Browsing assistance system, browsing assistance method, and computer-readable storage medium storing program

Abstract

An information processing terminal provided in this browsing assistance system has an input unit for creating a question having a content to be registered in medical-related information displayed on a display unit. A medical-related information management system has a creation assistance unit that outputs, to an input assistance system, creation information inputted by the information processing terminal in order to create the medical-related information, receives an answer from the input assistance system, and assists creation of the medical-related information. The input assistance system comprises a vector database in which knowledge vectorized and embedded in a vector space is stored, and a knowledge retrieval unit that retrieves the knowledge on the basis of the vectorized creation information from the vector database, and outputs, to the creation assistance unit, an answer obtained by shaping the retrieved knowledge.

Description

Browsing support system, browsing support method, and computer-readable storage medium storing a program

[0001] The present invention relates to a browsing assistance system, a browsing assistance method, and a computer-readable storage medium storing a program.

[0002] In health insurance treatment, the calculation rules and insurance points for dental diagnosis procedures are defined by the Ministry of Health, Labor and Welfare of Japan and are revised regularly. In particular, insurance points must reflect fluctuations in the price of gold, which is used in dental fillings, etc. For this reason, they are now revised every three months (four times a year) instead of twice a year.

[0003] Doctors and medical institutions that provide medical treatment covered by insurance must provide treatment in accordance with the rules of medical treatment covered by insurance, which are published in the Official Gazette issued by the Ministry of Health, Labor and Welfare of Japan, for example.

[0004] Patent Document 1 discloses a technology for improving the quality of medical data as a whole (e.g., all medical data related to a certain patient) by evaluating the consistency between reference data (reference data, first medical data) and other data (target data, second medical data).

[0005] JP 2024-011476 A

[0006] Traditionally, the writing of medical records was left to the ability of each individual. As a result, dentists who frequently had incomplete medical records experienced a lot of stress due to the constant return of records. As mentioned above, the rules for insured medical treatment are frequently revised, so not only medical records but also prescriptions must be written in accordance with the rules. Insured physicians are not only required to acquire knowledge and skills related to treatment, but also to be thoroughly familiar with the rules of insured medical treatment and to constantly update their knowledge.

[0007] However, the rules for insured medical treatment are vast and the information written is often difficult to understand. This makes it extremely difficult for insured physicians to keep up with the frequent cycle of updating their knowledge. If an insured physician makes an incorrect prescription, the prescription will be returned, and medical institutions with too many returns will be audited in the form of individual guidance. Dentists with little experience in insured medical treatment must continue practicing without being able to grasp such situations. Generally, the content taught to students at educational institutions such as dental schools mainly focuses on diagnostic and treatment methods. For this reason, dentists rarely receive sufficient training in how to write medical records and claim prescriptions at their educational institutions, and in most cases they only encounter these issues once they start working in the clinical field after graduation.

[0008] The above-mentioned Patent Document 1 describes, for example, support technology for detecting omissions in electronic medical records, but does not describe support technology for filling out electronic medical records in accordance with the rules of insured medical treatment.

[0009] The present invention has been made in view of the above circumstances, and aims to support the creation of medical-related information.

[0010] A browsing support system according to the present invention comprises an information processing terminal that displays medical-related information, a medical-related information management system that manages the medical-related information, and an input support system, all connected via a network. The information processing terminal has a display unit that displays the medical-related information and an input unit that creates questions to be registered in the medical-related information displayed on the display unit. The medical-related information management system has a creation support unit that outputs creation information for creating medical-related information input from the information processing terminal to the input support system, receives answers from the input support system, and supports the creation of the medical-related information. The input support system has a vector database that stores knowledge that has been vectorized and embedded in a vector space, and a knowledge search unit that searches the vector database for knowledge based on the vectorized creation information and outputs answers formed from the searched knowledge to the creation support unit.

[0011] According to the present invention, it is possible to support the creation of medical-related information. Note that the above-described browsing assistance system is one aspect of the present invention, and a browsing assistance method that reflects one aspect of the present invention and a computer-readable storage medium storing a program are also configured in the same manner as the above-described browsing assistance system. Problems, configurations, and effects other than those described above will become clear from the description of the following embodiments.

[0012] 1 is a functional block diagram showing an example of the overall configuration of a browsing support system according to a first embodiment of the present invention. FIG. 1 is a diagram showing an example of the data structure of data and information used in a medical record management system according to a first embodiment of the present invention. FIG. 2 is a diagram showing an example of the data structure of a medical record DB according to a first embodiment of the present invention. FIG. 3 is a block diagram showing an example of the internal configuration of a RAG system according to a first embodiment of the present invention. FIG. 4 is a diagram showing an example of the configuration of a knowledge DB according to a first embodiment of the present invention. FIG. 5 is a diagram showing an example of the configuration of a vector DB according to a first embodiment of the present invention. FIG. 6 is a diagram showing an example of the configuration of a prompt DB according to a first embodiment of the present invention. FIG. 7 is a block diagram showing an example of the hardware configuration of a computer according to a first embodiment of the present invention. FIG. 8 is a diagram showing a first display example of an electronic medical record creation display screen according to a first embodiment of the present invention. FIG. 9 is a diagram showing a second display example of an electronic medical record creation display screen according to a first embodiment of the present invention. FIG. 10 is a flowchart showing an example of knowledge addition processing in the RAG system according to the first embodiment of the present invention. FIG. 11 is a flowchart showing an example of knowledge update processing in the RAG system according to the first embodiment of the present invention. FIG. 12 is a flowchart showing an example of electronic medical record creation processing according to a first embodiment of the present invention. FIG. 13 is a flowchart showing an example of knowledge search processing by the RAG system according to a first embodiment of the present invention. FIG. 14 is a functional block diagram showing an example of the overall configuration of a browsing support system according to a second embodiment of the present invention.

[0013] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In this specification and drawings, components having substantially the same functions or configurations are designated by the same reference numerals, and redundant description will be omitted.

[0014] First Embodiment <Overall Configuration Example of a Viewing Assistance System> First, a configuration example and a processing example of a viewing assistance system according to a first embodiment will be described with reference to FIGS.

[0015] 1 is a functional block diagram showing an example of the overall configuration of a browsing support system 1 according to the first embodiment. Dental clinic staff who use the browsing support system 1 are assumed to be, for example, dentists, dental hygienists, reception staff, etc.

[0016] 2 is a diagram showing an example of the data structure of data and information used in the medical record management system 3. Below, in addition to the description of the browsing support system 1, the data and information used in the medical record management system 3 will also be described with reference to FIG.

[0017] The browsing support system 1 mainly comprises an information processing terminal 10 used in medical institutions such as dental clinics and medical universities, a medical receipt management system 2 operated by a service company, a medical record management system 3, and a Retrieval-Augmented Generation (RAG) system 4. The information processing terminal 10, the medical receipt management system 2, and the medical record management system 3 are connected via a network, and data can be sent and received safely using a secure communication method such as encryption and a VPN (Virtual Private Network). The medical receipt management system 2, the medical record management system 3, and the RAG system 4 are connected via the service company's internal LAN or the like.

[0018] The browsing support system 1 according to the first embodiment implements a medical record management function and an input support function using RAG in a cloud-based medical record management system 3 provided to dental clinics by a service company. Therefore, the medical records described below are "electronic medical records" that have been digitized using the medical record management function of the medical record management system 3, and these electronic medical records are insurance medical records that record the treatments given to patients and their insurance points.

[0019] The dental clinic staff accesses the medical record management system 3 by using their own information processing terminal 10 to input a user ID, password, etc. into the authentication screen of the medical record management system 3. The staff then browses the medical record DB 33 and creates an electronic medical record.

[0020] The RAG system 4 is an example of an input support system that presents answers to staff and supports their input. For example, when staff create electronic medical records, the RAG system 4 can provide input support. With the browsing support system 1 shown in Figure 1, dental clinics do not need to prepare individual servers, etc., which significantly reduces initial costs.

[0021] (Dental Clinic Staff) The information processing terminal 10 is an example of an information processing terminal such as a desktop PC (Personal Computer), a notebook PC, a smartphone, or a tablet. The information processing terminal 10 is connected to the medical receipt management system 2 and the medical record management system 3 via the Internet and displays medical-related information including medical receipts and medical records (electronic medical records). The medical receipt management system 2 and the medical record management system 3 are examples of medical-related information management systems that manage medical-related information. The information processing terminal 10 displays various screens provided by the medical receipt management system 2, the medical record management system 3, and the RAG system 4 through operation by the staff member, and accepts various information inputs by the staff member. For this purpose, the information processing terminal 10 includes a display unit 11 that displays medical-related information, and an input unit 12 that creates questions to be registered in the medical-related information displayed on the display unit 11.

[0022] For example, at a dental clinic, a dentist provides dental care to a patient and inputs the details of the treatment and medical fees into the information processing terminal 10. Also, at a dental clinic, a dental hygienist may operate the information processing terminal 10 to refer to the medical records and prescriptions displayed on the screen. At a medical university, a professor or a medical student may operate the information processing terminal 10 to obtain necessary information.

[0023] When a staff member operates the information processing terminal 10 to access the medical receipt management system 2, a medical receipt input screen (not shown) is displayed on the web browser of the display unit 11 based on the functions provided by the medical receipt management system 2. The information processing terminal 10 can also access the medical record management system 3 and display an electronic medical record creation display screen W1 shown in Figures 9 and 10 (described later) on the web browser of the display unit 11.

[0024] In addition, the dental clinic staff operates the input unit 12 to input words and phrases into the various screens displayed on the display unit 11 to create medical records and medical receipts. The insurance medical records created through the medical record creation process are electronic and are therefore called "electronic medical records." The words and phrases input from the input unit 12 are sent to the medical receipt management system 2 together with the information in the word and phrase input fields.

[0025] Furthermore, when the information processing terminal 10 accesses the medical record management system 3, it displays the display screen of the electronic medical record on the web browser of the display unit 11 based on the functions provided by the medical record management system 3. The electronic medical record is displayed in a fixed format that matches the layout of an actual book, allowing words on the page to be copied and pasted using the web browser or a dedicated application. In addition, dental clinic staff can switch to a reflowable display, which allows the font size to be changed to any size, or a horizontal or vertical display.

[0026] In addition, the dental clinic staff can highlight important points in the displayed electronic medical record, create indexes for pages, and convert the electronic medical record into a portable document format (PDF).The electronic medical record can be displayed on the information processing terminal 10 by turning one page at a time, or by turning multiple pages consecutively.

[0027] In addition to the "disease name" and "procedure name (or treatment name)," the electronic medical record may also record, for example, the patient's chief complaint (e.g., a throbbing pain in the area) and the dentist or other staff member's findings (e.g., bleeding when pressing the affected area with a finger). For example, the electronic medical record may include keywords such as the patient's chief complaint and the staff member's findings as trends in the disease name.

[0028] (Receipt Management System) Information entered into the receipt input field by dental clinic staff via the information processing terminal 10 is transmitted to the receipt management system 2 as medical treatment data d1 shown in Figure 2. The receipt management system 2 includes a receipt management unit 21 and a receipt DB (Data Base) 22.

[0029] The medical receipt management unit 21 manages data related to medical treatment and data related to the issuance of medical receipts, which are input from the information processing terminal 10. The medical receipt management unit 21 outputs an input screen for data related to medical treatment to the information processing terminal 10. The medical receipt management unit 21 also outputs information to be displayed on the information processing terminal 10 (such as an html (Hyper Text Markup Language) file, an xml (Extensible Markup Language) file, or an image file) to the information processing terminal 10. The web browser of the information processing terminal 10 displays a medical receipt input screen. The medical receipt management unit 21 also displays a screen for inputting medical treatment data, in addition to the electronic medical record input screen, on the display unit 11 of the information processing terminal 10 using a program that manages the issuance of medical receipts.

[0030] Medical data d1 is input to the medical receipt management unit 21 from the information processing terminal 10. The medical data d1 is data entered in the electronic medical record and data for creating medical receipts. The medical receipt management unit 21 imports the medical data d1 from the information processing terminal 10 and stores it in the medical receipt DB 22.

[0031] The medical receipt DB 22 stores medical treatment log data and basic medical institution information shown in FIG. 2, which will be described later. The medical receipt management unit 21 creates medical receipts based on data acquired from the medical receipt DB 22. For example, the medical receipt DB 22 stores medical treatment log data and medical receipt data created for each patient and each medical procedure. Staff at the dental clinic periodically operate the information processing terminal 10 to print medical receipts. For example, while the medical receipt data read from the medical receipt management system 2 is displayed, the staff presses a print button displayed on the screen of the information processing terminal 10, thereby outputting a printed receipt. This printed receipt is submitted to an examination and processing agency (not shown).

[0032] As shown in Figure 2, the medical data d1 stores the date and time, user ID, and user name. The date and time field stores the date and time when the dental clinic staff member entered the medical receipt using the information processing terminal 10. The user ID field stores a user ID that uniquely identifies the staff member as the user of the information processing terminal 10. The user name field stores the name of the user linked to the user ID, i.e., the dental clinic staff member. The medical receipt information field stores the medical receipt information entered by the user.

[0033] (Configuration of medical care log data) As described above, the medical care log data shown in FIG. 2 is stored in the medical receipt DB 22 of the medical receipt management system 2. Each item in the medical care log data includes a date and time, a medical professional ID, a patient name, and an event recorded in the medical record. The date and time item stores the date and time when the dental clinic staff filled out the electronic medical record. The medical professional ID item stores a medical professional ID assigned to identify the staff member.

[0034] The patient name field stores the patient's name. The event recorded in the medical record field stores information about events recorded in the electronic medical record by dental clinic staff. Events recorded in the electronic medical record include the quoted content (disease name, procedure name, summary, etc.), the date and time of citation, the citation by the citation (healthcare professional ID), the number of citations, etc., when the information on the page or phrase that the staff referenced in the electronic medical record is quoted in the electronic medical record.

[0035] (Configuration of medical institution basic information) Furthermore, the medical institution basic information shown in Figure 2 is registered in the receipt DB 22. The medical institution basic information is information that is also known by the service company. Each item of the medical institution basic information stores the name of the clinic, address, telephone number, home page URL (Uniform Resource Locator), and medical staff information.

[0036] The clinic name field stores the name of the dental clinic (or hospital). The address field stores the address of the dental clinic (or hospital). The phone number field stores the phone number of the dental clinic (or hospital). The website URL field stores the URL of the website managed by the dental clinic (or hospital). The medical staff information field stores information about medical staff working at the dental clinic (or hospital). The medical staff information includes, for example, information such as the university from which the medical staff graduated.

[0037] (Service Company) In addition to managing the entire viewing support system 1, the service company provides dental clinic staff with a service that supports the input of information to be recorded in the electronic medical record using the medical record management system 3. Typically, staff who can view the electronic medical record include, for example, other staff at the dental clinic where the staff member who created the electronic medical record works.

[0038] The medical record management system 3 displays the electronic medical record read from the medical record DB 33 on the display unit 11 of the information processing terminal 10 when the dental clinic staff operates the information processing terminal 10. The dental clinic staff operates the information processing terminal 10 connected to the medical record management system 3 to open the electronic medical record at any page, refer to the page corresponding to the treatment content, and create the patient's electronic medical record or medical receipt. The dental clinic staff can also copy and paste selected words from the electronic medical record into the receipt input field, or enter comments in the input field.

[0039] The medical record management system 3 includes a medical record management unit 31, a reference management unit 32, a medical record DB 33, and a reference log DB 34. The medical record management unit 31 manages electronic medical record information. The medical record management unit 31 also outputs creation information for creating medical-related information, which is input by the information processing terminal 10, to the RAG system 4. The medical record management unit 31 then receives a response from the RAG system 4 and supports the creation of medical-related information. In other words, the medical record management unit 31 can be said to be an example of a creation support unit that supports staff in creating medical-related information.

[0040] The creation information for creating medical-related information includes questions entered in the display area of ​​the medical-related information creation display screen (see Figures 9 and 10, described below) used for creation. This creation information also includes information about items selected by dental clinic staff on the medical-related information creation display screen. For example, the creation information includes questions asked by staff to create electronic medical records or medical claims, instructions for creating a summary document, etc. This creation information includes at least one of text, audio, and images. The answers also include the summary document and information about the reference source of the reference information summarized by LLM 45 (e.g., page number, chapter title, etc. of the reference information).

[0041] The medical record management unit 31 accepts input to the electronic medical record by dental clinic staff and outputs a screen displaying the electronic medical record to the display unit 11 of the information processing terminal 10. The medical record management unit 31 manages the electronic medical records whose creation is supported by the RAG system 4. The electronic medical record is generated based on information obtained from the RAG system 4 when the knowledge search unit 44 inputs creation information (prompts, described later) into the LLM 45.

[0042] The RAG system 4 performs embedding processing by reading past electronic medical records, official gazettes, etc., and mapping vectorized text and images into a vector space. The RAG system 4 also uses large language models (LLMs) to output information appropriate for the input content. A detailed configuration example of the RAG system 4 is shown in Figure 4, which will be described later, and only a brief explanation will be given here.

[0043] The RAG system 4 includes an LLM 45 (an example of a large-scale language model) shown in FIG. 4 (described later). The LLM 45 may be provided by a service other than the service company. The LLM 45 is a type of text generation AI (artificial intelligence) and is a language model trained on a vast amount of text data. The LLM 45 is capable of performing advanced natural language processing tasks (text generation, translation, summarization, question answering, etc.). Furthermore, the LLM 45 utilizes the Transformer mechanism to output appropriate answers based on generated information (e.g., questions) entered by dental clinic staff. The generated information may be text explicitly entered by staff, or text or images automatically extracted from information displayed on the screen.

[0044] The reference management unit 32 collects data such as reference log data d2 (see FIG. 2 ) from the reference log DB 34 in accordance with commands received from the information processing terminal 10, and passes the data to the medical receipt DB 22 of the medical receipt management system 2. The medical receipt management unit 21 manages medical receipt data stored in the medical receipt DB 22 for each dental clinic or each dental clinic staff member.

[0045] 2 includes information on the reference history of electronic medical records that dental clinic staff referred to when creating electronic medical records or inputting medical claims by operating the information processing terminal 10. The reference log data d2 is stored in the reference log DB 34 of the medical record management system 3.

[0046] The reference log data d2 is data that stores the type of electronic medical record that the staff member referenced, the version of the electronic medical record (such as the creation date), the amount of time the electronic medical record was left on the electronic medical record, etc. The reference management unit 32 reads out the version of the electronic medical record specified by the input unit 12 from the medical record DB 33 and displays it on the display unit 11 of the information processing terminal 10. In addition, operations that the dental clinic staff performed while referring to the electronic medical record when entering a medical receipt (for example, copying and pasting specific items) are also stored in the reference log DB 34 as reference log data d2.

[0047] (Configuration of reference log data) The reference log DB 34 stores the reference log data d2 shown in Figure 2. The reference log data d2 stores the medical record ID, book name, version, index, page information, and reference page. The medical record ID field stores the medical record ID that uniquely identifies the electronic medical record being viewed on the information processing terminal 10. The book name field stores the name of the electronic medical record linked to the medical record ID as the book name. The version field stores the version of the electronic medical record (date of creation, etc.). The page information field stores page information such as the number of pages in the electronic medical record.

[0048] In addition, the reference log DB 34 may store a receipt ID and return information indicated by a dotted line as reference log data d2. The receipt ID field stores a receipt ID for identifying a receipt submitted to the review processing agency. The receipt ID is unique identification information assigned to each receipt when the staff member creates the receipt. The return information field stores a return flag indicating whether the receipt submitted to the review processing agency has been returned, and return information including words and sections that the review processing agency pointed out for correction at the time of return.

[0049] 3 is a diagram showing an example of the data structure of the medical record DB 33. The medical record DB 33 has fields for patient ID, book ID, date, body part, disease name, and treatment. The patient ID field stores a patient ID that uniquely identifies a patient. The book ID field stores a book ID that uniquely identifies a book version of the electronic medical record. The date field stores the creation date or update date of the electronic medical record.

[0050] The site field stores information about the site that was treated. The disease field stores information about the name of the patient's disease. The treatment field stores information about the treatment given to the patient.

[0051] For a first-time patient, a new electronic medical record is created, and the creation date of the electronic medical record is stored in the date field. For a returning patient, an additional entry is made to the electronic medical record of the existing patient ID, and a new record is added to the medical record DB 33, and the details of the treatment for the returning patient are stored in that record.

[0052] <Example of Internal Configuration of RAG System> Fig. 4 is a block diagram showing an example of the internal configuration of the RAG system 4. Here, the cooperation between the medical record management system 3 and the RAG system 4 will be described.

[0053] First, let us explain RAG. RAG is a technology that combines information retrieval capabilities with a generative model (usually a text generation model), and is primarily composed of a search engine (Retriever) and a generative model (Generator). A search engine has the function of searching for relevant documents and information from a database based on a user's input (query). To do this, the search engine performs similarity calculations using an embedding model that embeds the user's query and documents in the database in a vector space to search for documents related to the query. Meanwhile, a generative model generates responses and text for the user based on the information retrieved by the search engine. To do this, the generative model generates responses by converting the contents of documents retrieved by the search engine into language expressions that are easy for humans to understand.

[0054] The RAG system 4 includes a knowledge upload unit 41, a knowledge update unit 42, an embedding unit 43, a knowledge search unit 44, an LLM 45, a knowledge database (knowledge DB 51), a vector database (vector DB 52), and a prompt database (prompt DB 53).

[0055] A knowledge upload unit 41, a knowledge update unit 42, and a knowledge search unit 44 are each connected to the medical record management system 3. The knowledge upload unit 41 is connected to a knowledge DB 51 and uploads knowledge acquired from the medical record management system 3 to the knowledge DB 51. Knowledge includes, for example, official gazettes published by the Ministry of Health, Labor and Welfare of Japan, guidelines for filling out medical insurance claims, past electronic medical records, etc. The knowledge uploaded by the knowledge upload unit 41 is stored in the knowledge DB 51.

[0056] The knowledge update unit 42 is connected to the embedding unit 43 and the knowledge DB 51. The knowledge update unit 42 updates the knowledge stored in the knowledge DB 51. The knowledge update unit 42 then acquires, from the knowledge DB 51, knowledge that has been instructed to be updated by the medical record management system 3, and outputs the acquired knowledge to the embedding unit 43.

[0057] The embedding unit 43 uses an embedding model to vectorize the knowledge input from the knowledge update unit 42 and embed it in a high-dimensional vector space. The knowledge embedding process uses a technique for mapping data such as text and images into a vector space. The embedding unit 43 performs tokenization to divide the knowledge into tokens, then vectorizes the tokens and generates a vector representation of the entire query.

[0058] A vector representation of the entire query is obtained by integrating token vectors in the query using an appropriate method (such as average pooling or max pooling). The embedding process in the embedding unit 43 converts words, sentences, and entire documents in the text into numerical vectors, and semantically similar words and documents are placed close to each other in the vector space. Using these numerical vectors facilitates machine learning tasks such as similarity calculations and clustering between different texts. The embedding unit 43 then stores the vectorized knowledge in the vector DB 52.

[0059] The vector DB 52 stores knowledge that has been vectorized and embedded in a vector space. For this reason, the vector DB 52 is constructed as a database specialized for vector search. The vector DB 52 stores chunks (e.g., text) that are obtained by dividing a file stored in the knowledge DB 51 into multiple chunks, as well as information about the chunks expressed as vectors.

[0060] The text embedded in the vector space is stored in the vector DB 52, for example, in a matrix format. Each row of the matrix represents an individual text vector, and each column represents a dimension of the vector. The vector may also be stored as a numerical array of floating-point numbers or in binary format. Searchability is improved by storing the vector in the vector DB 52 together with the text corresponding to the vector itself, the document ID of the document from which the text was written, the publication date and time, etc.

[0061] The knowledge search unit 44 is connected to the medical record management system 3, the embedding unit 43, the LLM 45, the vector DB 52, and the prompt DB 53. The knowledge search unit 44 searches the vector DB 52 for knowledge based on the vectorized creation information, and outputs a response formed from the searched knowledge to the medical record management unit 31.

[0062] When the knowledge search unit 44 receives a question (query) input from the medical record management system 3, it tokenizes the question by dividing it into tokens, and then vectorizes the tokens to vectorize the question.The knowledge search unit 44 then extracts vectorized knowledge based on the similarity between the vectorized knowledge stored in the vector DB 52 and the vectorized creation information (query vector).At this time, the knowledge search unit 44 extracts one record stored in the vector DB 52.

[0063] In a knowledge search, the knowledge search unit 44 calculates the similarity between a query vector and each vector in the database. Vector similarity is measured using, for example, cosine similarity or Euclidean distance. Cosine similarity is a method of calculating the cosine of the angle between two vectors, and its value ranges from -1 to 1. The closer the value is to 1, the more similar the vectors are determined to be. Euclidean distance is a method of calculating the linear distance between two vectors, and the smaller the value, the more similar the vectors are determined to be.

[0064] The embedding unit 43 embeds the question text in a vector space and searches the vector DB 52 for a vector that answers the question. The embedding unit 43 then determines the vector with the highest similarity as the answer with the highest confidence. The embedding unit 43 returns the text corresponding to the vector searched for in the vector DB 52 to the knowledge search unit 44 as a search result.

[0065] The knowledge search unit 44 also outputs the text of the knowledge found by the vector search to the LLM 45. The LLM 45 receives the text of the knowledge from the knowledge search unit 44 and a question (prompt) composed of regular characters. The LLM 45 then generates natural-sounding sentences based on the prompt and generates an answer based on the prompt. The LLM 45 also reads reference information in advance and summarizes the reference information based on prompts consisting of text or images. For example, the LLM 45 can summarize long texts contained in previously read reference information such as official gazettes and information on the website of the Ministry of Health, Labor and Welfare of Japan, and extract important information.

[0066] The LLM 45 can also translate text from one language into another and categorize text. By utilizing the LLM 45, the RAG system 4 can utilize the text generation capabilities of the LLM 45 to provide context-appropriate responses to staff based on the retrieved information. This allows the RAG system 4 to provide interactive functionality to the medical record management system 3. The knowledge search unit 44 can also output a summary document summarized by the LLM 45 to the medical record management system 3.

[0067] The following first to third modes are assumed as the entity that selects a prompt from the prompt DB 53. In the first mode, the knowledge search unit 44 specifies a fixed prompt ID, and retrieves and operates a prompt corresponding to the specified prompt ID from the prompt DB 53.

[0068] The second form is a derivative of the first form, and is implemented, for example, in such a way that the knowledge search unit 44 displays a list of prompts acquired from the prompt DB 53 on a screen, and the staff of the dental clinic can select any prompt.

[0069] In the third configuration, the LLM 45 and the prompt DB 53 are connected as shown by the dashed line in the figure. That is, in this third configuration, the LLM 45 directly references the prompt DB 53. In this third configuration, dental clinic staff search for the most appropriate prompt to process the input question. The LLM 45 then uses the searched prompt to perform processing such as creating a summary.

[0070] Next, configuration examples of the knowledge DB 51, vector DB 52, and prompt DB 53 will be described with reference to FIGS. 5 to 7. FIG. 5 is a diagram showing a configuration example (data structure) of the knowledge DB 51. The knowledge DB 51 has the following fields: document ID, file name, file type, and BLOB (Binary Large Object). The document ID field stores a document ID that can identify knowledge as a document. The document includes information on various printed materials that can be referenced when creating an electronic medical record.

[0071] The file name item includes the file names of the various printed materials mentioned above. For example, PDF format medical record case studies, insurance revisions, memos, etc. Also included are file names of files saved in text format, document file format, and presentation file format. The file type item includes the file type (e.g., extension) of the various printed material files. The BLOB item stores data that has been compiled into a book, i.e., BLOB data. BLOB data includes text data or binary data of image data.

[0072] FIG. 6 is a diagram showing an example of the configuration (data structure) of the vector DB 52. The vector DB 52 has the following fields: document ID, chunk ID, text, vector, file name, character position, and page. The document ID field stores a document ID. This document ID is the same as the document ID stored in the knowledge DB 51. The chunk ID field stores chunk IDs that identify chunks obtained by dividing a document into a predetermined number of characters. Multiple chunk IDs may be included for the same document ID. For example, a document with a document ID of "1" may include chunk IDs of "1" to "4."

[0073] The text field stores text that has been split into chunks from a document. A chunk is, for example, a group of 500 characters, and is automatically split from the original document. However, chunks identified by consecutive chunk IDs are likely to lose context. For this reason, consecutive chunks are split from the original document so that some characters overlap each other.

[0074] For example, a chunk with a chunk ID of "1" is the text from characters 0 to 499 of "Medical Record Case Collection," and a chunk with a chunk ID of "2" is the text from characters 449 to 948, with the division starting from a position 50 characters back from the division end character position of the chunk with a chunk ID of "1." Note that, to clarify the context, for example, the division start character position for dividing the next chunk may be, going back from the division end character position of the previous chunk to the first period. Alternatively, the heading and the main text may each be divided into a single chunk.

[0075] The vector item stores the vector calculated for each chunk. Based on the characters contained in the chunk, it is tokenized into the smallest unit of meaning, and the tokens are embedded in a vector space to calculate the numeric vector of the token. The chunk vector calculated by combining the numeric vectors of each token is stored in the vector item. This process results in a vector that retains the meaning of each chunk in a high-dimensional vector space.

[0076] The file name field stores the file name of the file that contains the text. The character position field stores the start and end character positions of the chunks split from the original document. For example, the character positions of a chunk with a chunk ID of "1" are start character position "0" and end character position "499." The page field stores the number of pages in the file that contains the chunk.

[0077] 7 is a diagram showing an example of the configuration (data structure) of the prompt DB 53. The prompt DB 53 has fields for ID and prompt.

[0078] The ID item stores a prompt ID that identifies the prompt. The prompt item stores a prompt that is input to the LLM 45. A prompt is a command that dental clinic staff inputs to the RAG system 4 through the medical record management system 3. For example, the prompt for ID1 has the following content. "You" in the prompt refers to the LLM 45. "You are a Japanese AI assistant for the RAG, and will converse in a friendly tone. The following knowledge has been found in response to the user's question. {"Text": (text from 0 to 499th character of "Medical Record Case Collection")} Please explain to the user that the above knowledge has been found, and provide a summary in your response."

[0079] <Example of Hardware Configuration of Information Processing Terminal 10 and Each Server> Next, a description will be given of the hardware configuration of the computer 100 that constitutes the information processing terminal 10, the medical receipt management system 2, the medical record management system 3, and the RAG system 4 of the browsing assistance system 1. FIG. 8 is a block diagram showing an example of the hardware configuration of the computer 100.

[0080] The computer 100 is an example of hardware used as a computer capable of operating as the information processing terminal 10, the medical receipt management system 2, the medical record management system 3, and the RAG system 4 according to this embodiment. The computer 100 includes a CPU (Central Processing Unit) 101, a ROM (Read Only Memory) 102, and a RAM (Random Access Memory) 103, each connected to a bus 104. The computer 100 further includes a display device 105, an input device 106, a non-volatile storage 107, and a network interface 108.

[0081] The CPU 101 reads out the program code of the software that realizes each function according to this embodiment from the ROM 102, loads it into the RAM 103, and executes it. Variables, parameters, etc. generated during the calculation process of the CPU 101 are temporarily written to the RAM 103, and these variables, parameters, etc. are read out by the CPU 101 as appropriate. However, an MPU (Micro Processing Unit) may be used instead of the CPU 101. Each function of the receipt management system 2, the medical record management system 3, and the RAG system 4 is realized by the operation of the CPU 101.

[0082] The display device 105 is, for example, a liquid crystal display monitor, and displays the results of processing performed by the computer 100 to the staff. The input device 106 is, for example, a keyboard, a mouse, etc., and allows the staff to input predetermined operations and give instructions. The functions of the display unit 11 and input unit 12 of the information processing terminal 10 are realized by the display device 105 and the input device 106. In addition, the functions of the display unit and input unit of a PC (not shown) connected to the medical record management system 3 are also realized by the display device 105 and the input device 106. The RAG system 4 does not need to be equipped with the display device 105 and the input device 106.

[0083] The nonvolatile storage 107 may be, for example, a hard disk drive (HDD), a solid state drive (SSD), a flexible disk, an optical disk, a magneto-optical disk, a CD-ROM, a CD-R, a magnetic tape, or a nonvolatile memory. In addition to an operating system (OS) and various parameters, the nonvolatile storage 107 also stores programs for operating the computer 100.

[0084] The ROM 102 and the non-volatile storage 107 store programs, data, etc. necessary for the operation of the CPU 101, and are used as an example of a computer-readable, non-transitory storage medium that stores programs executed by the computer 100. Data from the receipt management system 2, the medical record management system 3, and the RAG system 4 is stored in the non-volatile storage 107.

[0085] The network interface 108 may be, for example, a network interface card (NIC), and various data can be transmitted and received between devices via a LAN, dedicated line, or the like connected to a terminal of the NIC.

[0086] [Screen Display Examples] Next, display examples of screens displayed on the information processing terminal 10 will be described with reference to Fig. 9 and Fig. 10. The following screens are all displayed on a general-purpose web browser installed on the information processing terminal 10.

[0087] <First display example of electronic medical record creation display screen> Figure 9 is a diagram showing a first display example of the electronic medical record creation display screen W1. The display function of the electronic medical record creation display screen W1 is realized by the medical record management unit 31. The electronic medical record creation display screen W1 is used in the medical record management system 3 equipped with a chat-type RAG. This electronic medical record creation display screen W1 has an oral cavity information display area W1a, a medical record information display area W1b, and a chatbot display area W1d.

[0088] The oral cavity information display area W1a displays a schematic diagram of the patient's oral cavity. The schematic diagram of the oral cavity shows the arrangement and condition of the patient's teeth. The medical record information display area W1b displays the contents recorded in the electronic medical record as medical record information. The medical record information includes the date of the consultation as well as the details of the treatment performed on the patient. In addition, a PDF button W1c is displayed in the medical record information display area W1b.

[0089] The dental clinic staff presses the PDF button W1c to input an instruction to convert the medical record information into a PDF file. The PDF-converted medical record information is uploaded to the knowledge DB 51 by the knowledge upload unit 41 and vectorized by the embedding unit 43. The vectorized medical record information is then stored in the vector DB 52. This makes it possible to perform knowledge searches based on past medical record information.

[0090] The chatbot display area W1d displays an answer chat represented by an image of a robot and a question chat in which a staff member inputs a question. Initially, the answer chat displays "Is there something I can help you with?" The staff member inputs "Q1. What crown prostheses are covered by the crown and bridge maintenance fee?" into the question chat (shown in the figure as "Q1. Crown and bridge...?"). The answer chat displays "We will search for a relevant item in the knowledge base."

[0091] When a staff member clicks on the "Crown and Bridge Maintenance Fee (Br5 teeth or less)" item displayed in the medical record information display area W1b, the question chat may automatically be configured to input "What crown prostheses are covered by the crown and bridge maintenance fee?". The answer chat then displays a summary categorized as follows: "Prostheses covered by crown and bridge maintenance include the following: 1. ..., 2. ...." This answer is an explanation summarized by the LLM 45 from the search results in the vector DB 52 based on the question vectorized by the embedding unit 43.

[0092] The answer chat also displays the file name that is the basis for the answer and the relevant section within the file. For example, the answer chat displays both the content and the page number, such as "2024 Insurance Revision.pdf M015-2 CAD / CAM crowns are... Page 25." The basis for the answer may be a file name or a link containing a URL that can access the file. The answer chat may also display multiple answers in order of the confidence level of the answer, or the staff member may select any answer.

[0093] As shown at the bottom of the chatbot display area W1d, information entered by the staff member via keyboard operation or voice, such as "Q2. What crown prostheses are covered by the crown and bridge maintenance fee?" (shown as "Q2. Crown and bridge...?" in the figure) may be displayed. Alternatively, only "Kurabuuri," the abbreviation for "crown and bridge maintenance fee," may be entered and displayed in the question chat. Even with such a question, the staff member can press the send button to execute the knowledge search process shown in Figures 13 and 14, which will be described later. The question text is then sent to the RAG system 4 via the medical record management system 3, and an answer can be obtained.

[0094] <Second display example of electronic medical record creation display screen> Figure 10 is a diagram showing a first display example of the electronic medical record creation display screen W2. The display function of the electronic medical record creation display screen W2 is realized by the medical record management unit 31. The electronic medical record creation display screen W2 is used in a medical record management system 3 equipped with a type of RAG that automatically detects input. This electronic medical record creation display screen W2 has an input information display area W2a and a knowledge display area W2b.

[0095] The input information display area W2a displays the items to be input to the electronic medical record. For example, if a staff member selects the item "Crown and bridge maintenance fee (Br5 teeth or less)," this item is highlighted.

[0096] The knowledge display area W2b displays the results of an automatic search for knowledge related to the item selected in the input information display area W2a. For example, the procedure name, disease name, or comment selected by the staff member is automatically used as a search term by the medical record management system 3. Knowledge related to this search term is always displayed in the knowledge display area W2b. For example, the knowledge display area W2b displays "Related knowledge found" at the top. The middle section displays "Prosthetic devices eligible for crown and bridge maintenance include the following: 1. ... 2. ..." and the reference document information reads, "CAD / CAM crowns listed in 2024 Insurance Revision.pdf M015-2 are... Page 25." The bottom section displays, "Calculation of crown and bridge maintenance fees: To calculate crown and bridge maintenance fees, a document must be provided to the patient that includes the date of insertion, the name of the insured medical institution, the prosthetic location, and the purpose of crown and bridge maintenance. The reason is..." However, in the diagram, it is written as "Crown and bridge maintenance fee calculation: ..." This allows staff to reduce the time and effort required to check by themselves, while still being able to fully focus on medical treatment and communication with patients.

[0097] Note that the search target of the RAG system 4 is not limited to text. For example, the medical record management system 3 periodically takes screenshots of the input information display area W2a. By inputting the captured image into the multimodal RAG system 4, the medical record management system 3 can search for information corresponding to the captured image and display it in the knowledge display area W2b. In this case, information related to the knowledge displayed in the input information display area W2a is always displayed in the knowledge display area W2b.

[0098] [Example of operation of medical record management system and RAG system] Next, an outline of an example of operation of the medical record management system 3 and RAG system 4 according to this embodiment will be described with reference to Figures 11 to 14. In the following description, reference will be made as appropriate to the medical record management system 3 and RAG system 4 shown in Figure 1 and the internal configuration example of the RAG system 4 shown in Figure 4.

[0099] First, an example of the process of adding and updating knowledge in the RAG system 4 will be described with reference to Figures 11 and 12. These processes are performed before the service company provides the service to the dental clinic, but can also be performed at any time after the service is provided. Furthermore, these processes can be performed not only by the service company staff but also by staff at the dental clinic using the medical record management system 3.

[0100] 11 is a flowchart showing an example of knowledge addition processing in the RAG system 4. First, the knowledge upload unit 41 selects a file to be uploaded to the knowledge DB 51 (S1). Next, the knowledge upload unit 41 checks whether a file with the same file name as the selected file is stored in the knowledge DB 51 (S2).

[0101] If a file with the same file name is not stored in the knowledge DB 51 (NO in S2), the knowledge upload unit 41 uploads the selected file to the knowledge DB 51 (S3), and ends this process.

[0102] On the other hand, if a file with the same file name is stored in the knowledge DB 51 (YES in S2), the knowledge upload unit 41 ends this process without adding the knowledge. However, if the file has the same file name but a different version, the knowledge upload unit 41 may upload the different version of the file to the knowledge DB 51 and then end this process.

[0103] 12 is a flowchart showing an example of a knowledge update process in the RAG system 4. First, the knowledge update unit 42 acquires data including the knowledge to be updated from the knowledge DB 51 (S11). Next, the embedding unit 43 vectorizes the data using an embedding model (S12). This vectorization process includes, for example, a process of dividing a document into chunks, a process of tokenizing words and phrases in the chunks, and a process of vectorizing the tokens.

[0104] Next, the embedding unit 43 stores the vectorized data in the vector DB 52 (S13). Then, the knowledge update unit 42 determines whether or not the vectorized knowledge stored in the vector DB 52 has been updated with all the data acquired from the knowledge DB 51 (S14).

[0105] If the vectorized knowledge has not been updated with all the data (NO in S14), the process returns to step S11 and is repeated. On the other hand, if the vectorized knowledge has been updated with all the data (YES in S14), the process ends.

[0106] Next, an example of the processing of the viewing support system 1 when a staff member creates an electronic medical record will be described with reference to Figures 13 and 14. Figure 13 is a flowchart showing an example of the processing of creating an electronic medical record. Operations on the information processing terminal 10 are performed via a web screen displayed when the staff member accesses the medical record management system 3. Therefore, each operation on the information processing terminal 10 is executed by each functional unit of the medical record management system 3.

[0107] When a dental clinic staff member starts creating an electronic medical record by operating the input unit 12 of the information processing terminal 10, the medical record management unit 31 determines whether the created electronic medical record is stored in the medical record DB 33 (S21). This determination is made based on whether the patient is visiting for the first time or a repeat visit. The dental clinic staff member inputs the patient number written on the patient's patient card into the information processing terminal 10 to check whether the patient number is registered in the medical record DB 33.

[0108] If the created electronic medical record is stored in the medical record DB 33 (YES in S21), the medical record management unit 31 retrieves the created electronic medical record from the medical record DB 33 (S22) and displays it on the display unit 11. This electronic medical record is displayed on the electronic medical record creation display screen W1 (see FIG. 9) or W2 (see FIG. 10).

[0109] On the other hand, if the created electronic medical record is not stored in the medical record DB 33 in step S21 (NO in S21), a new electronic medical record is created (S23).

[0110] The dental clinic staff inputs the necessary information while checking the electronic medical record. If there is an item that the staff member is unsure about, they select that item and input a question into the question chat. The selected item is input as a prompt (text or image) to the medical record management system 3 (S24). The medical record management unit 31 outputs the input prompt to the RAG system 4 (S25). The RAG system 4 performs a knowledge search process based on the prompt input from the medical record management system 3 (S26).

[0111] Here, the knowledge search process performed by the RAG system 4 will be described with reference to Fig. 14. Fig. 14 is a flowchart showing an example of the knowledge search process by the RAG system 4. When the knowledge search unit 44 shown in Fig. 4 receives a prompt input from the medical record management system 3, it inputs the prompt as a search query to the embedding unit 43 (S31). As described above, the prompt includes text or an image.

[0112] Next, the embedding unit 43 vectorizes the search query using the embedding model (S32), and then acquires knowledge that has a high similarity to the vectorized search query from the vector DB 52 (S33).

[0113] Next, the embedding unit 43 determines whether or not knowledge has been acquired from the vector DB 52 (S34). If knowledge has not been acquired from the vector DB 52 in step S34 (NO in S34), the embedding unit 43 ends this process and proceeds to step S27 in FIG.

[0114] On the other hand, if knowledge can be acquired from the vector DB 52 in step S34 (YES in S34), the embedding unit 43 passes the knowledge to the LLM 45. The LLM 45 acquires an arbitrary prompt from the prompt DB 53 based on the knowledge (S35).

[0115] Next, the search query, knowledge, and prompt are input to the LLM 45 (S36).The LLM 45 then generates an explanation of the search results and a summary of the knowledge in accordance with the prompt (S37), and the process proceeds to step S27 in FIG.

[0116] Returning to the explanation of Figure 13, the medical record management unit 31 acquires an answer including an explanation of the search results and a summary of the knowledge from the RAG system 4 (S27). The medical record management unit 31 displays the acquired answer on the screens W1 and W2. Next, the medical record management unit 31 reflects the answer in the electronic medical record and saves the electronic medical record in the medical record DB 33 (S28), and this process ends.

[0117] The dental clinic staff may use voice input instead of operating the input unit 12 to write the prompt. For example, the input unit 12 may capture a conversation between the staff and the patient via a microphone, convert the voice input into text, and output the text information as a prompt to the RAG system 4. By using voice input, the staff does not need to hold the information processing terminal 10 and can concentrate on treating the patient.

[0118] In the browsing support system 1 according to the first embodiment described above, staff can use the medical receipt management function and the medical record management function simply by preparing the information processing terminal 10, without the need for a dedicated server. The medical record management function incorporates a response support function using the RAG system 4, and whenever staff enter information when creating an electronic medical record, information related to this entry is displayed. Therefore, dental clinic staff can create electronic medical records in the correct format while referring to the information displayed on the display unit 11.

[0119] In addition, when dental clinic staff enter a question when creating an electronic medical record, the answer to that question is displayed. The answer is a summary of the original long document (reference information) in LLM45, making it easy for staff to understand. Furthermore, the reference information of the document that is the source of the answer is also displayed, so when staff want to check more detailed information, they can immediately access the reference information of the document.

[0120] Furthermore, since the service company manages the medical record management system 3, there is no need to install a server or the like at the dental clinic. This reduces the introduction costs and ongoing costs of the server or the like required for introducing the browsing assistance system 1. Furthermore, the information processing terminal 10 does not need to be installed in a specific location.

[0121] For example, dental clinic staff can access the medical receipt management system 2 and the medical record management system 3 via the Internet using an information processing terminal 10 (e.g., a notebook PC or a tablet terminal) that they have taken outside the dental clinic. This makes it easy for dental clinic staff to check the input status of the electronic medical records and medical receipts, and also makes it easy to check the contents of the electronic medical records.

[0122] Answers to questions entered by dental clinic staff are generated using AI technologies such as LLM45. The answers include, for example, explanations of search results and knowledge summaries, making it easier for staff to understand complex documents.

[0123] Furthermore, when a new document is published, the knowledge is uploaded. Furthermore, when a document is revised, the knowledge is also revised. Therefore, the RAG system 4 can output answers based on the latest knowledge to the medical record management system 3 and reflect the answers in the electronic medical records.

[0124] In the first embodiment described above, the link between the medical record management system 3 and the RAG system 4 was described, but the receipt management system 2 and the RAG system 4 are also linked in a similar manner. For example, a receipt creation display screen similar to screens W1 and W2 shown in FIGS. 9 and 10 is displayed on the display unit 11. Therefore, when a staff member specifies an unknown item on the screen displayed when using the receipt management system 2, the answer obtained from the RAG system 4 is displayed on the display unit 11. In addition, a summary of the appropriate content is displayed from a document containing the outline of receipt entry. This ensures that receipts created by staff members are accurate, reducing the possibility of them being returned.

[0125] Furthermore, the knowledge upload unit 41 and the knowledge DB 51 may be removed from the RAG system 4 according to the first embodiment. In this case, the document (knowledge) sent to the RAG system 4 by the medical record management unit 31 of the medical record system 3 is sent to the embedding unit 43 by the knowledge update unit 42. The embedding unit 43 vectorizes the document (knowledge) and stores the vectorized knowledge in the vector DB 52. Even with this configuration, the knowledge search unit 44 can display the chunked text in the vector DB 52 on the screens W1 and W2, allowing staff to check the chunked text.

[0126] [Second embodiment] Next, a configuration example of a browsing support system according to a second embodiment of the present invention will be described with reference to Fig. 15. Fig. 15 is a functional configuration diagram showing an example of the overall configuration of a browsing support system 1A according to the second embodiment. The browsing support system 1A is capable of generating electronic medical records and providing answer support using RAG even in an offline state.

[0127] The browsing support system 1A includes an information processing terminal 10 and a medical record computer 6 for each dental clinic. The medical record computer 6 includes a medical record management system 60 and an RAG system 70. The medical record management system 60 includes a medical record management unit 61, a reference management unit 62, a medical record DB 63, a reference log DB 64, and a medical receipt DB 65.

[0128] The functions of the medical record management unit 61 and the reference management unit 62 are similar to those of the medical record management unit 31 and the reference management unit 32 according to the first embodiment. Therefore, the medical record management unit 61 can generate and manage electronic medical records offline.

[0129] The medical record DB 63 stores information of the electronic medical record created by the medical record computer 6. Since the information of the electronic medical record includes personal information of the patient, it remains within the medical record computer 6. Medical treatment data d1 is sent from the medical record computer 6 to the medical receipt management system 2, and the medical treatment data d1 is registered in the medical receipt DB 22. In addition, reference log data d2 is sent from the medical record computer 6 to the medical record management system 3, and the reference log data d2 is registered in the reference log DB 34.

[0130] The data configurations of the medical record DB 63, reference log DB 64, and receipt DB 65 are the same as the data configurations of the medical record DB 33, reference log DB 34, and receipt DB 22. The inside of the RAG system 70 is configured in the same way as the RAG system 4 shown in Fig. 4. The RAG system 70 is installed within a dental clinic, and there is no need to output data such as personal information outside the dental clinic.

[0131] The data in each DB constructed in the medical record computer 6 is limited to information that can be referenced by staff. Even when the medical record computer 6 is offline, staff can operate the information processing terminal 10 to view the information recorded in each DB in the medical record computer 6. Furthermore, of the information written in each DB of the medical record computer 6, information other than the medical record DB 63 is periodically synchronized with each DB provided in the medical receipt management system 2 and the medical record management system 3 and used.

[0132] In the browsing support system 1A according to the second embodiment described above, the information in each DB of the medical record computer 6 is synchronized with the information in each DB of the medical receipt management system 2 and the medical record management system 3. Therefore, even if the medical record computer 6 goes offline, the staff can check the information recorded in each DB of the medical record computer 6.

[0133] [Modification] In the above-described embodiments, the creation of medical records (electronic medical records) and medical receipts in dental treatment has been described, but the present invention can also be applied to the creation of medical records (electronic medical records) and medical receipts for treatment by doctors, nurses, physical therapists, chiropractors, etc. For example, not only in dentistry but also in medical, pharmaceutical, and veterinary care departments, staff from each department can check the electronic medical records.

[0134] The LLM 45 may also provide data summarizing the electronic medical record to the patient recorded in the electronic medical record. The patient can confirm the details of their treatment by checking the summarized data. Furthermore, if the patient visits another dental clinic, they can bring the summarized data with them to continue receiving treatment from their previous dental clinic.

[0135] The present invention is not limited to the above-described embodiments, and various other applications and modifications are possible without departing from the spirit and scope of the present invention as defined in the claims. For example, the above-described embodiments provide detailed and specific descriptions of the device and system configurations to clearly explain the present invention, and are not necessarily limited to those including all of the described configurations. Furthermore, it is possible to replace part of the configuration of one embodiment with the configuration of another embodiment, or even to add the configuration of another embodiment to the configuration of one embodiment. Furthermore, it is also possible to add, delete, or replace part of the configuration of each embodiment with other configurations. Furthermore, the control lines and information lines shown are those considered necessary for the explanation, and not all control lines and information lines are necessarily shown in the product. In reality, it can be assumed that almost all components are interconnected.

[0136] 1, 1A... Viewing support system, 2... Receipt management system, 3... Medical record management system, 4... RAG system, 10... Information processing terminal, 11... Display unit, 12... Input unit, 21... Receipt management unit, 31... Medical record management unit, 41... Knowledge upload unit, 42... Knowledge update unit, 43... Embedding unit, 44... Knowledge search unit, 45... LLM

Claims

1. A browsing support system in which an information processing terminal that displays medical-related information, a medical-related information management system that manages the medical-related information, and an input support system are connected via a network, wherein the information processing terminal has: a display unit that displays the medical-related information; and an input unit that creates questions about content to be registered in the medical-related information displayed on the display unit, and the medical-related information management system has: a creation support unit that outputs creation information for creating the medical-related information that is input by the information processing terminal to the input support system, receives answers from the input support system, and supports the creation of the medical-related information, and the input support system has: a vector database that stores knowledge that has been vectorized and embedded in a vector space, and a knowledge search unit that searches the vector database for the knowledge based on the vectorized creation information, and outputs the answer formed from the searched knowledge to the creation support unit.

2. The browsing support system according to claim 1, wherein the knowledge search unit extracts the vectorized knowledge based on the similarity between the vectorized knowledge stored in the vector database and the vectorized creation information.

3. The browsing assistance system according to claim 2, wherein the input assistance system comprises a large-scale language model that reads reference information and summarizes the reference information based on prompts consisting of text or images, and the knowledge search unit outputs a summary document summarized by the large-scale language model to the medical-related information management system.

4. The browsing assistance system described in claim 3, wherein the creation information is information contained in a question entered in a display area of ​​a medical-related information creation display screen used to create the medical-related information, or information contained in an item selected on the medical-related information creation display screen, and the answer includes the summary document and information on the reference source of the reference information summarized by the large-scale language model.

5. The browsing assistance system according to claim 4, wherein the creation information includes at least one of text, audio, and images.

6. The browsing assistance system according to claim 5, comprising: a knowledge update unit that updates the knowledge; and an embedding unit that vectorizes the updated knowledge, embeds it in the vector space, and stores the vectorized knowledge in the vector database.

7. The browsing support system according to claim 5, wherein the input support system comprises: a knowledge database that stores the knowledge; and a knowledge upload unit that uploads the knowledge to the knowledge database.

8. The browsing assistance system according to claim 6, wherein the input assistance system comprises: a knowledge database that stores the knowledge; and a knowledge upload unit that uploads the knowledge to the knowledge database.

9. A browsing assistance method carried out in a browsing assistance system in which an information processing terminal having a display unit that displays medical-related information and an input unit that edits the medical-related information displayed on the display unit, a medical-related information management system that manages the medical-related information, and an input assistance system are connected via a network, the browsing assistance method comprising: a step in which a creation assistance unit of the medical-related information management system outputs creation information for creating the medical-related information, which is input by the information processing terminal, to the input assistance system; a step in which the input assistance system vectorizes the creation information; a step in which the knowledge is searched for based on the vectorized creation information from a vector database in which vectorized knowledge embedded in a vector space is stored; a step in which the searched knowledge is shaped into an answer and output to the creation assistance unit; and a step in which the creation assistance unit receives the answer from the input assistance system and supports the creation of the medical-related information.

10. A computer-readable storage medium storing a program for causing a computer to execute the following steps: vectorizing creation information input by an information processing terminal for creating medical-related information; searching for knowledge based on the vectorized creation information from a vector database in which vectorized knowledge embedded in a vector space is stored; and outputting an answer formed from the searched knowledge to the information processing terminal.

Images