device 13 is an input device (such as a keyboard, a touch panel, a touch pad, or a pointing device such as a mouse) for receiving input operations from the user. The output device 14 is an output device (such as a display or a speaker) for presenting information to the user. The memory 15 is for temporarily storing programs, data processed by the programs, etc., and is, for example, a volatile memory such as a DRAM. The storage unit 16 is a storage device for saving data, such as a flash memory or a HDD. The processor 19 is hardware for executing an instruction set written in a program, and is composed of an arithmetic unit, registers, peripheral circuits, etc.

[0035] <Specific server configuration> 3 is a block diagram showing an example of the functional configuration of server 20. Server 20 includes communication means 220, input device 230, display device 240, notification device 250, storage means 280, and control means 290, and these components are electrically connected via a bus. In addition, storage means 280 includes a data storage area 283.

[0036] The communication means 220 performs modulation / demodulation processing and the like to enable the server 20 to communicate with the user terminal 10, performs transmission processing on signals calculated by the control means 290, and transmits the signals to external devices and equipment. The communication means 220 also performs reception processing on signals received from the outside and outputs them to the control means 290. In this way, the communication means 220 interprets commands or input contents and provides them to each means, and also functions as an interface that interprets various display commands issued from the storage means 280 and controls output.

[0037] The input device 230 is a device through which an administrator who operates the server 20 inputs instructions or information as needed, and may be a keyboard, mouse, reader, or touch-sensitive device. The input device 230 converts instructions input by the administrator into electrical signals and outputs the electrical signals to the control means 290. The input device 230 also includes a receiving port that receives electrical signals input from an external input device.

[0038] The display device 240 is a device having a display 241 such as an LCD or organic EL for presenting information to an administrator who operates the server 20 as needed. The display 241 can display data according to the control content of the control means 290, and can confirm the communication status between the server 20 and other external devices 10. The notification device 250 is a device (such as a speaker 251) for presenting information to the administrator.

[0039] The storage means 280 is realized by a memory (RAM) 25 and storage 26 such as a disk device (floppy disk, hard disk, magneto-optical disk, etc.), and stores data, programs, etc. used by the server 20. The storage means 280 stores an application program 282 of the present system as well as data in a work area 281, a data storage area 283, and a screen definition storage area 284.

[0040] The work area 281 is allocated when the system is started up, and is an area in which various data input and output in the system are temporarily stored. The data storage area 283 is an area in which data temporarily stored in the work area 281 is semi-permanently stored through writing control by the data management means 285 when a save request is made for the data temporarily stored in the work area 281. The screen definition area 284 is an area in which screen definition information for various screens to be output and displayed on the user terminal 10 is stored in advance, and includes format information for screen settings to be displayed by the display control means 292.

[0041] The data storage area 283 is an area for integrated management of various data in the document output system, and holds databases such as an item name list DB 283A and a template DB 283B. The data storage area 283 not only functions as a simple storage, but can also operate as a data processing means for optimizing document generation processing and performing data preprocessing, and by linking with the data warehouse 4, it can also handle the processing of large-scale medical data.

[0042] The item name list DB283A is a database that supports document generation, storing the names and related information of various data items used in medical systems and document output systems, realizing unified data management across the entire system and ensuring consistency of information between different systems. For example, if electronic medical record systems and workflow management systems each have their own unique data items, data exchange and integration processing between different systems will be difficult if the item names are not unified. Therefore, this database manages standardized data items used across the entire system, allowing each system to refer to the same names and definitions.

[0043] Specifically, the database stores a list of various document types and the item names contained in each document. For example, a "medical record" document type includes items such as basic patient information (name, date of birth, gender, and date of consultation), medical details (chief complaint, examination findings, diagnosis, and treatment plan), and prescription information (drug name, dosage, and medication instructions). Meanwhile, a "lab report" stores a list of item names, including not only patient information but also the type of examination (blood test, MRI, CT scan, etc.), measurement results (blood glucose level, hemoglobin level, and imaging findings), examination date, and the name of the attending physician. A document such as a "discharge summary" stores comprehensive information about a patient's hospitalization, including the diagnosis at the time of admission, length of hospitalization, treatment progress, whether surgery was performed, medications used, and post-discharge follow-up plans. Because the required items differ depending on the document type, the item name list DB 283A appropriately manages the items corresponding to each document and accurately reflects the required information when generating the document. Furthermore, documents such as "letters of referral" and "medical certificates" contain information such as the patient's medical history, current symptoms, diagnostic results, progress of treatment, and information about the medical institution to which the patient is referred. Although the formats of these documents may differ depending on the specific medical institution, by utilizing the item name list DB283A, it is possible to manage standardized items across different formats and ensure information compatibility. The item name list DB283A stores the item names of existing OCR documents, the item names of learned documents, and the item names of the medical document to be created.

[0044] The template DB283B stores templates for various documents used in the document output system. It provides a unified format for each different document type, improving efficiency and standardizing document creation. Additionally, the template DB283B stores template information such as fonts, layouts, table placement, header and footer information, and fixed wording within documents. This allows documents to be displayed and printed in a unified format even when exchanged between different systems or different medical institutions. For example, a medical certificate created at one hospital can be submitted to another medical institution or insurance agency with a consistent format.

[0045] A PDF template designed to be shareable across multiple facilities and departments is one that separates customizable and fixed sections, with the fixed sections (layout, legal text) immutable and the variable sections (patient information, diagnosis details) editable. To ensure compatibility across different facilities, industry-standard formats such as HL7 FHIR, PDF / A, and XML-based templates (XFA) are used. Cloud-based template management and API access and updating of templates allow for consistent, up-to-date templates.

[0046] For example, formats for medical documents such as medical records, test reports, referral letters, and medical certificates are predefined, and the document output system automatically inputs data to create medical records in a unified format. PDF templates, which define the components and layout of medical documents, are used by document output systems to create various medical documents. They serve as the basis for outputting documents such as medical records, test reports, medical certificates, referral letters, and discharge summaries in a unified format. PDF templates contain the necessary fields for each document type. For example, a medical certificate layout includes basic patient information (name, date of birth, and gender), diagnosis, treatment history, prescription information, and a physician's signature field. PDF templates also clearly separate fixed and variable sections. Fixed elements such as the hospital name, logo, document title, description, electronic signature, and electronic seal always appear in the same format, while individual patient diagnosis results and treatment information vary depending on the data retrieved from the system. This structure enables documents to be generated with a consistent layout even for different patients and treatments.

[0047] PDF templates are designed based on PDF data structure technology and have a blueprint for a structure that combines a fixed layout with the ability to embed dynamic data. They automatically input patient information and diagnosis results obtained from electronic medical record systems and data warehouses using form functions (AcroForm) and XML-based Extensible Form Architecture (XFA). Note that PDF templates can be any template that has the functionality to control print formatting and document layout, and automate data entry. For example, they can be PDF forms, AcroForms, XFA (XML Forms Architecture), or fixed layout PDFs.

[0048] The control means 290 realizes various processes of the input means 291, display control means 292, OCR processing means 293, LLM processing means 294, user IF 295, and management means 296 using a processor. The control means 290 uses the processor to oversee the processing of the entire system and control various processes related to input, search, display, high-speed processing, and user interface (UI). Note that the processor may be one or more processors. The at least one processor is typically a microprocessor such as a CPU (Central Processing Unit), but may also be other types of processors such as a GPU (Graphics Processing Unit). The at least one processor may be single-core or multi-core. Furthermore, the at least one processor may also be a processor in a broad sense, such as a hardware circuit (e.g., an FPGA (Field-Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit)) that performs part or all of the processing.

[0049] The input means 291 analyzes the contents of the user's input via a keyboard, touch screen, or the like, and accepts the data. When the patient's symptoms and diagnosis results are entered using a keyboard, the input means 291 accepts text data. When voice input is possible, the diagnosis dictated by the doctor is analyzed by a voice recognition engine, converted into text, and the data is accepted. When a touch screen is used, selection operations using drop-down menus and check boxes are also included.

[0050] The display control means 292 controls the visual presentation method through the user interface. The display control means 292 provides appropriate information presentation and interaction functions to facilitate the process of creating, checking, and correcting medical documents. Specific control examples include real-time input feedback, correction support, information highlighting, data visualization, and function allocation.

[0051] For example, when doctors or administrative staff operate electronic medical records or document output systems, real-time input feedback is provided so that the information they enter is immediately reflected on the screen. This allows users to proceed with their work while visually checking the contents of their input. If an input error is detected, the system controls the system by changing the color of the input area or displaying a pop-up error message, helping the user to correct the error immediately. In addition, when corrections are required, the system works in conjunction with the OCR processing means 293 and LLM processing means 294 to provide a function to present recommended correction candidates. For example, if the characters read by OCR are ambiguous, the system highlights the part and designs an interface that allows the user to select the appropriate candidate.

[0052] The OCR processing means 293 is composed of an acquisition unit 293A, a registration unit 293B, and a template registration unit 293C. The LLM processing means 294 is composed of an assignment unit 294A and a template registration unit 294B. The cooperation between the OCR processing means 293 and the LLM processing means 294 enables the automation of medical document creation. The OCR processing means 293 accurately acquires character information, and the LLM processing means 294 organizes it into an appropriate document structure, thereby streamlining the creation of medical documents. In particular, it learns the correction history of misrecognitions, leading to improved accuracy in future generations, and by coordinating with document templates, it achieves format standardization.

[0053] The acquisition unit 293A extracts text information from paper medical certificates, screen capture images, etc. For example, it analyzes scanned image data and acquires the patient name, diagnosis results, doctor's findings, etc. contained therein as text data. It uses optical character recognition (OCR) technology to recognize not only printed characters but also handwritten characters.

[0054] The registration unit 293B registers the character information recognized by the acquisition unit 293A in a database in an appropriate format. For example, it checks whether the text data after OCR processing correctly corresponds to each item in the medical document (such as name, diagnosis, and test results). During registration, to ensure data accuracy, the data undergoes format conversion, normalization, and classification by item, ensuring consistency with the electronic medical record and DWH data, thereby maintaining consistency throughout the system.

[0055] The analysis unit 293C performs detailed analysis of the character data acquired through OCR processing, identifies possible misrecognition parts, and checks the consistency of the data. For example, it identifies characters suspected of being misrecognized (ambiguous parts of handwritten characters or mistakes in medical terminology) and provides an interface to prompt the user to correct them. It also supports more accurate text conversion by comparing them with past input data and referring to the correction history of similar data.

[0056] The assignment unit 294A classifies and organizes the text information acquired and analyzed by the OCR processing means 293 into an appropriate document structure. For example, data read by OCR, such as "Blood Test Results: HbA1c 7.8," is assigned to the "Test Results" section, and information such as "Diagnosis: Diabetes" is automatically classified into the "Diagnosis" field. The assignment unit 294A also performs classification based on context. For example, if the diagnosis information of "high blood pressure" is recorded, treatment plan and medication information is extracted from the surrounding information and mapped to an appropriate format. Utilizing the natural language processing (NLP) function of LLM (Large Scale Language Model), the assignment unit 294A not only recognizes the characters in the OCR data but also understands their meaning and performs classification. The assignment unit 294A provides a function to automatically assign patient profile information and adopt a format that can be shared across multiple facilities and departments, allowing a unified document format to be used in different environments.

[0057] The management means 296 manages the approval process of the generated medical documents in cooperation with the workflow management system.

[0058] 4 and 5 are flowcharts showing the processing procedure of the document generation program, which is executed by the application program of the document output system according to the embodiment of the present invention. Fig. 4 shows pre-processing, and Fig. 5 shows post-processing. In the pre-processing flow, OCR is performed on screen captures of medical certificates and electronic medical records to extract text information. The acquired data then proceeds to a document type identification step, where the appropriate document type is determined through analysis by a large-scale language model (LLM), and the extracted text information is saved in a database and passed on to the post-processing flow. In the post-processing flow, an appropriate PDF template is retrieved based on the saved text information, and document data is embedded into it. The consistency of the embedded data is checked by LLM, and the appropriateness of the content is determined. If inappropriate data is detected, the user is asked to correct it, and the data is re-embedded. Once the document is confirmed as appropriate, the medical document is saved in a database and approval processing is carried out in conjunction with the workflow management system. The final document is then output, ready for printing or electronic transmission.

[0059] 4, first, the input means 291 accepts data acquired through the keyboard, touch screen, scanner, or camera of the user terminal (step S401), and the acquired data is sent to the OCR processing means 293. In the OCR processing (step S402) of the OCR processing means 293, character information is extracted from the scanned image or captured data, and the character information is stored as text data in the registration unit 293B.

[0060] Next, a process for identifying the document type is performed (step S403). The character information acquired by OCR is analyzed to identify the document type (medical certificate, referral letter, test report, etc.). Next, the LLM processing means 294 automatically determines whether the document type is appropriate (step S404). After the automatic determination, the display control means 292 may request feedback from the user as to whether the document type is appropriate or not. The document type is identified based on the title of the OCR document or the item names described in the OCR document.

[0061] If the document type is determined to be appropriate in step S404, the process proceeds to the next step (step S406); if it is not appropriate, the process proceeds to step S405. In step S405, if the document type is incorrect, the user manually specifies the correct document type and corrects it, and then re-executes the steps S403 to S404. In step S406, the character information acquired by OCR is saved for each appropriate document type. The registered character information is used for document generation in post-processing.

[0062] 5, the template acquisition unit 294B acquires a PDF template corresponding to the document type from a database (step S501). Also, character information acquired from the OCR document saved in the preprocessing flow is acquired (step S502). Next, the template acquisition unit 294B applies the document layout, and the allocation unit 294A embeds the acquired character information in an appropriate position in the PDF template (step S503).

[0063] Next, the LLM processing means 294 determines whether the content of the document is correct or appropriate (step S504). If it is appropriate, the process proceeds to step S506. If it is inappropriate, the process proceeds to step S505. If it is inappropriate, the user manually corrects it and re-embeds the corrected content (S503). If it is appropriate, the completed medical document is saved in a database (step S506), and the saved medical document is generated in PDF format and subjected to output processing such as printing on paper or electronic transmission (step S507).

[0064] Before the completed medical document is output, an approval flow may be interrupted. This approval process works in conjunction with the workflow management system, and the doctor in charge or a manager will approve the medical document after checking its contents. Once approval is complete, the final medical document is output and can be printed or sent electronically. If the approval flow determines that corrections are necessary, the data will be requested to be corrected again, and the document generation process may be partially reset.

[0065] The example user display screen in Figure 6 is an interface screen for OCR scanning a medical information report from another hospital and creating a medical document using the text information. The scanned data of the source document can be displayed on the screen, or the source document and the medical document to be generated can be displayed by switching between them.

[0066] The text information extracted by OCR is listed in a visible format in the right column of the screen. The user can directly edit the text information on the right and make corrections. The listed text information may be all of the information obtained from the OCR document, or it may be a partial extraction. When extracting a partial information, the information can be extracted by referencing a database based on the item names required for the medical document to be generated. Furthermore, the listed text information may be the text information contained in the OCR document itself, or it may be converted to text information suitable for the medical document to be generated. For example, if the OCR document is in Gregorian calendar format and the medical document to be generated is in Japanese calendar format, the Gregorian calendar year is converted to the Japanese calendar year and listed. Alternatively, the list may be in Gregorian calendar format and converted to Japanese calendar year format when the user operates the system.

[0067] Meanwhile, the medical document being created is displayed on the left side of the screen, and text information on the right side can be assigned to the appropriate field by dragging and dropping. This function reduces the burden of manual input and allows users to intuitively arrange necessary information. For example, important information such as diagnosis, test results, and findings can be dragged to the appropriate field in the appropriate template, allowing for quick creation of formatted documents. In addition, collaboration with LLM (large-scale language model) analyzes the meaning of text information extracted by OCR and supports automatic assignment to the appropriate field. Potentially incorrectly recognized text is highlighted to assist users in correcting the text. When generating text using drag-and-drop, the assignment unit 294A recognizes the text as appropriate by dragging it to the field corresponding to the original text information and assigns it accordingly. Even if the drag operation is attempted to assign an unrelated text, the assignment unit 294A recognizes the text as inappropriate and does not assign it. For example, when assigning date information, the appropriateness of the field name is determined by whether it is the date of birth or the date of consultation / onset, rather than relying solely on the date information, preventing incorrect input.

[0068] To improve the matching accuracy of drag-and-drop operations, data categories are assigned to the input source (the information to be dragged) and output destination (the information to be dropped), and dropping is only possible if they match. In ambiguous cases, LLM can analyze the context and correct them. Proper matching can be achieved by setting "acceptance rules" (e.g., patient profile only) and "format information" (e.g., data conversion rules such as "convert from the Gregorian calendar to the Japanese calendar" or "comma-separated to space-separated") for the output destination. Furthermore, rule-based consistency checks can be combined with LLM's context analysis. [Industrial Applicability]

[0069] The present invention is useful in that it enables the automation of documents frequently created in medical settings, such as medical certificates, and absorbs differences in document formats between different medical institutions, enabling document creation in a unified format. Furthermore, by utilizing the PDF template function of the present invention, it is possible to standardize document formats and output electronic documents in a unified layout. It is useful not only for medical documents but also for prescription work, and as a tool for streamlining document management and reducing workloads in fields other than medical care. [Example]

[0070] Examples of this system in the medical field include the following clinical settings: The present invention is particularly useful when submission is required by the national government, but the formats differ between local governments and hospitals. - Prepare an infectious disease notification form based on electronic medical record data and submit it to the local public health center or administrative agency. ·Create test result reports from information in electronic medical records and testing systems. · Prepare a letter of introduction. · Fill in the necessary information in the document template commonly used by the admission and discharge support center. · Prepare surgery application forms / hospitalization application forms, etc. The present invention is applicable not only to the medical field, but also to a wide range of official documents that may differ from region to region, including birth registrations, tax returns, etc.

[0071] (2) Flow of creating and submitting an infectious disease notification Document type determination and template creation for infectious disease notification: A PDF equivalent to the official format for an "infectious disease notification" is imported, and it is automatically or manually identified as an "infectious disease notification" based on the OCR recognition results, metadata, etc. (Claim 1). In the manual case, a doctor or administrative staff member selects "Document type: Infectious disease notification." Detect candidates for item names and input areas: Based on the document type, a (standardized) list of item names (patient name, gender, date of birth, etc.) is obtained, the similarity with the strings in the OCRed document is evaluated, and the relevant areas are extracted as candidates for the "area where the item name is written." Optical area identification: The coordinate data (bounding box) and layout structure obtained from the OCR results are analyzed to identify the input area (the part corresponding to the text field) on the document. Linking field names to input areas: The system identifies the correspondence between the candidate areas and the character strings in the OCRed document, and matches the (standardized) item names (e.g., patient names) registered in the database with the input areas in the actual document. This allows the system to clearly understand "which items should be entered in which positions." Initial generation of the template document: The input fields for the field names obtained through OCR analysis are mapped to PDF templates prepared for each document type. If there are multiple standardized field names, fields corresponding to each field are defined on the layout, and the template is expanded or modified. Get character information: The system also OCRs patient information (name, ID, etc.), diagnosis, onset date, etc. read by the OCR module of other document types, such as referral letters, and aggregates them into (standardized) item names. Data from electronic medical records and DWHs can also be obtained as needed (Claim 2). Get a list of field names from the database: The item names (basic patient information, notified disease name, diagnosis date, notification date, testing, etc.) corresponding to the identified document type "infectious disease notification" are obtained from the database (Claim 1). Assigning items using LLM. LLM analyzes the text information collected above and automatically assigns which information corresponds to which item. For example, the part where the "date of onset" is clearly written as a date is assigned to the "date of onset" field (Claim 1). If necessary, the user manually enters corrections (Claim 4). Placement in PDF template: The above template is called up and the allocation results are automatically entered there (claims 3 and 6). Patient profile information, notified disease name, diagnosis date, etc. are automatically entered. Approval flow and output: Before submitting the notification, the doctor will make a final check and electronically sign or stamp the document via the workflow management system (Claim 5). After approval, the final submission is made by printing or electronically sending the PDF (Claim 1: Printing medical documents).

[0072] <Examples of accompanying functions> In this embodiment, in addition to the above flow, the following functions are additionally provided. Automatic text size change: The font size is automatically adjusted according to the length of the field name and input value to prevent the document layout from being disrupted. Color Settings: To highlight the input field, you can change the text color to red or blue. Handling of concatenated characters / Transparency of concatenated characters: It has a function that makes unnecessary spaces and symbols between concatenated characters transparent, for example when concatenating item names to fit on one line, and outputs the text as natural-looking text. For example, "Patient name: Yamada Taro" can be changed to "Patient name: Yamada Taro" to improve visibility. In addition, numbers and units such as 10 mg can be concatenated and handled. Bulk processing of date notation: Year, month, and day notations such as "Reiwa Year XX Month XX Day" are treated as a single string, and the "year," "month," and "day" are replaced with spaces or other symbols as necessary. It is also possible to convert it to a format such as "XX / XX / XX" to match the output format. Patient Name Composition: Even if the patient data source is divided into "patient last name" and "patient first name," they are automatically linked and combined into a single item called "patient name." Conversely, if you want to output the last name and first name separately, specify any delimiter (such as a space or hyphen) to separate and output them. Auto-fit text size and images: Even if the length of the string changes due to compositing, deletion, or replacement, the font size and layout will be readjusted so that it fits naturally within the template frame. When inserting a patient's photograph or examination image, the system uses automatic resizing and cropping functions to position the image at the appropriate size.

[0073] <Effects of this Example> Document identification and item name standardization, improving natural language processing performance of OCR results: The system automatically identifies various document types and associates them with standard field names, thereby streamlining the creation process. Automatic mapping of input fields and identifying input fields based on OCR analysis and layout information reduces placement errors caused by manual input or visual inspection. Furthermore, the system improves the performance of natural language processing by identifying fields that need to be entered using OCR. Improved operational efficiency through templating: Create and update templates to standardize document formats across your facility, simplify version control, and promote standardization of work. Extensive customization features: It has a wide range of functions, including changing font size and color, processing concatenated characters, batch conversion of dates, and image fitting, so you can flexibly adjust the format to suit on-site needs.

[0074] (3) Flow of creating test result report Select document type: Select "Create a new document" on the system and select the document type "Coronary Artery Examination Result Report." Manual selection is the norm, but automatic identification is also possible based on OCR information, subject line, etc. Acquisition of text information and test data: The numerical values ​​on the test result form read by OCR or test values ​​linked from the testing system or electronic medical record are imported. Patient profiles (name, ID, test date and time, etc.) are also automatically obtained from the DWH or electronic medical records at the same time (Claim 2). List of items and LLM analysis: The item names required for the "coronary artery test result report" (test items, measurement values, reference values, attending physician, etc.) are obtained from the database, and the LLM assigns the character information (Claim 1). Placement in a PDF template: The acquired data is automatically embedded in a PDF template for "test result reports" that has been standardized within the hospital in advance (claims 3 and 6). The result value and judgment comment are placed in the specified position. User Modifications and Approvals: The attending physician or technician can then check the data, move the insertion location, and add comments as necessary (Claim 4). Once approval is complete on the workflow management system, the data can be attached to the patient's electronic medical record as a PDF. Alternatively, it can be printed out on paper and used for explanations to the patient or for storage within the hospital (Claim 5).

[0075] (4) Flow of creating a letter of introduction Document type determination: Whether it is a "letter of introduction" is determined from data such as OCR reading or screen capture, or the user can manually specify the type of "letter of introduction." Automatic patient information capture: The items required for the referral letter (clinic name, patient name and ID, attending physician, diagnosis, progress summary, etc.) are checked in the database, and information is imported from the electronic medical record or DWH (Claim 2). Automatic assignment by LLM: If the patient's current medical history and past medical history are extracted in text, the LLM analyzes the content of the text and assigns it to the specified items on the referral letter (such as "chief complaint," "past medical history," and "treatment progress") (Claim 1). Placement and editing in PDF templates: The system reads a hospital-specific or regional collaboration "letter of referral" PDF template and automatically fills in the necessary information (claims 3 and 6). The user can add or correct information and finally complete the letter of introduction that can be submitted to the referral institution (Claim 4). Workflow management (if required): Depending on hospital rules, referral letters may also require an approval process, and a workflow management system can be used to centrally manage everything from creation to approval (Claim 5).

[0076] (5) Fill out the admission and discharge support center template Document type / template selection: Specify a common template used by the admission and discharge support center (for example, "Pre-admission interview sheet" or "Discharge coordination plan") as the document type and obtain a PDF template. Automatic assignment of patient profile information: The basic information of patients (name, ID, age, main illness, planned date of admission, etc.) registered in the electronic medical record or DWH is automatically assigned to the designated fields of the template (claims 3 and 6). At that time, the value restrictions and value types are checked. The restrictions are made to match the value restrictions and value types of the electronic medical record template. LLM analysis for additional information: When name, ID, age, main illness, expected date of admission, etc. are entered as text, LLM will sort them into the appropriate columns and map them to item names as necessary (Claim 1). Sharing with multiple departments: The template can be shared not only with the admission and discharge support center, but also with related wards and social workers, allowing each department to use information in the same format (Claim 7). Input PDF information into electronic medical record information: The information after checking the above values ​​is entered into the electronic medical record via template linkage. If necessary, the value type is converted.

[0077] (6) Flow of creating surgery / hospitalization application form Document type determination or selection: Document types such as surgery application forms and hospitalization application forms are registered in the database, and the user selects the appropriate form. Integration with electronic medical records and OCR: Patient information may be obtained from electronic medical records, and surgery dates and details may be obtained from other systems or OCR of handwritten input (Claim 2). LLM Allocation: The LLM automatically matches and maps patient names, IDs, scheduled surgery dates, surgical procedures, etc. to the appropriate items (Claim 1). The user can check and correct as needed (Claim 4). Automatic placement in PDF templates: Data is automatically embedded into the official "surgery application form" and "hospitalization application form" templates established by the hospital (claims 3 and 6). Workflow Management Approval: It is expected that surgery applications will require multiple steps, such as approval by a doctor, head nurse, and operating room manager. This will be linked to a workflow management system to allow online approval (Claim 5).

[0078] <Effects of this Example> Efficient documentation: The creation of various medical documents (infectious disease notifications, test result reports, referral letters, etc.) is greatly streamlined through the automatic assignment of OCR, LLM, and PDF templates. Reduced human error: This reduces errors caused by manual input or copy-paste, and ensures that patient information is reflected accurately. Users can make corrections as needed, making it highly reliable. Easier approval flow: Approval processes can be completed electronically through a workflow management system, reducing the hassle of circulating paper documents and stamping. Sharing templates across multiple facilities and departments: While each department, such as the admission and discharge support center or operating room, has its own customized PDF template, it can be standardized and shared across all facilities as needed, promoting collaboration between facilities and standardizing formats. Example 1

[0079] <Flow of creating and submitting an infectious disease notification report> The present invention improves the efficiency of creating infectious disease notifications. 1. Import the PDF of the infectious disease notification form and perform OCR. 2. Analyze the OCR results and metadata and classify it as an "infectious disease report." 3. Obtain a list of item names corresponding to infectious disease notifications from the database. 4. Match the character information obtained by OCR with the item name. 5. Auto-fill information into PDF templates. 6. After the doctor confirms, electronic approval is given through the workflow management system. 7. Generate a PDF for submission and print or submit electronically. Example 2

[0080] <Creating test result reports> The present invention is applicable to creating reports of test results. 1. Obtain test data via OCR or electronic medical record integration. 2. Identify the document type and select a template. 3. Obtain a list of test item names from the database. 4. Use LLM to match test data with item names. 5. Auto-fill data into PDF templates. 6. The doctor in charge checks the contents and approves them in the workflow system. 7. Generate a PDF report and attach it to the electronic medical record. Example 3

[0081] <Preparing a letter of introduction> 10 is a flow chart showing the application of the present invention to the creation of a letter of introduction. 1. Identify document type through OCR analysis or manual selection. 2. Obtain patient information from electronic medical records and DWH. 3. The LLM analyzes the document contents and assigns them to categories such as "chief complaint" and "medical history." 4. Auto-fill PDF templates. 5. After the doctor checks, the application is approved and submitted. Example 4

[0082] <Common Summary> The processes common to each embodiment (creating an infectious disease notification, creating a test result report, and creating a referral letter) are summarized in chronological order as follows: 1. Acquiring documents Import PDFs, perform OCR, or pull data from electronic medical records or data warehouses (DWHs) 2. Document Classification Analyzes OCR results and metadata to identify document type (automatic or manual selection). 3.Get a list of item names from the database The standardized field name corresponding to the document type is obtained from the database. 4. Mapping text information to item names Character information obtained by OCR and data from electronic medical records are matched with item names, and an LLM (large-scale language model) is used to assign the appropriate item. 5. Fill in the PDF template Automatically input acquired data into a PDF template. 6. User confirmation and correction If necessary, the doctor or administrative staff will manually correct and enter the information. 7. Implementing the approval flow Doctors and others make final confirmation and electronic approval through the workflow management system. 8. Output and save the final document Generate a PDF after approval and attach it to the electronic medical record or print and submit it.

[0083] Figure 7 shows the relationship between input fields and field names. When processing a document, there are input fields identified using OCR, and standardized field names are associated with them. The input fields include specific fields from various documents (patient name, date of birth, disease name, etc.), and it is important to properly organize them and match them with templates. The layout portion of the ●●●● notification (shown with a dashed line in the illustration) is subjected to OCR analysis to identify (specify) the document type, and the "standardized item names" for the items listed in the document are obtained from the database. Next, the "OCRed item name" (patient name) and the area where the item name is written are identified, and the input area is then identified. Next, the "OCRed item name" is linked to the input area, and the input area is then optically identified. Finally, the "(standardized) item name" for the document type and the "OCRed item name" are linked to a large-scale language model to create a template.

[0084] Using Figure 8, we will explain how the item names obtained through OCR processing are standardized and organized. The character information read by OCR is matched with standardized item names according to the document type, enabling uniform data processing even across documents of different formats. A prompt is set to link the standardized item names with the OCR results of the referral letter (Document 1). For example, standardized item names are entered into the item list linked to the document type "referral letter," and the OCR results of the referral letter are entered. The OCR results of the referral letter are linked to the standardized item names and extracted. Next, based on the results of OCR on the template for the infectious disease outbreak report (Document 2), the item names from the OCR results of the infectious disease outbreak report are extracted and retrieved, and a prompt is used to link them to the standardized item names extracted from the referral letter, simultaneously linking the values ​​corresponding to the item names.

[0085] Figure 9 describes the text visualization and formatting features, including: Change the font size Change the text color Character concatenation (e.g., combining names into "last name + first name") Transparency of concatenated characters (e.g., make unnecessary separators such as "year·month·day" transparent and hide them) In addition, the system shows a function to internally standardize the item names obtained by OCR and store them in association with data such as the "name of the doctor in charge."

[0086] Figure 10 shows the relationship between the referral letter and its OCR analysis results. The text information acquired when the referral letter is scanned by OCR is converted into a standardized data format and compared with a document template. This makes it possible to manage and utilize data in a consistent format, even for referral letters in different formats.

[0087] Figure 11 is an explanatory diagram that applies different output formats depending on each item name and its value. When cough, phlegm, fever, etc. are specified as the symptom selection format, a circle is selected for fever. Also, when a Yes / No binary option for whether the symptom is confirmed or estimated is specified, a circle is selected for confirmation. [Explanation of symbols]

[0088] 1: User terminal, 2: Medical information search system, 3: Medical system, 4: Data warehouse, 5: Search engine 10: User terminal (communication IF (12), input device (13), output device (14), processor (19), memory (15), storage unit (16)) 20: Server (Processor (29), Memory (25), Storage (26))

[0089] The program of the present invention can be installed in computer terminals having computer functions such as a CPU, memory, and storage, as well as mobile devices such as smartphones, tablets, and wearable devices, digital home appliances such as smart TVs, smart speakers, and smart home appliances, recording media such as USB memory, SD cards, hard disk drives (HDDs), and solid-state drives (SSDs), dedicated devices and terminals such as POS terminals, vending machines, ATMs, and medical equipment, and game consoles (home and portable).When the program is installed in a medical device, it may be linked to a hospital's electronic medical record (EHR / EMR) system.

[0090] [Appendix C1] a processor and a memory, A program for operating a computer having a database storing document types and item names corresponding to the document types, Execute a step of acquiring an input area list from a document through optical recognition, a step of acquiring recognized character information by analyzing through optical character recognition, and a step of automatically or manually identifying the document type of the document; a step of acquiring a list of item names corresponding to the identified document type from the database, and then assigning the acquired recognized character information to the item names by a program; A program comprising a step of specifying an input area corresponding to the item name from the input area list, and outputting a template document. [Appendix C2] a processor and a memory, A program for operating a computer having a database storing document types and item names corresponding to the document types, A step of acquiring recognized character information of the document 1 by analyzing the document 1 through optical character recognition, a step of acquiring an input area list through optical recognition, and a step of automatically or manually identifying the document type of the document 1 are carried out; a step of acquiring from the database a list of item names corresponding to the document type of the identified document 1, and then a step of assigning the recognized character information of the acquired document 1 in correspondence with the item names using a large-scale language model; a step of identifying an input area corresponding to the recognized character information of the document 1 from the list of input areas; and outputting a template document; performing a step of acquiring recognized character information of document 2 by analyzing document 2 through optical character recognition; and a step of automatically or manually identifying the document type of document 2; a step of acquiring from the database a list of item names corresponding to the document type of the identified document 2, and then assigning the acquired recognized character information of the document 2 to the item names and values ​​associated with the item names by using a large-scale language model; A program that inputs a value associated with the item name of the document 2 into an input area corresponding to the item name of the template document of the document 1, and outputs a new document. [Appendix C3] [Appendix C2] A program according to the present invention, A program characterized in that the character information of document 2 is at least one of recognized character information of document 2 extracted and obtained by optical character recognition, character information obtained from a data warehouse in which electronic medical record information is stored, or recognized character information extracted and obtained by optical character recognition from screen capture data of an electronic medical record. [Appendix C4] A program according to [Appendix C1] or [Appendix C2], A program characterized by generating or selecting and assigning recognized character information of the document 2 to a PDF template that defines a layout related to the list of item names and the list of input areas of the document. [Appendix C5] [Appendix C2] A program according to the present invention, A program characterized in that the value associated with the item name of the document 2 is associated with the item name by a user's operation and / or correction input. [Appendix C6] [Appendix C4] A program according to the present invention, A program that cooperates with a workflow management system to manage the approval flow of the document 2 generated based on the PDF template. [Appendix C7] [Appendix C1] A program according to the present invention, The item name list stores one or more of item names, value types, and value restriction information, When a value associated with the item name is entered in the input area of ​​the document, the type of the value or the restriction information of the value is checked, and if the value does not conform to these, a warning message is output on the screen or the color of the text is changed; A program characterized by obtaining a list of item names and value types or restriction information for the values ​​corresponding to the document type from an electronic medical record or matching the list with the electronic medical record. [Appendix C8] [Appendix C7] A program according to the present invention, An output method can be set in association with the document template or the item name, A program characterized by outputting the value associated with the item name of document 2 in the input area in one or more of the following expression formats when the value type is a Yes / No type value or a selection type value. [Appendix C9] [Appendix C3] A program according to the present invention, The document template can be shared among a plurality of facilities or departments.

[0091] [Appendix D1] A computer system comprising a processor and a memory, Operate a computer having a database storing document types and item names corresponding to the document types; Execute a step of acquiring an input area list from a document through optical recognition, a step of acquiring recognized character information by analyzing through optical character recognition, and a step of automatically or manually identifying the document type of the document; a step of acquiring a list of item names corresponding to the identified document type from the database, and then assigning the acquired recognized character information to the item names by a program; A system comprising a step of specifying an input area corresponding to the item name from the list of input areas, and outputting a template document. [Appendix D2] A computer system comprising a processor and a memory, Operate a computer having a database storing document types and item names corresponding to the document types; A step of acquiring recognized character information of the document 1 by analyzing the document 1 through optical character recognition, a step of acquiring an input area list through optical recognition, and a step of automatically or manually identifying the document type of the document 1 are carried out; a step of acquiring from the database a list of item names corresponding to the document type of the identified document 1, and then a step of assigning the recognized character information of the acquired document 1 in correspondence with the item names using a large-scale language model; a step of identifying an input area corresponding to the recognized character information of the document 1 from the list of input areas; and outputting a template document; performing a step of acquiring recognized character information of document 2 by analyzing document 2 through optical character recognition; and a step of automatically or manually identifying the document type of document 2; a step of acquiring from the database a list of item names corresponding to the document type of the identified document 2, and then assigning the acquired recognized character information of the document 2 to the item names and values ​​associated with the item names by using a large-scale language model; A system characterized by inputting a value associated with the item name of document 2 into an input area corresponding to the item name of the template document of document 1, and outputting a new document. [Appendix D3] [Appendix D2] The system according to the present invention, The system is characterized in that the character information of document 2 is at least one of recognized character information of document 2 extracted and obtained by optical character recognition, character information obtained from a data warehouse in which electronic medical record information is stored, or recognized character information extracted and obtained by optical character recognition from screen capture data of an electronic medical record. [Appendix D4] A system according to [Appendix D1] or [Appendix D2], A system characterized by generating or selecting and assigning recognized character information of the document 2 to a PDF template that defines a layout related to the list of item names and the list of input areas of the document. [Appendix D5] [Appendix D2] The system according to the present invention, The system is characterized in that the value associated with the item name of the document 2 is associated with the item name by a user's operation and / or correction input. [Appendix D6] [Appendix D4] The system according to the present invention, A system that cooperates with a workflow management system to manage the approval flow of the document 2 generated based on the PDF template. [Appendix D7] [Appendix D1] The system according to the present invention, The item name list stores one or more of item names, value types, and value restriction information, When a value associated with the item name is entered in the input area of ​​the document, the type of the value or the restriction information of the value is checked, and if the value does not conform to these, a warning message is output on the screen or the color of the text is changed; A system characterized in that a list of item names and value types or restriction information for the values ​​corresponding to the document type are obtained from an electronic medical record or are matched with the electronic medical record. [Appendix D8] [Appendix D7] The system according to claim 1, An output method can be set in association with the document template or the item name, A system characterized in that when the value type is a Yes / No type value or a selection type value, the value associated with the item name in document 2 is output to the input area in one or more of the following expression formats: a circle, an X, or a check. [Appendix D9] [Appendix D3] The system according to claim 1, The system is characterized in that the document template can be shared among multiple facilities or departments.

Claims

1. 1. A document output method implemented by a computer having a database, comprising: (a) accepting a document image; (b) detecting input areas on the document image by image analysis and obtaining a list of input areas including information about their positions; (c) performing optical character recognition (OCR) on the document image to obtain character information; (d) identifying the document type of the document image and obtaining from the database a list of item names associated with the document type; (e) analyzing the acquired character information using an artificial intelligence engine including natural language processing, extracting values ​​corresponding to each item name in the item name list or the list of item names, and associating the item names with the values; (f) identifying an input area corresponding to the item name from the list of input areas and outputting a template document; A document output method comprising:

2. 2. The document output method according to claim 1, The character information is (i) character information extracted by optical character recognition; (ii) textual information obtained from a data warehouse stored in an electronic medical record system; (iii) Character information obtained by optical character recognition of a captured image of an electronic medical record screen; A document output method comprising at least one of the steps of:

3. 2. The document output method according to claim 1, Obtain a PDF template in which the layout of the input area is defined; The document output method further comprises a step of assigning the item name and value associated in step (e) to the corresponding input area.

4. 2. The document output method according to claim 1, presenting a UI that allows a user to display and edit the correspondence results between the item names and values; The document output method further comprises a step of accepting a correction input from a user and updating the corresponding result.

5. 4. The document output method according to claim 3, The document output method further comprises a step of managing an approval flow for a document generated based on the PDF template in cooperation with a workflow management system.

6. 2. The document output method according to claim 1, the item name list or the item name table includes value type information or value restriction information corresponding to each item name, The document output method further comprises the step of outputting a warning on a screen when a value does not conform to the type information or the restriction information.

7. 7. The document output method according to claim 6, If the value related to the type information or the value restriction information is a Yes / No type or a selection type, The document output method further comprises a step of expressing the layout of the input area with at least one symbol of a circle, an X, or a check mark when outputting to a PDF template in which the layout of the input area is defined.

8. 4. The document output method according to claim 3, The document output method is characterized in that the PDF template is configured to be shareable among multiple facilities or departments.

9. A recording medium storing a program for causing a computer to execute the document output method according to any one of claims 1 to 4.

Images