Medical conversation support

Abstract

Providing appropriate responses that take into account past conversation history during patient interactions is a significant burden for healthcare professionals, requiring the maintenance of consistency and appropriateness in their speech. Furthermore, in situations where speech control based on treatment plans or ensuring informed consent is required, adjustments to the content of speech become necessary. [Solution] The present invention provides a medical conversation support system that performs the following processes: creating utterances based on conversation history, searching and matching them; directly searching and uttering from conversation history; and creating utterances using individual key point data. It generates utterance examples using individual dialogue data and individual key point data, and selects the optimal utterance by comparing it with a medical professional utterance database. For utterance selection, keyword matching and vectorization evaluation using natural language processing are utilized to calculate the similarity between individual utterance examples and utterance examples in the utterance database and determine the optimal one.

Description

Technical Field

[0001] The present invention relates to dialogue support technology in the medical field, and relates to a tool (system, program, method) for assisting medical staff in conducting conversations with patients smoothly and accurately.

Background Art

[0002] Conventionally, there have been many technologies with the function of predicting the next utterance based on the conversation history in order to support daily conversations. In recent years, technologies with the function of supporting utterances by generative AI have emerged, for example, as disclosed in Patent Document 1.

[0003] Patent Document 1 is a dialogue support device that determines an utterance example to be presented to the user from among the generated utterance examples by referring to a database that stores the dialogue history for each user. The generated utterance examples correspond to the target function, and by using the dialogue history to calculate the utilization rate of the target function, the aim is to present highly useful utterance examples based on the user's usage situation. It is also described that the utterance examples may be generated by extracting utterance examples from the user utterances included in the dialogue history to generate utterance examples corresponding to the target function.

[0004] That is, Patent Document 1 provides utterance examples regarding a technology realized by a terminal capable of executing a function corresponding to a user utterance, and is fundamentally different from that which supports conversations between medical staff and patients, as will be described later.

[0005] In the medical field, in interviews (interviews) with patients, online consultations, the use of assessment tools such as checklists and screening tests, and counseling situations, the conversation with the patient is one of the important factors that affect the quality of medical treatment. Through the conversation, it is required to accurately grasp the patient's situation, complaints, and symptoms, and to determine appropriate diagnosis and treatment policies. In the conversation, while considering maintaining consistency and providing appropriate information, even if the conversation continues for a long time, the content of the medical staff's utterance must be appropriate. [Prior art documents] [Patent Documents]

[0006] [Patent Document 1] Patent No. 7541533 (Paragraph

[0025] etc.) [Overview of the project] [Problems that the invention aims to solve]

[0007] However, many existing dialogue support technologies aim to predict utterances for general users and provide example utterances based on dialogue history, and do not adequately address the specific characteristics of the medical field. For example, the technology described in Patent Document 1 primarily extracts example utterances based on the user's utterance history, focusing on providing utterances related to specific functions. In contrast, medical dialogue requires not just assistance in function execution, but the provision of appropriate utterances in real time according to the patient's condition. Furthermore, since the content of conversations with patients is highly dependent on the context of the situation, it is difficult for conventional dialogue support technologies to maintain appropriate medical conversations.

[0008] Furthermore, the utterance examples generated by Patent Document 1 are extracted from user utterances included in the dialogue history, targeting the history of unspecified users, and are not intended to support dialogue with specific users. Moreover, the appropriateness of the generated utterance examples is judged based on indicators such as first usage rate and performance information, and does not directly make judgments based on the dialogue history. Rather, the dialogue history is used indirectly to calculate indicators such as first usage rate and performance information.

[0009] Furthermore, healthcare professionals need to process a lot of information while communicating with patients during the course of treatment. However, it is a significant burden to choose appropriate words while considering each patient's different background and medical history. In particular, when fatigued or under time constraints, the accuracy of speech tends to decrease, which can lead to problems where important information is not properly conveyed.

[0010] This invention has been made in view of these problems and provides a medical conversation support tool (system, program, method) to help healthcare professionals communicate with patients appropriately and efficiently.

[0011] The first process (conversation history → creation → search), which creates individual utterance examples based on conversation history (individual dialogue data) and performs search and matching, can provide new utterances that take into account the individual patient's past utterances, and allows for flexible utterance provision by selecting appropriate expressions through matching with a medical professional utterance database. The second process (conversation history → search → utterance), which directly searches and performs utterances from conversation history, allows for highly real-time utterance provision by inputting past conversation history (individual dialogue data), directly searching and matching it, and obtaining associated appropriate utterances. This eliminates the need for utterance creation and allows for immediate selection of appropriate utterances through search. Furthermore, the third process, which utilizes individual key point data, eliminates unnecessary information and provides utterances that focus on the essentials, enabling more accurate utterance provision compared to the first process. In addition, the contradiction evaluation function, notification control function, and explanation control function can improve the quality of medical conversations with patients by ensuring appropriate informed consent. [Means for solving the problem]

[0012] This invention is a tool to support dialogue between healthcare professionals and patients and to enable appropriate and consistent medical conversations. This system has integrated functions of speech creation, retrieval, matching, evaluation, and control, and by providing advanced support for medical conversations, it differs from conventional simple speech history retrieval technology in that it realizes integrated medical conversation support of speech creation, retrieval, contradiction evaluation, control, and customization, making it extremely practical in medical settings.

[0013] (1) Utterance creation and search matching (Claim 1, 2, 3) Claim 1: Based on the past conversation history with an individual patient, a speech generation program generates new speech and further compares it with a medical professional speech database to provide appropriate speech examples. Claim 2: To improve real-time performance by directly searching the patient's dialogue history and immediately providing similar existing utterance examples. Claim 3: By utilizing an individual key points database and extracting and organizing key points for each patient, it supports more concise and accurate speech. By linking with data from electronic medical records and referral letters, it is possible to provide more contextually relevant speech.

[0014] (2) Management of the context in which the utterance is applied (Claim 4) It provides appropriate utterances tailored to various medical scenarios, such as initial consultations, follow-up visits, hospitalization, and admission / discharge support centers, enabling optimal conversations in each situation.

[0015] (3) Inconsistency evaluation and policy control (Claim 5, 6, 7) Claim 5: A function to evaluate conversational inconsistencies, which adjusts utterances so as not to be inconsistent with the patient's existing information or past conversations. Claim 6: For example, if a policy of "not disclosing cancer information" is set, search control is performed so that this information is not included in the conversation. Claim 7: Controls are in place to ensure that explanations are given before consent is given during the examination explanation, and informed consent is thoroughly implemented.

[0016] (4) Integration and customization functions with medical systems (Claim 8, 9, 10) Claim 8: The template information, checklist information, and speech data of the electronic medical record can be associated and used as structured data. Electronic Medical Record Profile Information: This refers to profile data within the electronic medical record that contains basic patient information and individual information necessary for treatment. Example: Patient's name, age, sex, medical department (internal medicine, surgery, etc.), medical history (diabetes, hypertension, etc.), allergy information, family history (presence or absence of genetic disorders, etc.) Current treatment plan (conservative therapy, planned surgery, etc.) Electronic medical record template information: This refers to the template information for medical records registered in the electronic medical record system. Examples: Consultation record templates (initial consultation, follow-up consultation, follow-up, pre- and post-surgery formats), test result report templates, prescription record templates, and medical record formats for hospital admission and discharge. Task Information: This refers to tasks (medical work and instructions) managed by healthcare professionals using electronic medical records or clinical support systems. Examples: "Explain MRI results at the next appointment," "Provide medication guidance to the patient," "Obtain informed consent before surgery," "Instruct the patient to measure blood pressure at the next appointment." Checklist Information: This refers to checklist information that lists items to be confirmed during a medical consultation. Examples: Pre-operative checklist (allergy confirmation, fasting instructions confirmation), hospital admission interview checklist (medical history, current medication status confirmation), pre-examination explanation checklist (contrast agent use, side effect explanation), discharge instruction checklist (dietary guidance, medication guidance, confirmation of next appointment date). Ordering information: This refers to information that manages the order and priority of utterances in medical conversations. Example: Always explain the examination in the first half of the consultation; strictly adhere to the order of pre-operative explanation → confirmation of surgical procedure → obtaining consent. Maintain a standard order of answers to patient questions and prioritize verbal responses according to the consultation flow (interview → physical examination → tests → explanation of results). In this invention, by linking this information with a medical professional speech database, appropriate speech is provided in accordance with the flow of medical treatment, thereby optimizing the medical treatment process. In particular, its innovative feature lies in its ability to provide dynamic speech tailored to each patient's situation and treatment task, thereby reducing the burden on healthcare professionals while enabling consistent medical conversations. Thus, claim 8 achieves optimization of speech and improved visibility of medical treatment through integration with medical data, thereby greatly contributing to the standardization and efficiency of medical treatment. Claim 9: It is equipped with a mail merge function that allows for the replacement of hospital names, department names, etc., making it easy to customize for different medical institutions. Long-form explanation content: General medical explanations (procedures for examinations, precautions during hospitalization, post-operative care, etc.) are registered to ensure consistency in speech. Insert document function: Register hospital names, department names, doctor names, medical department names, etc. in a templated and replaceable format as appropriate. Claim 10: New utterance examples can be generated and registered from electronic medical record information and case reports, and the utterance data can be continuously updated.

Advantages of the Invention

[0017] According to the present invention, medical staff can provide new utterances based on the patient's past utterance history, and by performing search and comparison with a standard medical conversation database, an appropriate utterance sentence can be selected, enabling appropriate communication optimized for each patient. In particular, the provision of utterances utilizing individual key data leads to highly accurate utterance sentences, contributing to an improvement in the reliability of medical treatment.

[0018] In addition, it becomes possible to realize smooth and accurate medical conversations and provide a beneficial medical environment that reduces the burden on medical staff.

Brief Description of the Drawings

[0019] [Figure 1] Diagram showing the network configuration of the conversation support system. [Figure 2] Block diagram showing an example of the system configuration of the conversation support system. [Figure 3] Block diagram showing an example of the functional configuration of an in-hospital terminal. [Figure 4] Block diagram showing an example of the functional configuration of a system server. [Figure 5] Block diagram showing an example of the functional configuration of a processing means. [Figure 6] Flowchart showing the processing procedure of the utterance creation program of the support system. [Figure 7] Flowchart showing the processing procedure of the utterance creation program of the support system.

Best Mode for Carrying Out the Invention

[0020] Figure 1 shows the network configuration centered on the conversation support system 2. In addition to the computer system server running the conversation support system 2, the hospital terminals 1 (telephones, mobile phones, telecommunications terminals) and the medical system 4 owned by the medical institution are connected by a telephone / communication network via telecommunication lines. Patients 3 can converse with medical professionals via telephone, chat, etc. using their own telephones or telecommunications terminals through the hospital's terminals 1, and patients and medical professionals can converse face-to-face. In face-to-face conversations, it is sufficient for medical professionals to collect the content of the conversation with the patient by operating the hospital terminal 1. For example, by recording the conversation with the patient using a telecommunications terminal such as a tablet, and displaying the next example utterance on the tablet screen for the medical professional to speak, the conversation with the patient can proceed smoothly. Alternatively, a simple voice recorder with communication functions may be used to support the speech of medical professionals who are remotely conversing with patients online.

[0021] Medical system 4 includes systems such as appointment scheduling, electronic medical records, and electronic patient interviews. Each of these devices is connected to the internet / intranet / hospital LAN, etc., to form a network. Note that the control lines and information lines shown are those deemed necessary for explanation and do not necessarily represent all control lines and information lines in the product, and all components may be interconnected.

[0022] Conversation support system 2 is a computer that acquires the content of conversations between medical professionals and patients as medical conversation data, and also creates and generates the content of speech spoken by medical professionals. By linking with medical system 4 via a network, conversation support system 2 can also create and generate speech content utilizing electronic medical record data and interview data.

[0023] Figure 2 is a block diagram showing an example of the configuration of conversation system 2, which is a computer connected to a network such as the Internet. It includes a communication interface 22, an input / output interface 23, memory 25, storage 26, and a processor 29.

[0024] Communication IF22 is an interface for inputting and outputting signals so that the system 2 can communicate with external devices. Input / Output IF23 functions as an interface to an input device for receiving input operations from the user and an output device for presenting information to the user. Memory 25 is for temporarily storing programs and data processed by programs, etc., and is a volatile memory such as DRAM. Storage 26 is a storage device for saving data, such as flash memory, HDD, or SSD. Processor 29 is hardware for executing the instruction set written in the program, and is composed of an arithmetic unit, registers, peripheral circuits, etc.

[0025] Note that "processor" refers to one or more processors. At least one processor is typically a microprocessor such as a CPU (Central Processing Unit), but may be other types of processors such as a GPU (Graphics Processing Unit). At least one processor may be single-core or multi-core. Furthermore, at least one processor may be a broader type of processor, such as a hardware circuit that performs some or all of the processing (e.g., an FPGA (Field-Programmable Gate Array) or ASIC (Application Specific Integrated Circuit)).

[0026] Figure 3 is a block diagram showing an example of the functional configuration of the in-hospital terminal 1. Regardless of whether it is a telephone, mobile phone, or telecommunications terminal, it includes a communication IF 12, an input device 13, an output device 14, a memory 15, a storage unit 16, and a processor 19, or implements equivalent functions. Note that the telephone (mobile phone) may be connected to the network via an in-hospital exchange that functions as a gateway for connecting to the public telephone network or the internet. The communication interface is an interface for inputting and outputting voice and signals so that the in-hospital terminal 1 can communicate with external devices via telephone. It also functions as an interface to an input device 13 for receiving user input operations and an output device 14 for presenting information to the user. The input device 13 is an input device (such as a keyboard, touch panel, touchpad, mouse, or other pointing device) for receiving user input operations. The output device 14 is an output device (such as a display or speaker) for presenting information to the user. Memory 15 is for temporarily storing programs and data processed by programs, etc., and is a volatile memory such as DRAM. Storage unit 16 is a storage device for saving data, such as flash memory, HDD, or SSD. Processor 19 is hardware for executing the instruction set written in the program, and is composed of an arithmetic unit, registers, peripheral circuits, etc.

[0027] In-hospital terminal 1 connects to the network by communicating with communication devices such as wireless base stations compatible with various communication standards including LTE, and wireless LAN routers compatible with IEEE and wireless LAN standards. If in-hospital terminal 1 is a desktop or laptop PC, it acquires voice signals using broadband or fiber optic lines. If it is a mobile device such as a tablet or smartphone, it can make calls and data communications using a mobile phone line (cellular network), and instead of a fixed telephone line, it connects to a mobile phone network such as 4G or 5G and utilizes mobile communication infrastructure. Furthermore, it is also possible to make calls via the internet using a VoIP application.

[0028] Patient 3's device is a tablet or smartphone (registered trademark), such as a telephone, mobile phone, or telecommunications terminal, and has generally known and general-purpose functions that allow for conversations and communications with medical personnel operating the in-hospital terminal 1.

[0029] Figure 4 is a block diagram showing an example of the functional configuration of the server 20 of the system terminal 2. The server 20 comprises a communication means 220, a storage means 280, and a control means 290, with each block being electrically connected by a bus or the like. The communication means 220 implements the functions of the communication IF 22, the storage means 280 implements the functions of the memory 25 and storage 26, and the control means 290 implements the functions of the processor 29.

[0030] The communication means 220 performs modulation and demodulation processing to enable the system 2 to acquire voice and communication data from other terminals, such as the medical system 4 and the in-hospital terminal 1 / patient 3, and transmits the signal calculated by the control means 290. The communication means 220 performs reception processing on the signal received from the outside and outputs it to the control means 290. In this way, the communication means 220 interprets commands or input content and provides them to each means, and also functions as an interface that interprets various display commands issued from the storage means 280 and performs output control.

[0031] The storage means 280 is implemented by memory (RAM) 25 and storage 26 such as a disk device (floppy disk, hard disk, or magneto-optical disk, etc.) and stores data, programs, etc. used by the system 2. The storage means 280 holds and stores various data in the application program 282 of the system, as well as in the work area 281, data storage area 283, and screen definition storage area 284.

[0032] The work area 281 is allocated upon startup of this system and is an area where various data input and output by this system are temporarily stored. The data storage area 283 is an area where data temporarily stored in the work area 281 is semi-permanently stored through write control when a save request is made. The screen definition area 284 is an area where screen definition information for various screens to be output and displayed on the display devices of the hospital terminal 1 and user 3's terminal is pre-stored and includes format information for the screen settings to be displayed.

[0033] The input device 230 is a device used by the user operating the system 2 to input instructions or information, and can be a keyboard, mouse, reader, touch-sensitive device, etc. The input device 230 also converts the instructions input by the user into electrical signals and outputs the electrical signals to the control means 290. The input device 230 also includes a receiving port that accepts electrical signals input from external input devices. The output device 240 is a display device 241 such as an LCD or organic EL for presenting information to the user operating the system 2. The display 241 can display data according to the control content of the control means 290 and can check the communication status with other external devices.

[0034] The control means 290 is realized when the processor 29 reads the application program 282 stored in the storage means 280 and executes the instructions contained in the application program 282. The control means 290 also controls the operation of the system 2 and, by operating according to the application program 282, performs the functions of an input means 291, a user interface 292, a display control means 293, and a processing means 294.

[0035] The data storage area 283 stores various data to support medical conversations, including a medical conversation DB 283A that holds medical conversation data, individual conversation data 283C that holds the individual conversation content for each patient, and individual key point data 283C1 which extracts only the key points from the individual conversation data 283. The medical conversation DB 283A holds the medical professional utterance DB 283A1 and patient-medical professional conversation data 283A2, and the medical professional utterance DB 283A1 includes the content of examinations 283B1 and the content of notifications 283B2, as well as the content of examinations and notifications 283B2, in accordance with the medical guidelines / treatment protocol DB 283B. This data storage area is not merely for record keeping, but also functions as a data processing means for this support system 2, performing real-time conversation suggestions, consistency checks with past information, and presentation of utterances in accordance with standard medical protocols.

[0036] Medical Conversation DB283A contains information necessary to support dialogue in medical settings. This database is further subdivided and holds Medical Professional Utterance DB283A1 and Patient-Medical Professional Conversation Data 283A2. Medical Professional Utterance DB registers standard utterances used by medical professionals to conduct medical conversations during examinations and explanations. These standard utterances include content compliant with medical guidelines and treatment protocols DB283B, and organize standard examination explanations 283B1 and medical condition disclosures 283B2. Individual Dialogue Data 283C holds the dialogue history for each patient. To ensure continuity of care and enable responses based on past interactions, the content of dialogues with individual patients is recorded chronologically. Furthermore, Individual Key Points Data 283C1 is stored, which extracts only the most important elements (key points) from the Individual Dialogue Data 283C, such as the patient's chief complaint and turning points in treatment plan. The medical professional speech database 283A1 includes speech examples from the initial consultation, follow-up consultations, hospitalization, and during hospitalization. Furthermore, the medical professional speech database 283A1 is registered in relation to the patient-healthcare professional conversation data 283A2, meaning that the medical professional speech database 283A1 is not simply accumulating speech examples, but is managed in relation to the patient-healthcare professional conversation data 283A2. In other words, because the speech examples are registered considering individual contexts and the flow of medical treatment, it can provide appropriate speech according to the specific dialogue situation with the patient, rather than just being a generic template for medical speech.

[0037] Figure 5 is a block diagram showing an example of the specific functional configuration of the processing means 294, and includes a speech creation processing unit 294A, a speech search and matching processing unit 294B, a conversation contradiction evaluation processing unit 294C, an explanation control processing unit 294D (policy control function and inspection explanation function), an individual key point extraction processing unit 294E, and a dialogue history management unit 294F (dialogue history management function and search processing function).

[0038] The speech creation processing unit 294A extracts and generates new speech examples that are natural and contextual based on past dialogue data with individual patients and medical conversation data. It implements the function of a speech creation program that outputs candidate speech examples as input data according to past conversation content and the patient's situation. The speech search and matching processing unit 294B searches and matches the medical professional speech database 283A1 based on the dialogue history with individual patients and selects additional speech examples. Furthermore, the speech creation program implements the function of inputting individual key point data and individual dialogue data and outputting additional individual speech examples in either the speech creation processing unit 294A or the speech search and matching processing unit 294B.

[0039] The conversation inconsistency evaluation processing unit 294C detects whether an additional utterance contradicts one or more individual key point data or individual dialogue data in order to maintain the consistency of the conversation, and evaluates the degree of inconsistency. It detects inconsistencies to ensure that utterances are consistent with treatment policies and medical records, and if an inconsistency is detected, it can also suggest correction options.

[0040] The explanation control processing unit 294D combines policy control and examination explanation functions to manage adherence to medical policies and appropriate explanations for individual patients. The policy control function controls speech to ensure that it follows a specific treatment policy and adjusts speech content based on data regarding the necessity of examination explanations and the necessity and appropriateness of the disclosure content. For example, if the disclosure NG flag is set in the disclosure 283B2 data for a patient who is "not subject to cancer disclosure," the disclosure-related speech is excluded. The examination explanation function supports appropriate explanations about examinations that patients will undergo, enabling healthcare professionals to make appropriate speech based on treatment policies and examination procedures. For example, if the examination 283B1 data includes "contrast agent will be used in the next MRI examination," the function supports speech regarding pre-examination dietary restrictions and appropriately explains the purpose, risks, and procedures of the examination.

[0041] The individual key point extraction processing unit 294E extracts important elements (key points) in medical treatment from the dialogue data with the individual patient, which holds the individual dialogue data 283C, and stores them in the individual key point data 283C1. The individual dialogue data 283C is updated in real time, and the key points stored in the individual key point data 283C1 may be changed sequentially as desired. In this case, the individual key point data 283C1 may be temporarily held in the work area 281. For example, if the patient says, "I'm worried about the side effects of the medication," "concerns about side effects" is recorded as individual key point data, and utterance examples containing the content of "side effects" are extracted or generated by search as candidate utterance examples for the next utterance.

[0042] The Dialogue History Management System 294F combines dialogue history management and search processing functions to properly manage past dialogue history and allow it to be searched as needed. The dialogue history management function stores and organizes each patient's dialogue history and medical history, making it easily accessible to healthcare professionals. The search processing function quickly searches past dialogue history, medical history, and individual key point data to retrieve the necessary information.

[0043] The flowcharts illustrating the processing procedure of the speech generation program of this system are explained in Figure 6: Example of the first embodiment and Figure 7: Example of the second embodiment.

[0044] [Figure 6: First Embodiment] In step S601, based on the commands of the dialogue history manager 294F, the individual dialogue data 283C and individual key point data 283C1 are referenced to obtain the dialogue history / dialogue key points. The dialogue key points may be obtained as individual key point data 283C1 by performing a real-time process to extract the necessary key points from the individual dialogue data 283C. Next, in step S602, the speech creation processing unit 294A executes the speech example output process to output a speech example. The speech creation processing unit 294A extracts and generates a new speech example based on the acquired dialogue history / dialogue key points. The functions of the speech creation processing unit 294A will be described later.

[0045] In step S603, when an utterance example is output, in S604, the utterance search and matching processing unit 294B performs a search or matching based on the match or similarity in the medical professional utterance DB 283A1 and obtains the search result or matching result. The utterance search and matching processing unit 294B evaluates the one that matches or is most similar (the one with the highest similarity) from among the matched utterance example candidates and determines whether the utterance sentence is applicable. If no suitable utterance example is found, the utterance creation processing unit 294A creates another utterance example. The suitable utterance example obtained from the search result or matching result is output via the in-hospital terminal 1 as an utterance sentence to be presented to the patient (step S606). The method for evaluating the one with the highest match or similarity is based on evaluation using text matching technology, evaluation using natural language processing technology, or an evaluation method that combines these. Text matching can use keyword-based search as a simple method, extracting the main words contained in individual utterance examples (e.g., "examination," "side effect," "blood pressure," etc.) and comparing them with the words contained in utterance examples in the medical professional utterance DB. Alternatively, an embedded model utilizing natural language processing can be used for similarity evaluation using vector representations. Additional individual utterance examples and utterance examples from the healthcare professional utterance database are converted into vectors (numerical representations), the cosine similarity between the vectorized utterance sentences is calculated, and the one with the highest score is selected as the optimal utterance example candidate. Furthermore, weighting based on medical guidelines and clinical protocols can be applied to take into account constraints specific to medical conversation. For example, "diagnostic utterances" may be judged to be more important than "everyday conversation utterances," and the similarity score may be adjusted accordingly.

[0046] In step S604, the candidate utterances deemed appropriate undergo further contradiction evaluation by the conversation contradiction evaluation processing unit 294C and policy control processing by the explanation control processing unit 294D (step S605), and are output as the final utterance (step S606).

[0047] Next, the system determines whether additional utterances are necessary (step S607). If additional utterances are deemed necessary, the system returns to step S601 to obtain the content of the additional utterances. On the other hand, if additional utterances are deemed unnecessary, the system obtains an example of the utterance that concludes the conversation as the final utterance, and records a log indicating that all utterances have been completed (step S608).

[0048] [Figure 7: Second Embodiment] In step S701, based on the instructions of the dialogue history manager 294F, the system references individual dialogue data 283C and individual key point data 283C1 to obtain dialogue history / dialogue key points. Dialogue history / dialogue key points may be obtained from both or both. Next, in step S702, based on the instructions of the speech search and matching processing unit 294B, the system searches and matches the medical professional speech database 283A1 and patient-medical professional conversation data 283A2. That is, based on the obtained individual dialogue data 283C and individual key point data 283C1, the speech search and matching processing unit 294B searches and matches appropriate candidate utterances from medical conversations. Searching and matching within the healthcare professional speech database, which is linked to patient-healthcare professional conversation data, is performed using the healthcare professional speech database 283A1, which is classified by medical scenario (initial consultation, follow-up consultation, explanation of tests, explanation of treatment plan, etc.) and situation, and the patient-healthcare professional conversation data 283A2, which is linked to past interactions for each patient. The speech search and matching processing unit 294B takes past dialogue data 283A2 as input and searches the related healthcare professional speech data 283A1 to obtain the most appropriate candidate utterance. This association allows for the search of optimal utterances that take into account the patient's past dialogue history, rather than just referring to standard utterance examples.

[0049] Next, additional associated utterance examples are extracted (step S703). After utterance examples are extracted in step S703, the conversation contradiction evaluation processing unit 294C and the explanation control processing unit 294D perform policy control processing (step S704), and then the final utterance is output (step S705).

[0050] Next, the system determines whether additional utterances are necessary (step S706). If additional utterances are deemed necessary, the system returns to step S701 to obtain the content of the additional utterances. On the other hand, if additional utterances are deemed unnecessary, the system obtains an example of the utterance that concludes the conversation as the final utterance, and records a log indicating that all utterances have been completed (step S707).

[0051] [Speech creation process] The program-based creation process by the speech generation processing unit 294A is suitable for providing standard speech examples based on medical guidelines and treatment protocols, as well as fixed speech examples for each treatment phase. Furthermore, the speech generation process utilizing the generation AI by the speech generation processing unit 294A can generate more flexible and individualized speech examples by considering the patient's past conversation history and the context of the dialogue. Considering this adaptability, the speech generation processing unit 294A can choose to determine standard speeches using rule-based program processing or generate additional context-adapted speeches using AI processing, or use both methods in combination. Furthermore, by setting conditions for the speech creation processing unit 294A to create, search, and match speech examples, such as the phase of treatment (initial consultation, follow-up consultation, examination explanation, hospitalization, discharge, pre-operative explanation, post-operative follow-up, etc.), diagnosis content / symptoms (hypertension, diabetes, heart disease, cancer, allergies, etc.), examinations / procedures (MRI, CT, blood tests, endoscopy, surgery, etc.), prescription / medication information (medications currently being taken, start of medication, change of medication, side effects, etc.), patient's speech content (questions / complaints), and medical policies / guidelines (cancer disclosure policy, restrictions on explaining specific diseases, in-hospital rules, etc.), the risk of unintended speech examples being output can be reduced, and more accurate speech creation can be achieved.

[0052] Furthermore, this system may include a conversation trend analysis program that analyzes the conversation history database and extracts conversational trends, and the conversation trend analysis program may provide healthcare professionals with the following utterances based on the extracted conversational trends. [Example 1]

[0053] (1) How to create a conversation with a first-time patient Background: Finish the conversation between medical staff and the patient at the initial diagnosis before the patient's visit. <STEP1: Preparation of the medical staff's utterance database and appropriate registration of utterance examples (Claims 1, 4)> Overview: Register a large number of utterance examples (such as inquiry phrases and explanation phrases required at the initial diagnosis) used by medical staff for a large number of symptoms in the medical staff's utterance DB283A1. The types of conversations are open questions and closed questions. Objective: Accumulate standard and typical conversation patterns (such as inquiry items and explanation items) for dealing with patients at the initial diagnosis, and make them referable in subsequent steps. <STEP2: Input of conversation data between patient and medical staff and creation of additional utterance sentences (Claims 1, 3)> Overview: Input the conversation data (such as the patient's chief complaint, symptoms, past history, etc.) exchanged between an individual patient and medical staff into the utterance creation program, and automatically output additional utterance examples. Specific example: · First, ask "What brings you here today?" · Suppose the patient says "I've been feeling feverish since this morning" and a conversation ensues. → Based on this conversation, the program converts "feeling feverish" into the medical term "fever", and identifies that the patient's expression is {feeling feverish} and the medical topic is "fever". The utterance creation program automatically generates additional phrases such as "Are there any other symptoms besides fever?" <STEP3: Search using the output result of the utterance creation program (Claim 1)> Overview: Use the "additional utterance examples" (additional phrases) created in STEP2 as input, and search and match the medical staff's utterance database 283A1 based on identity or similarity to obtain the corresponding utterance examples. Specific example: · Search and match the database to see if there are any sentence examples similar to the additional utterance example "Are there any other symptoms besides fever?" → Obtain the utterance example "{{symptom}}. Do you have any other symptoms?" which shows a high degree of similarity. Replace "{{symptom}}" with "having a fever", and obtain the phrase "Do you have any other symptoms besides having a fever?" By doing this, it becomes possible to have a conversation using the words the patient himself / herself used in the conversation. <STEP4: Output the search results to the first-time visiting patient (Claim 1)> Summary: Present the obtained utterance examples to the patient either directly or after the doctor or nurse checks them on the screen and makes minor adjustments as necessary. In this way, a thorough inquiry and explanation without omission can be carried out at the first visit. Since the system presents standard phrases as candidates, check whether there is no misinformation or the like from medical staff. This contributes to reducing the burden and improving the efficiency of medical staff. <Industrial effectiveness of Example 1> By applying the present invention to the inquiry of first-time visiting patients, the inquiry work of medical staff can be made more efficient and information leakage can be reduced. Also, by comparing with standardized utterance examples, appropriate, safe, and consistent medical communication can be realized. Furthermore, it is easy to expand to other hospitals, home medical care, call centers, etc., and the industrial application range is wide and the effectiveness is high.

Example 2

[0054] (2) Method for creating a conversation with a follow-up patient Background: Finish the conversation between medical staff and the patient in advance at the follow-up visit. <STEP1: Prepare a database of medical staff's utterances for follow-up visits (Claims 2, 4)> Summary: In past follow-up cases, register the phrases that medical staff / doctors usually check and ask as "next utterance examples" in the medical staff's utterance DB283A1. · For example, prepare utterance examples for matters commonly asked at follow-up visits, such as "How was the effect of the medicine prescribed last time?" "Did you feel any side effects?" · Also, as an utterance to end the conversation, register "Thank you very much for telling me so much. Since I will convey this content to the doctor, could you please wait for a while at {{waiting area}}?" Aim: Develop a "Database of Conversation Examples at Follow-up Visits" that asks all necessary questions without omission for follow-up patients and make it searchable in subsequent steps. <STEP2: Input Conversation Data at Follow-up Visits with Individual Patients (Claim 2)> Overview: Input the conversation during the follow-up visit with a patient who has already been examined as individual dialogue data into the system. For example, import "Course Since the Previous Visit", "Previous Prescribed Medication", "Patient's Physical Condition Memo", etc. recorded in the electronic medical record, and input the conversation situation during this follow-up visit ("Cough still persists slightly", "The fever has subsided but headache remains", etc.). Specific Example: · Information such as "I finished taking the medicine prescribed last time, but still have a little cough" is obtained from the patient. · The system recognizes this as "Individual Dialogue Data at Follow-up Visits" and prepares to search for additional questions and confirmation items to be asked next. <STEP3: Search and Match Based on Individual Dialogue Data at Follow-up Visits (Claim 2)> Overview: Based on the above-mentioned "(at follow-up visit) individual dialogue data", search or match the medical staff's conversation database to obtain "Next Conversation Example". In this embodiment, instead of automatic generation by AI, the optimal one is extracted from the "Question Phrases at Follow-up Visits" registered in the database. Specific Example: · The system uses as input the dialogue data stating "Doctor: How was the effect of the medicine prescribed last time? Patient: I finished taking the medicine prescribed last time, but still have a little cough". · Match the database to obtain one or a list of "Standard Questions to Ask Patients with Persistent Cough during Follow-up Visits" and "Side Effect Confirmation Phrases". Example: · "Did you have any side effects after taking the medicine?" · "How often does the cough occur?" · "If the cough at night seems severe, it may be necessary to prescribe another medicine additionally. How do you feel about that?" <STEP4: Reflect and Output the Obtained Results in the Conversation with the Patient (Claim 2)> Summary: The system presents the "next conversation example" obtained in STEP3 by voice or on the screen, and presents it to the patient after being corrected by the doctor or medical staff as necessary. To ensure the smooth progress of the interview in natural language, the doctor actually uses the selected phrases with the patient. Conversation example (natural language): Doctor: "After that, how often does the cough occur?" Patient: "I often cough at night. It doesn't happen much during the day." Doctor: "I see. It only gets worse at night. Have you taken any other medicine and felt any side effects?" Patient: "There were no particular side effects." ※ Such a process proceeds by the doctor actually using the phrases extracted from the database in the actual conversation. <STEP5: Determine comprehensiveness based on the output content> Summary: Based on the questions from the patient obtained in STEP4, check the database of conversation examples at the follow-up visit for the answers to the questions, determine whether all the answers to the questions have been obtained according to the follow-up protocol, and return to STEP2 to make additional statements if necessary. If it is determined that all the answers to the questions have been obtained, present the statements registered as statements for ending the conversation to end the conversation (steps S608, S707). Effect: Appropriate, safe, and standardized follow-up interview Points that medical staff tend to overlook, such as the progress of the previous visit and the confirmation of the effect of the prescription drug, are also reminded by database search, so the quality of the follow-up visit is improved. Reduction of the burden on medical staff By automatically listing the necessary items based on past conversation data, doctors do not need to talk about everything from scratch, and an efficient follow-up visit is possible. Industrial applicability and effectiveness in implementation: · According to this embodiment, by accumulating and sharing in the database the confirmation items (such as side effects, changes in symptoms, and the necessity of the next examination) that medical staff usually perform during the follow-up interview, an interview without omission or duplication can be realized. · The system can operate a consistent reexamination protocol and can also be applied to home medical care, house calls, and patient information sharing among multiple hospitals. · It is possible to standardize the "additional questions at the time of reexamination," which conventionally relied on the personal experience of doctors, thus contributing to industrial efficiency improvement and medical quality improvement.

Example 3

[0055] (3) Customization for Each Medical Institution · Insertion Sentences (Claim 9) Overview: In multiple medical institutions (such as Hospital A, Clinic B, etc.), there are cases where the spoken content used for explaining and guiding patients is slightly different. For example, fine-tuning is required for each facility, such as the hospital name, examination procedures, policy explanations, etc. Using the medical conversation support system of the present invention, explanatory texts and insertion texts specific to each facility are registered in the medical staff spoken data database 283A1 so that the system can replace and apply them without errors and can be easily customized for each medical institution. <STEP1: Registration of Templates for Each Medical Institution> Hospital Information Database 283D: In order for each medical institution to give unique explanations, a "hospital information database" that registers the hospital name ({hospitalName}), location, features, policies, etc. is prepared in advance, and they are expressed as insertion texts and registered. Example: Register spoken examples such as "At {hospitalName}, it is the policy to use sedatives during endoscopy." Register template sentences including the following "insertion texts" in the medical staff spoken data database. · "This is {{hospitalName}}. Today, it is an out-of-hours consultation." · "Please explain the endoscopy at {{hospitalName}}. {{Endoscopy Explanation}}" Manage these templates as "spoken examples for each facility" so that they can be dynamically replaced when creating explanatory texts for patients. <STEP2: Input, Search, and Verification of Individual Dialogue Data> · Designation of Facility Information When the medical institution (such as Hospital A) where the patient visits is specified, the system refers to the "facility information (hospital name, etc.)" and sets to use the custom utterance examples of the corresponding facility. ·Search·Verification For example, when the dialogue data with the patient is input and contains content such as "I want a detailed explanation about the examination", the system searches the medical staff utterance database. If a template of "examination explanation" (example: "At {{hospitalName}}, there is a diet restriction the day before the examination", etc.) is found and shows a high similarity, the utterance example is obtained. ·Insertion text with difficult collision Insertion text containing the hospital name, policy, etc. is separated and managed by assigning unique placeholders such as {{hospitalName}} so as not to collide with the text of other facilities. This can prevent the accidental mixing of "names of other facilities" and "different examination policies". <STEP3: Automatic replacement of insertion text and generation of utterance sentences> ·1. Replacement process for template sentences Placeholders such as "{{hospitalName}}" contained in the utterance sentence obtained as the search result are automatically replaced with the corresponding facility name or features. Example: "This is {{hospitalName}}." → "This is Takahashi Hospital." ·2. Additional registration for disease·examination explanation It is also possible to insert the policies and guidelines for each facility into the explanation of diseases and examinations. Register the phrase "In order to perform an endoscope with safety and security, if you undergo a gastroscope examination, please refrain from eating after 8 pm the previous night." in {{Gastroscope Explanation}}. Example: "When undergoing a gastroscope examination at {{hospitalName}}, {{Gastroscope Explanation}}" → "When undergoing a gastroscope examination at Suzuki Clinic, in order to perform an endoscope with safety and security, if you undergo a gastroscope examination, please refrain from eating after 8 pm the previous night." 3. Output The system presents the text after substitution to medical staff and patients in the form of screen display, voice output, etc., and realizes conversations and explanations optimized for each medical institution. <STEP4: Continuous customization through registration and update operations> ·1. Addition and modification by the facility staff The on-site staff of the medical institution (office staff, nurses, etc.) add newly emerged in-hospital rule changes and explanatory texts to the database. For example, changes such as "Since a new parking lot has been added in our hospital, you can come to the hospital by car from today." can be immediately reflected. ·2. Learning of selections by the system The system monitors which template texts are frequently used in which scenarios, etc., and preferentially proposes frequently occurring custom utterances. As a result, the facility-specific phrases are gradually optimized during actual operation, and the utilization rate increases. Industrial applicability and effectiveness in implementation: · Prevention of mixing of facility names, inspection policies, etc. Since each facility prepares its own explanatory template and registers it as non-conflicting insertion text, the risk of accidentally providing explanations of other hospitals can be reduced. · Reduction of the workload of medical staff By automatic substitution using templates, it is possible to provide a unified text surface without having to modify the description of the hospital name and inspection procedures one by one. · Ease of multi-facility deployment Even among large hospital groups and affiliated clinics, while sharing the basic template, only the words specific to each medical institution need to be overwritten. It is possible to flexibly respond even in scenarios across facilities such as telemedicine (remote medical care) and home nursing. · Improvement of patient satisfaction Since the explanations on the medical staff side are given in consistent words such as the hospital name, disease explanation, inspection content, etc., it is easy for patients to understand and convincing communication is realized.

Example 4

[0056] (4) Acquisition of Structured Data by Inserting Template (Claim 8) Summary: A medical staff registers inquiries made to a patient, confirmation of medical history, etc. by associating them with any one or more of electronic cal profile column information, electronic medical record template information or task information, or checklist information, or order information. Note that by associating with any one or more of electronic cal profile column information, electronic medical record template information or task information, or checklist information, or order information, it becomes possible to associate speech or a response to speech with electronic cal profile column information, electronic medical record template information or task information, or checklist information, and the subsequent confirmation work of medical staff is made more efficient. Example: · For example, in the case of electronic cal profile column information or electronic medical record template information, the burden of data transfer to the subsequent profile column is alleviated. Or in the case of task information or checklist information, since a conversation is properly logged, the confirmation work for forgetting to talk or forgetting to confirm is reduced, and at the same time, when something has been forgotten to be said, medical staff can make additional confirmations. · In one example, category information, etc. can be put in the following parentheses and then the speech sentence can be associated. · “(Past History) Have you had any major illnesses in the past?” · “(Confirmation of Visiting Another Hospital) Have you visited a hospital other than {Hospital name} since your last visit?” Note that in a medical conversation, since the past history is generally described below the normal medical record, it is also possible to assign an order such as “(Order 1) What is the reason for this visit?” ··· “(Order 10) Have you had any major illnesses in the past?” in the conversation and display the responses sorted in that order. <STEP1: Register Template Sentence and Category Information> · 1. Creation of Template Sentence Prepare a question text "Have you ever had a major illness in the past?" as a template, and set the category as "Past medical history" and the order (or priority). Prepare a question text "Have you visited a hospital other than ABC Hospital since your last visit?" as a template, and set the category as "Confirmation of visits to other hospitals" and the order. 2. Linkage with the database of healthcare provider speech data Register the above template text in the database, and manage it by attaching meta-information such as "Category: Past medical history", "Category: Confirmation of visits to other hospitals", "Order: Priority (1, 2,...)". Based on this information, the system can sort and present the information in an order that is easy for healthcare providers to understand. In this case, this control is performed through search, but control is also performed by training the program to output "category information". <STEP2: Input of dialogue data with the patient and presentation of template candidates> ·1. Input of dialogue data The patient inputs information related to the medical institution visit history and past medical history, such as "I have recently visited another hospital" or "I had surgery 5 years ago", into the system (or voice input → text conversion). ·2. Template search and order management The system analyzes the input dialogue content, searches for the template text of the corresponding category (past medical history, visits to other hospitals, etc.) in descending order of priority, and presents it. <STEP3: Register the obtained answer as structured data> ·1. Mapping of patient answers When the patient answers the question "(Past medical history) Have you ever had a major illness in the past?" with "I had surgery for gastric cancer 10 years ago", the system can save it in the database as an answer in the (past medical history) category. When the patient answers correctly according to the intention of the questioner, the data is linked to the appropriate item. If the past medical history is heard from the family and the patient has not been informed yet, it should be recorded as (past medical history) gastric cancer (not informed). If the answer is "I visited DEF Clinic yesterday", it is registered as an answer in the (Other Hospital Visit Confirmation) category. Also, when answering yes or no, operations such as checking the checkbox can be performed. ·2. Advantages of Structured Data Conversion By organizing the conversation content, which was previously often saved as free text, into categories (Medical History / Other Hospital Visits) + answer content, it becomes easier to utilize in electronic medical records and data analysis. For example, it becomes easier to narrow down and search for "the number of patients with a medical history of 'gastric cancer'". <STEP4: Expansion to Subsequent Processing> · Automatic Template Generation and Verification Using structured data, it is possible to automatically generate and expand "questionnaire templates for patients with a medical history and visits to other hospitals". Even templates created independently by the facility side can be presented by the system at an appropriate timing if relevant categories are specified. · Document Creation and Report Output Summarize the above answer data and reflect it in documents such as hospitalization consent forms and referral letters. Using insertion elements such as {Hospital name}, it becomes possible to generate patient-oriented documents such as "This will be an examination at our hospital (ABC Hospital). Although endoscopes are very rare, there is a risk of perforating the digestive tract." · Customization and Continuous Update Since categories and order (priority) can be changed and added later, it can be flexibly operated to reflect the voices of medical staff on-site. Even if new question items (e.g., allergies, family history) are added, they can be incorporated into the existing flow. <Industrial Applicability and Effectiveness in Implementation> · Structuring of Conversations By registering templates + categories + order candidates, miscellaneous free-text conversations are organized as questions and answers for each category. Not only does the information retrieval performance improve when conducting searches and analyses later, but the visibility also improves. · Reduction of the Burden on Medical Staff By inserting templates, necessary responses can be reliably obtained while preventing omission and duplication of questions. Since the data is structured, subsequent processes such as report creation can be smoothly carried out. · Easy-to-understand explanation to patients Even for the same content, the quality and consistency of the explanation to patients are ensured because the hospital name, examination method, etc. are accurately displayed in an inserted format.

Example 5

[0057] (5) Barber (hair salon) reservation system Overview: A conversation support system for making reservations and checking menus in barbershops (hairdressing and beauty salons), beauty parlors, etc. When customers check and request dates and times, desired styles, and optional services (shampoo, head spa, etc.), the system presents optimal additional questions and proposed sentences to smoothly complete the reservation. <STEP1: Registration of salon conversation data in the database> Menus and services specific to each store: In addition to basic menus such as cuts, colors, and perms, additional services unique to the store such as shampoo, treatment, and head spa are registered in the database. Template phrases: Examples: Register fixed customer service phrases such as "What date and time would be convenient for your reservation?" and "How about the set menu of cut and perm?" <STEP2: Input of conversation data with customers> Incorporation of customer requests: Input the customer's wishes (such as "I would like a cut and color on the weekend" and "Can I make a reservation for the evening time slot?") from the customer (chat, telephone voice recognition, etc.). Menu proposal (example of additional conversation): The system analyzes the customer's requested content (date and time, treatment content) and automatically generates or searches the database for candidate phrases (such as "There are several types of colors. Do you have any preferences?") and proposes them. <STEP3: Service proposal through search and verification> Search in the salon DB (substitute for hospital information DB): Compare with the reservation frame information on the store side such as "Surgery can be performed for one hour starting at 17:00 in the evening", and extract the optimal additional conversation examples according to the availability and the required time of the corresponding menu. Database response: Example: The system presents examples such as "There is an available slot from 17:00 to 18:00. It will take about 90 minutes for the combination of color and cut. Is that okay?" <STEP4: Output and reservation confirmation> · Confirmation with the customer Generate a text with the store name and date inserted, and finally present it to the customer. "Is it okay for a color + cut from 17:00 on [Month] [Day]?" · Reservation confirmation Confirm the reservation when the user's consent is obtained, and the system updates the store's reservation calendar. If necessary, issue a reminder message. <Industrial applicability and effectiveness in implementation> · Business efficiency improvement: Since the communication by phone or chat is standardized, the workload of the staff is reduced. · Customer satisfaction improvement: The desired time and options can be smoothly guided and proposed, and the customer service quality is improved.

Example 6

[0058] (6) Explanation of the usage method of mobile phones Overview: Provide conversation support for operation explanations and fee plan explanations in mobile phone sales stores and call centers. According to the content of the user's inquiry (such as fees, functions, setting methods, etc.), the system presents appropriate additional questions and explanatory texts to handle the inquiry with less confusion. <STEP1: Registration of mobile phone support conversation database (replacement of call DB283A1)> · FAQ and manual information Register support information as templates, from basic operations of the phone (such as setting ringtones, Wi-Fi connection, etc.) to option explanations of fee plans (such as family discounts, additional data traffic, etc.). · Categories of inquiries Define categories (such as operating system, fee plan, failure / repair, initial settings, etc.), and register standard explanatory texts and additional questions in the database by associating them with each category. <STEP2: Interactive Data Input (Customer Inquiry)> · The customer makes an inquiry Example: Obtain requests such as "I feel that the monthly fee is high and I want to review it" and "I don't know how to set the incoming call ringtone" via chat or phone. · Content analysis by the system Use natural language processing to identify keywords such as "fee review" and "setting method", and search for relevant category texts and additional questions. <STEP3: Database Matching and Proposal> · Obtain template phrases For example, present additional questions to be confirmed, such as "May I ask for the details of the fee plan?" and "Do you know the name of the plan you are currently using?" · Progressive hearing Receive the user's answer, the system identifies the fee plan, and automatically generates or searches for supplementary questions such as "Is the family discount or student discount applicable?" and guides the user. <STEP4: Output of Guided Content> · Automatic creation of explanatory texts Automatically replace the template summarizing the advantages and disadvantages of plan changes according to the user's usage situation, and present it in an easy-to-understand manner. · Display of operation procedures Extract the incoming call ringtone setting procedure from the procedure manual template, and display specific steps such as "On the [Settings] screen → [Sound & Vibration] → Tap [Incoming Call Ringtone]" on the screen or explain them orally. <Industrial Applicability and Effectiveness in Implementation> Efficiency improvement in call center response: Staff can surely handle standard FAQs and additional questions, and the answers are unified. User-friendly: Since the system proposes supplementary questions, it can provide a certain level of response without depending on the operator's proficiency.

Example 7

[0059] (7) Examples of inspection explanation content Summary: Doctors and nurses explain blood tests and imaging tests (such as CT, MRI, and endoscopy) to patients. When explaining the procedures, purposes, risks, and precautions of the tests to ensure that patients understand the test content correctly, appropriate explanation phrases are retrieved from the medical staff speech database and provided comprehensively. <STEP1:: Registration of inspection explanation template (Claim 8)> · Standard explanation text by inspection type (Inspection 283B1 data) Referring to Inspection 283B1, for each inspection such as blood test, X-ray, CT / MRI, and endoscopy, a set of assumed Q&A and explanatory texts are registered in the medical staff speech database in advance. Example: 「(CT inspection explanation) In a CT inspection, cross-sectional images of the body are taken using X-rays. Please inform the staff if you have claustrophobia or may be pregnant.」 「(Endoscopy inspection explanation) The dietary restriction before an endoscopy is until ○ hours before.」 etc. · Linking of risks and precautions For example, items explaining the allergic risk of contrast agents and precautions regarding anesthesia during endoscopy inspection are added to the template of Inspection 283B1. Categories such as 「Preparations before inspection」, 「Procedure on the inspection day」, and 「Precautions after inspection」 are set up so that explanations can be given in sequence. <STEP2: Input and verification of dialogue data with patients> Individual dialogue data 283C: If the patient asks questions such as 「Is the inspection painful?」 or 「How long does it take?」, the content is input into the system as individual dialogue data. Speech generation program or verification (Claim 8): The system searches the database for templates and FAQs corresponding to the inspection, and collates and extracts the optimal additional speech examples (such as 「Pain when using anesthesia」, 「The inspection time is approximately ○ minutes」, etc.). Explanatory speech: When the patient has not mentioned contrast agent allergy in the individual conversation data, the system proposes an additional question such as "This is an examination using a contrast agent. Have you ever had an allergic reaction?" <STEP3: Confirmation> Physician / Nurse Confirmation: The physician or nurse checks the example utterances extracted and presented by the system on the screen. After making fine adjustments as necessary, they verbally explain them to the patient. Presentation to the Patient: The examination procedures and precautions are presented in an easy-to-understand order, and there may be a function to print them as materials. Insertion items such as "{{hospital Name}} requires you to fill out the following documents before the examination" are set in the template, enabling it to adapt to the operations of each hospital. <Industrial Applicability and Effectiveness in Implementation> Since the precautions and prior preparations for each examination are guided without omission, the risk that the patient may not be able to undergo the examination due to insufficient understanding or may encounter troubles is reduced. The quality of explanations is standardized, improving medical safety and patient satisfaction.

Example 8

[0060] (8) Conversation collection considering the patient's disease notification status Overview: In the case of a serious illness including cancer notification, the situation varies depending on the wishes of the patient and family, such as "desiring notification" or "not desiring notification". The system grasps the patient's notification status (policy information) and, as necessary, changes the explanation content or avoids specific information to support medical staff to communicate safely and appropriately. In addition to cancer notification, it is possible to link "situations where conversation should not be carried out (hereinafter referred to as taboo situations)" to the conversation examples and save them in Notification 283B2. For example, for a patient who is a Jehovah's Witness, proceeding with a blood transfusion, etc., situations that are usually referred to as "taboo situations" are held in the database, and it has a function to ensure the appropriateness and safety of the output of conversation examples. In this case, this control is performed through search, but there are also examples where control is performed by training the program to output "situations where conversation should not be carried out". Example: Save the utterance example “(Blood transfusion explanation and consent confirmation) In a major surgery like this, blood transfusion may be necessary during the operation. Please sign the blood transfusion consent form.” linked to the taboo state (a person known to be a Jehovah's Witness). Save alternative conversation examples linked to this situation. For example, the alternative example in this case is “(Autologous blood collection and blood transfusion explanation and consent confirmation) In a major surgery like this, blood transfusion may be necessary during the operation. Although it takes time to prepare, there is a method called autologous blood transfusion, and it is possible to save your own blood and transfuse it back. Please sign the autologous blood transfusion consent form.” (Note that this state also cannot be spoken during an emergency surgery. This conversation during an emergency surgery becomes a taboo state.) <STEP1: Management of notification status> Status information: Register status information such as “Do not notify of cancer,” “Notify only a part,” and “Actively notify” in the database (Notification 283B2). This is reflected as a policy setting in the electronic medical record and this system based on the patient's intention survey and discussion with the family. Utterance examples by notification pattern: Prepare a conversation collection (conversation collection) according to each status in the medical staff utterance database. Example: Non-notification policy: “Text example to avoid excessive elaboration regarding physical condition and test results” Partial notification: “Balanced type that hides the disease name but explains the necessity of treatment” Full notification: “Explicitly explain the disease name, treatment content, prognosis, etc.” <STEP2: Input of conversation data and control of utterance candidates> Obtaining dialogue data: Input the content of the interview with the patient and family. If there is a history that the patient has said “I don't want to know the disease name in detail,” the system recognizes the non-notification status. Utterance generation based on status: When the medical staff explains the condition, phrases that do not directly mention the disease name are candidates for the “non-notification policy.” Conversely, in the “full notification” status, a text including detailed treatment content and future prospects is automatically proposed. <STEP 3: Contradiction Check and Presentation (Claim 6)> Contradiction detection: The system recognizes the "non-disclosure status" and makes it impossible to output a search for conversations for cancer disclosure. Or, an alert is issued to reconfirm the use of the utterance examples. This suppresses the risk of making incorrect disclosures against the patient's intention. <STEP 4: Status Update and Sharing> Response to situation changes: If the patient later wishes to be informed, updating the status from "partial disclosure" to "full disclosure" will cause the system to recommend presenting detailed information in subsequent explanations. In-hospital collaboration: By sharing this status (status change) on the electronic medical record, it becomes visible to nurses and doctors in other departments that "the patient himself / herself is not fully informed of the disease name", etc., and a consistent response is made. <Industrial Applicability and Effectiveness in Implementation> It becomes possible to provide information in line with the intentions of the patient and family, and prevent careless remarks and unnecessary confusion among medical staff. Through status management, it is possible to systematically control so that the description of the medical condition does not deviate from the patient's intention.

Example 9

[0061] (9) Examination Explanation Scenario: Signing the Consent Form, Confirming Side Effects and Risks <STEP 1: Registration of Conversation Templates According to the Examination Consent Procedure> Templates regarding the consent form: · "If you understand the content of this examination, please sign this consent form." · "If you have any questions, please feel free to ask." These phrases are registered in the database under the category of "examination consent". Side effect and risk explanation templates: · "(Allergy confirmation) There is a possibility of using a contrast agent. Have you ever had an allergy to a contrast agent in the past?" · This is a test with rare risks such as bleeding and infection, but we will take sufficient precautions in advance. Please note that depending on individual constitutions and pre-existing conditions, side effects as described below may occur. Allergy reaction... These are summarized as "Side Effect and Risk Explanation", and additional items for each case (such as during pregnancy and pre-existing conditions) are also added to the database. Individual Risk Confirmation: "Based on the past history of ●● (patient name), the points that require attention in this test are ××. Are you okay?" For example, if there are conditions such as hypertension or heart disease, precautions during the test are prepared as a template and can be output by combining with individual dialogue data. <STEP2: Input of Individual Dialogue Data (Interaction with the Patient)> Import of Patient Profile and Past History: The patient's medical condition and allergy history registered in the electronic medical record, etc., are obtained as individual dialogue data. If the patient reports "I have had a rash due to medicine before", the system automatically determines that "there is an allergy risk". Scene of Test Application: The patient hopes and agrees to undergo the test, but has not yet received a detailed explanation. Doctors or nurses operate the system to obtain an example of speech by combining the "Test Explanation" template and the patient's individual data. <STEP3: Generation and Search of Additional Speech> Dialogue Proposal According to the Presence or Absence of Side Effects: The system checks the "Side Effect and Risk Explanation" template and displays a list of items that need to be further explained based on the patient's past history and reported content. Example: In the case of a past allergy → "Caution is required when using the contrast agent. If symptoms occur, please inform the staff immediately", etc. Invocation of the Consent Form Signing Process: Since the "Test Consent" template is also stored in the database, the following example of speech is presented at the timing of entering the test consent process. Example: "Since I have already explained the outline of the examination, if you seem to have understood, please sign this consent form." Emphasis on individual risk items: If the patient has hypertension, slightly emphasize the bleeding risk during the endoscopic examination and propose follow-up questions such as "What is your usual blood pressure?" <STEP4: Final output and consent confirmation> Presentation of explanations and precautions: The system combines templates and presents the combined explanations to medical staff as on-screen displays or in written form, enabling explanations in an easy-to-understand manner for patients. For consent forms that require signatures, confirm them on the terminal or in paper form so that patients and their families can read them. Signature or electronic signature: If the patient is satisfied, the patient makes a physical signature or electronic signature on the consent form, and the system records the date and time of consent acquisition and the signature in the electronic medical record or database. At the same time, also leave logs such as "verbally explained" and "no particular objections from the patient" regarding side effects and risks. Follow up on individual risks without omission: If the patient states that "I haven't heard about such side effects", the system presents "Side effect FAQs" and "Emergency contact information" etc. again from the template and can provide additional explanations. <Industrial applicability and effectiveness in implementation> Prevention of information leakage and insufficient risk explanation: Since the system templates the examination explanation process and automatically prompts individual risk explanations according to past medical history and allergy information etc., it is possible to avoid oversights and insufficient explanations as much as possible. Improvement of medical safety: The risk of performing an examination without grasping the possibility of side effects and special risks is reduced, and the safety level for patients is increased. Promotion of patient understanding: When signing the consent form, it becomes possible to create a process where the patient gives consent after fully understanding their condition and the necessity and risks of the examination, improving the quality of informed consent. Reducing the burden on healthcare workers: The templates and matching results make it easier for healthcare professionals to comprehensively explain complex risks. This reduces the burden of repetitive explanations and improves work efficiency. [Example 10]

[0062] (10) Examples of building a speech example database <1. Data Source Collection and Input> Case report: We collect case reports shared at academic conferences and in-hospital conferences, and transcribe the treatment process and key points for patient management into text. Example: "Case report of a case in which the following test was performed when the following symptom appeared." Medical record information: From past medical records, we extracted interview phrases and explanatory statements that doctors and nurses actually used in their interactions with patients. Examples: "If you have had a fever for more than two days, you will need to undergo a certain test," or "Are you receiving nutritional guidance?" DWH Information: Standard questions and explanations corresponding to frequently occurring symptoms and treatments are extracted from a large-scale medical database (DWH). Example: Extract examples of "dietary guidance for diabetic patients" that were frequently used in medical interviews over the past few years. <2. Summary of Program-Generated Speech Examples> Automatic generation and editing of example speech: The above sources (case reports, medical record information, DWH information) are input into a natural language processing program to extract and reorganize the core elements of the conversation (question patterns, explanation patterns, attention-grabbing phrases, etc.). Example: - "If you have a fever, check the degree and duration of the fever" → "What is your temperature? How long has it been going on?" Here, tags are added indicating the situation (first visit, follow-up visit, emergency, etc.), contraindications (allergies, refusal to disclose cancer diagnosis, etc.), category (medical history, explanation of tests, lifestyle guidance, etc.), and context (open-ended questions, closed-ended questions, etc.). Merge fields such as {{Hospital Name}} or {{...}}: Placeholders such as {{hospitalName}} and {{test name}} are set up in the spoken text to allow embedding facility-specific information such as the hospital name, test description, and precautions. Example: "At {{hospitalName}}, fasting is required before this test, so please refrain from eating for {{fasting time}} hours beforehand." <2. Integration with checklist information> Manage closed questions (yes / no questions or questions with predetermined options) as a checklist and define when (at what timing) to ask them, along with a timeline. Examples: "Do you have any allergies? (Yes / No)" "When was your last examination? (Enter date)" etc. <3. Adding contextual information to each utterance example> Open Question: A question format that allows patients to freely describe their symptoms and situation. Examples: "What symptoms are you experiencing today?" "Is there anything that's bothering you?" Closed Question: Questions with limited options, such as "Yes / No" or "A / B / C". By creating a checklist and presenting it **chronologically (e.g., at the initial consultation, immediately before the examination)**, we can prevent oversights and facilitate data collection. Background question: Questions designed to gather background information about the patient (medical history, lifestyle, etc.). Examples: "How often do you exercise?" "Do you have a history of diabetes?" <4. Customization and editing> Responding to contraindications: If the database contains information such as "has allergies" or "cannot use certain medications," the program automatically adjusts by hiding the corresponding speech examples or adding a "warning." Modifications for presentation to another patient: The program controls the patient's dialogue history and context to rephrase things in a natural way and ask questions at the appropriate time. Example: Avoid asking a patient who has already said they have a fever, "Do you have a fever?" Facility-specific wording and insert text: The system automatically replaces phrases like "The reception hours for {{hospital Name}} are until 5 PM on weekdays" and "{{hospital notices}}" to accurately explain hospital rules and facility information. <5. Examples of prompts actually used> "Based on the following case report and medical record information, generate five open-ended questions for first-time patients. However, the output must include notes regarding {{hospital name}}." The text is extracted and generated using "first-time patient," "open-ended question," and "{{hospital name}}" as keys. "Based on the data extracted from the data warehouse, create a checklist-type closed question, assign categories (medical history / lifestyle, etc.), set a priority for each, and output the results." The program provides answers by listing the items that need to be confirmed with the patient and indicating the order in which to ask them (priority). <6. Summary> The effectiveness of the speech example database: Insights obtained from various sources such as case reports, medical records, and data warehouse analysis can be aggregated and centrally managed as standardized interview and explanation phrases. By dynamically inserting hospital names and notes using the {{}} format, it's easy to accommodate customization for each facility. Benefits of adding context: Speech examples are classified into categories such as "open questions," "closed questions," and "background questions," and the contraindicated situations and timing are controlled as needed. This will reduce the burden of conversation for healthcare professionals, enabling smoother information gathering and explanation. Practical applications: By leveraging its customization features, it can be used in a wide range of medical settings, such as by replacing hospital names or test names to expand to multiple facilities, or by simplifying electronic medical record entry using closed-list questions. [Example 11]

[0063] <Example prompt> Based on the following information from the "Consultation Report Content," "Conversation Scene," and "Summary of Reason for Consultation," and in accordance with the conversation summary creation policy, output the conversation following the "Previous Conversation" in the "{data1} Conversation Scene at the Time of Consultation" within code blocks. The conversation should be designed to maximize the evaluation axis of the conversation below. Output the conversation, the evaluation of the conversation, and the summary of the reason for consultation according to the format. If there are situations where conversation should not take place, indicate the contraindication condition (e.g., Contraindication condition: When the patient is awake). <Prompt for creating a medical interview conversation> [the purpose] This prompt aims to systematize an effective conversation flow in medical interviews and appropriately elicit information useful for diagnosis and treatment. [Rules for creating dialogue scene names] Using conversational terminology as a reference, create a name that is as concise as possible, consisting of 2 to 6 characters.

[0064] [Conversation Creation Rules] [Table 1]

[0065] [Structure of a conversation scene] <1. Open-ended questions (to elicit the chief complaint)> -3-4. Clarify the patient's chief complaint through question-and-answer sessions and guide them towards closed questions. Once the timeline and reason for the visit (mainly the chief complaint) are determined, the session can be concluded (e.g., acute gastritis, traffic accident injury). Do not delve into closed questions or background questions. [Table 2]

[0066] <2. Organizing the reason for seeking medical attention (organizing the main points for diagnosis)> (Symptoms, Findings, and Diseases): Describe the patient's specific symptoms, findings observed by the doctor, and suspected diseases. (Social): Describe the social background that may influence the patient's symptoms, such as lifestyle, work environment, and stressors. (Psychology): Analyze the psychological aspects, such as the patient's anxiety, psychological burden, and hesitation about whether or not to seek treatment. <3. Closed Questions (Gathering Detailed Information)> Based on the reason for the visit, extract examples of additional questions for appropriate question-and-answer sessions, such as 20 questions and answers, using the framework described below. Do not delve into background questions. [Table 3] [Table 4]

[0067] <4. Background Questions (Patient Background Information)> The patient's medical history and lifestyle are confirmed through approximately 20 questions and answers. The dosage is adjusted according to the severity of the patient's symptoms. [Table 5]

[0068] <**Example conversation output**> The conversation scene is described in plain text, followed by the conversation itself. For open-ended questions, if there is a hypothetical chief complaint, write "Hypothetical Chief Complaint," and for closed-ended questions, write the chief complaint as is. (Conversation scene: Open-ended questions assuming stomach pain) Doctor: (Confirming the chief complaint) Good morning. What can I help you with? Patient: Good morning, doctor. For the past three days or so, I've had a stomach ache and haven't been able to eat much. Doctor: (Empathy) That must be tough. Doctor: (Location of symptoms) Where exactly is the stomach pain located? Patient: It's around my stomach. Doctor: (Symptoms) What kind of pain is it? Patient: I have a constant dull ache. Sometimes it becomes a sharp, stabbing pain. <**Example of a summary of the reason for seeking medical attention**> ```plaintext (Symptoms / Findings / Disease) Stomach pain, nausea, loss of appetite, suspected stomach ulcer × 3 days (Society) Increased stress due to busy work schedules, irregular eating habits, and increased frequency of alcohol consumption. (Psychology) I'm anxious because I don't know the cause of my stomach pain, I tend to feel down due to work stress, and I'm unsure whether I should go to the hospital. [Table 6]

[0069] <**Evaluation Format:**> plaintext Fluency: X, Comprehensiveness: X, Empathy: X, Depth: X, Conversation Density: X, Lack of Harmfulness: X <**Conversational Terminology Classification**> • Symptom name (e.g., acute fever, chronic headache, acute stomach ache, acute dizziness, hypothetical acute stomach ache, etc. Combine the timeline and symptoms. Include hypothetical statements when it's an open question.) • Abnormal test result (e.g., abnormal health checkup result) • Trauma (traffic accidents, falls, slips, etc.) • Disease name (hypertension, dyslipidemia, diabetes, etc.) <**Category Glossary**> - Empathy - Acknowledgment - Chief complaint - Characteristics - Time course - Precipitating factors - Relief factors - Presence or absence of complications - Related symptoms - Location - Radiation - Severity - Impact on ADL - Habitual changes - Past medical history - Family history - Medication history - Surgical history - Developmental history - Pregnancy history - Accident details - Complications <**Information for creating conversation**> <**Medical consultation article content**> ·{{text1}} ·**Previous conversation** ·{{text2}}{{text3}}{{text4}} [Industrial applicability]

[0070] The medical conversation support system according to the present invention can efficiently and effectively support communication between medical professionals and patients. In particular, the function of generating additional speech examples based on a speech example database and searching for them based on match or similarity promotes appropriate communication in medical settings and contributes to reducing patient anxiety and improving treatment effectiveness. For this reason, the present invention is useful as a tool that can be widely adopted in medical institutions such as hospitals and clinics.

[0071] Furthermore, the medical conversation support system according to the present invention is a technology for supporting dialogue between medical professionals and patients, and has potential applications in a wide range of industrial fields, as described below. <General Medical Institutions> By introducing this system to hospitals, clinics, and other settings where patient interaction is a daily occurrence, it can streamline the process of taking medical interviews and providing explanations, thereby reducing the workload of doctors, nurses, and other medical staff. In particular, by integrating with medical interview support systems and electronic medical record systems, it contributes to improving the efficiency of medical operations and enhancing the quality of medical care. <Elderly Care and Home Healthcare Field> It can also be used in elderly care facilities and home healthcare settings to support communication with patients and users, and to collect and explain necessary information without fail. For visiting nurses and home healthcare providers, the ability to accumulate and reuse conversation data is a major advantage. <Medical Call Center / Consultation Service> This system can be used in medical consultation hotlines and call centers to support callers by instantly searching and matching anticipated questions and example answers to provide appropriate advice. <Expansion into fields other than medicine> The present invention has the technical features of accumulating dialogue data and performing automatic generation programs and similarity searches, making it applicable to industries other than the medical field, such as handling product inquiries and general customer support.

[0072] <1. Improved efficiency of communication and enhanced medical safety> According to the present invention, additional speech sentences are automatically generated based on speech examples and individual dialogue data registered in the medical professional speech database, thus eliminating the need for medical professionals to devise interview questions and explanations from scratch. On the other hand, searching or matching against a database enables safe communication that does not deviate from standard guidelines or past conversation examples, thereby reducing the risk of medical errors and insufficient explanation. <2. Continuous improvement through learning and updating> As new dialogue data is accumulated from daily interactions in medical settings, speech generation programs and databases are updated to produce more accurate additional speech examples. This allows the system to learn the expertise of healthcare professionals over time, continuously improving its conversational support capabilities. <3. Flexible response to multiple scenarios> This invention allows for the registration of speech examples corresponding to various medical scenarios, such as initial consultations, follow-up consultations, and support for hospitalization and discharge, enabling flexible application to diverse situations. By utilizing features such as template linking (Claim 8) and electronic medical record category linking (Claim 9), further optimization tailored to practical needs becomes possible. <4. Compatibility with other systems> By integrating with existing electronic medical record systems, workflow systems, and data warehouses (DWH), it leads to centralized management of medical information and increased efficiency in the entire clinical process. Furthermore, documents and templates presented to patients can be automatically generated using a program that embeds variables for output, contributing to paperless operations and a reduction in administrative work. From these points, the medical conversation support system of the present invention can effectively address a variety of issues related to communication between healthcare professionals and patients, and can be a useful technical means not only in medical settings but also in a wide range of dialogue fields.

[0073] <5. Suppression of Hallucination> Conventional generative AI-based conversational systems (such as large-scale language models) carry a high risk of introducing non-existent or inaccurate information (so-called "hallucination"). In this invention, by combining the step of comparing additional utterance examples with a medical professional utterance database (such as searching or similarity determination), it is possible to ensure that the automatically generated utterances match the correct context and facts. This significantly reduces the risk of inaccurate explanations or information about non-existent treatments being presented, as the content generated by the AI ​​is backed by reference information. <6. Introduction and flexibility of custom utterances> This invention makes it possible to register and update custom speech examples in a medical professional speech database, tailored to the specific circumstances of each medical institution and workplace. For example, by reflecting each facility's treatment policies, precautions, examination procedures, and unique wording used in patient explanations, we can achieve both general applicability and facility-specific responses. These custom utterance examples are incorporated into the generation of additional utterance examples by the utterance creation program and into database matching, which offers a significant practical advantage: they can be updated immediately in response to changes in the medical field. <7. Improving medical safety and accountability> To prevent inaccurate AI output and omissions in explanations, a double-check system involving human review and database verification is crucial. This invention combines these two approaches to enable healthcare professionals to safely and smoothly interact with patients. Template explanations using custom utterance examples reduce individual variability among healthcare professionals and enable standardized dialogue that reassures patients. <8. Accuracy improvement through continuous refinement> By learning from registered custom utterances and actual dialogue history, and continuously correcting and supplementing the database as new errors or omissions are discovered, the risk of misgeneration, including hallucination, will be further reduced in the long term, and the system's accuracy will improve.

[0074] [Note B1] A medical conversation support program based on medical conversation data for operating a computer equipped with a processor and memory, The memory holds a medical professional speech database in which examples of utterances for medical professionals to perform medical conversations are registered. The processor, in the program, We have implemented a speech generation function that takes patient-healthcare worker conversation data as input and outputs additional speech sentences. By realizing the aforementioned speech generation function, Taking individual patient-healthcare worker conversation data, which is a conversation between an individual patient and a healthcare worker, as input, it outputs additional individual utterance examples. Using the aforementioned additional individual utterance examples as input, the utterance examples are searched or matched based on the matching or similarity of the medical professional utterance database to obtain search results or matching results. A medical conversation support program characterized by providing a function to output one or more of the following to the individual patient: the individual utterance example, the search result, or the matching result. [Note B2] A medical conversation support program based on medical conversation data for operating a computer equipped with a processor and memory, The memory maintains a healthcare worker speech database in which additional utterances by healthcare workers are registered in relation to patient-healthcare worker conversation data. The processor, in the program, The individual patient-healthcare worker conversation data, which is a conversation between an individual patient and a healthcare worker, is used as input to search or match the healthcare worker utterance database and obtain associated additional utterance examples. A medical conversation support program characterized by outputting the aforementioned additional speech examples to the individual patient. [Note B3] A medical conversation support program based on medical conversation data for operating a computer equipped with a processor and memory, The memory includes a medical professional speech database in which examples of utterances for medical professionals to perform medical conversations are registered, It maintains an individual key points database that holds individual key points data extracted from individual patient-healthcare worker conversation data, which is a conversation between an individual patient and a healthcare worker, The processor, in the program, The system implements a speech creation function that takes the aforementioned individual key point data and one or more of the following as input: individual dialogue data, electronic medical record data, data on a data warehouse, or referral letter data, and outputs additional individual speech examples. By realizing the aforementioned speech generation function, In addition to implementing a function to output the aforementioned additional individual utterance examples, Using the aforementioned additional individual utterance examples as input, the utterance examples are searched or matched based on the matching or similarity of the medical professional utterance database to obtain search results or matching results. A medical conversation support program characterized by providing a function to output one or more of the individual utterance examples, the search results, or the matching results to the individual patient. [Note B4] A medical conversation support program characterized by associating the aforementioned speech examples described in any of Appendix B1 to Appendix B3 with one or more of the following situations: initial consultation, follow-up consultation, admission, during hospitalization, or admission / discharge support center, and inputting the aforementioned situation when performing the search or matching. [Note B5] A medical conversation support program described in any of Appendix B1 to Appendix B3, The aforementioned utterance examples are linked to contradictory situations, The memory holds a conversation inconsistency evaluation program that has the function of evaluating inconsistencies in conversations, The additional utterance examples and one or more of the individual key point data or individual dialogue data are input into the conversation contradiction evaluation program. A medical conversation support program characterized by its ability to detect or evaluate contradictions between two parties. [Note B6] A medical conversation support program described in any of Appendix B1 to Appendix B3, The speech examples registered in the medical professional speech database are linked to a flag indicating whether or not they contain information about cancer diagnosis. The aforementioned processor, A medical conversation support program characterized by implementing a function that, when it is detected that policy information indicating not disclosing cancer is stored in one or more of the individual key point data or the individual dialogue data, controls the search for medical conversations so that conversations containing information about disclosing cancer are not searched. [Note B7] A medical conversation support program described in any of Appendix B1 to Appendix B3, The aforementioned medical professional speech database contains the content of the examination explanations registered in the database. The processor controls the inclusion of the pre-consent examination explanation content in the medical conversation. A medical conversation support program characterized by its ability to record as a log that the utterance was completed in the entire conversation. [Note B8] A medical conversation support program described in any of Appendix B1 to Appendix B3, The system associates and stores speech examples registered in the aforementioned medical professional speech database with one or more of the following: electronic medical profile information, electronic medical record template information or task information, checklist information, or sorting information. When the processor searches or matches the medical professional speech database based on the individual dialogue data, The method for displaying conversation items to be presented according to one or more of the following: items in the associated electronic medical profile field, items in the electronic medical record template, task information, checklist information, or sorting information, is managed. Furthermore, the medical conversation support program is characterized by its ability to improve the visibility of which conversation items have been completed by linking and saving the patient's responses to the conversation items. [Note B9] A medical conversation support program described in any of Appendix B1 to Appendix B3, The aforementioned medical professional speech database is registered with one or more associated functions: either the function to insert long explanatory text, or the function to insert documents to replace hospital names, department names, etc. The medical conversation support program is characterized in that the processor implements a function to customize and control the inclusion of medical conversation information by replacing it with medical institution-specific information using the function of performing the mail merge. [Note B10] A medical conversation support program described in any of Appendix B1 to Appendix B3, A medical conversation support program characterized by its ability to create and register speech examples using one or more of either electronic medical record information or case reports.

[0075] [Note C1] A medical conversation support method based on medical conversation data for operating a computer equipped with a processor and memory, The memory holds a medical professional speech database in which examples of utterances for medical professionals to perform medical conversations are registered. The processor, in the program, The system takes patient-healthcare worker conversation data as input and executes a speech generation function that outputs additional speech sentences. By realizing the aforementioned speech generation function, Taking individual patient-healthcare worker conversation data, which is a conversation between an individual patient and a healthcare worker, as input, it outputs additional individual utterance examples. Using the aforementioned additional individual utterance examples as input, the utterance examples are searched or matched based on the matching or similarity of the medical professional utterance database to obtain search results or matching results. A medical conversation support method characterized by performing a function to output one or more of the individual utterance examples, the search results, or the matching results to the individual patient. [Note C2] A medical conversation support method based on medical conversation data for operating a computer equipped with a processor and memory, The memory maintains a healthcare worker speech database in which additional utterances by healthcare workers are registered in relation to patient-healthcare worker conversation data. The processor, in the program, The individual patient-healthcare worker conversation data, which is a conversation between an individual patient and a healthcare worker, is used as input to search or match the healthcare worker utterance database and obtain associated additional utterance examples. A medical conversation support method characterized by outputting the aforementioned additional speech examples to the individual patient. [Note C3] A medical conversation support method based on medical conversation data for operating a computer equipped with a processor and memory, The memory includes a medical professional speech database in which examples of utterances for medical professionals to perform medical conversations are registered, It maintains an individual key points database that holds individual key points data extracted from individual patient-healthcare worker conversation data, which is a conversation between an individual patient and a healthcare worker, The processor, in the program, The speech creation function is executed by inputting the aforementioned individual key point data and one or more of the aforementioned individual dialogue data, electronic medical record data, data on the data warehouse, or referral letter data, and outputting additional individual utterance examples. By executing the aforementioned speech creation function, In addition to implementing a function to output the aforementioned additional individual utterance examples, Using the aforementioned additional individual utterance examples as input, the utterance examples are searched or matched based on the matching or similarity of the medical professional utterance database to obtain search results or matching results. A medical conversation support method characterized by performing a function to output one or more of the individual utterance examples, the search results, or the matching results to the individual patient. [Note C4] A medical conversation support method characterized in that the aforementioned speech examples described in any of Appendix C1 to C3 are associated with one or more of the following situations: initial consultation, follow-up consultation, admission, during hospitalization, or admission / discharge support center, and the aforementioned situations are entered when performing the search or matching. [Note C5] A medical conversation support method described in any of Appendix C1 to Appendix C3, The aforementioned utterance examples are linked to contradictory situations, The memory holds a conversation inconsistency evaluation program that has the function of evaluating inconsistencies in conversations, The additional utterance examples and one or more of the individual key point data or individual dialogue data are input into the conversation contradiction evaluation program. A medical conversation support method characterized by performing a function to detect or evaluate contradictions between the two parties. [Appendix C6] A medical conversation support method described in any of Appendix C1 to Appendix C3, The speech examples registered in the medical professional speech database are linked to a flag indicating whether or not they contain information about cancer diagnosis. The aforementioned processor, A medical conversation support method characterized by detecting that policy information regarding not disclosing cancer is stored in one or more of the individual key point data or the individual dialogue data, and then executing a function to control the search of medical conversations so that conversations containing information regarding cancer disclosure are not searched. [Note C7] A medical conversation support method described in any of Appendix C1 to Appendix C3, The aforementioned medical professional speech database contains the content of the examination explanations registered in the database. The processor controls the inclusion of the pre-consent examination explanation content in the medical conversation. A medical conversation support method characterized by performing a function that records as a log that the utterance was completed in the entire conversation. [Note C8] A medical conversation support method described in any of Appendix C1 to Appendix C3, The system associates and stores speech examples registered in the aforementioned medical professional speech database with one or more of the following: electronic medical profile information, electronic medical record template information or task information, checklist information, or sorting information. When the processor searches or matches the medical professional speech database based on the individual dialogue data, The method for displaying conversation items to be presented according to one or more of the following: items in the associated electronic medical profile field, items in the electronic medical record template, task information, checklist information, or sorting information, is managed. Furthermore, the medical conversation support method is characterized by performing a function that improves the visibility of which conversation items have been completed by linking the patient's responses to the conversation items and saving them. [Note C9] A medical conversation support method described in any of Appendix C1 to Appendix C3, The aforementioned medical professional speech database is registered with one or more associated functions: either the function to insert long explanatory text, or the function to insert documents to replace hospital names, department names, etc. A medical conversation support method characterized in that the processor performs a function to customize and control the inclusion of information specific to each medical institution in a medical conversation by replacing it with the function of performing the mail merge. [Note C10] A medical conversation support method described in any of Appendix C1 to Appendix C3, A medical conversation support method characterized by performing a function to create and register speech examples using the program based on either electronic medical record information or case reports. [Explanation of Symbols]

[0076] 1. In-hospital terminal (processor 19, memory 15, memory 16, output device 14, input device 13, communication interface 12) 2. Support System (Processor 29, Memory 25, Storage 26, Communication IF 22, Input / Output IF 23) 4. Medical systems (electronic medical records, electronic questionnaires, etc.) 20 servers

Claims

1. A medical conversation support system comprising a processor and memory, and based on medical conversation data, The aforementioned memory is A medical professional speech database containing examples of utterances used by medical professionals to conduct medical conversations, It holds a speech generation program that takes patient-healthcare worker conversation data as input and outputs additional speech sentences. The execution of the speech generation program by the processor, Taking individual patient-healthcare worker conversation data, which is a conversation between an individual patient and a healthcare worker, as input, it outputs additional individual utterance examples. A medical conversation support system characterized by taking the aforementioned additional individual utterance examples as input, searching or matching the aforementioned utterance examples based on the matching or similarity of the medical professional utterance database to obtain search results or matching results, and outputting one or more of the aforementioned individual utterance examples, search results, or matching results to the individual patient.

2. A medical conversation support system comprising a processor and memory, and based on medical conversation data, The aforementioned memory is A medical professional speech database containing examples of utterances used by medical professionals to conduct medical conversations, An individual dialogue database that holds individual dialogue data, which is a conversation between an individual patient and a healthcare professional, A database of individual key points that holds individual key point data extracted from the individual dialogue data, The system holds the aforementioned individual key point data and one or more of the aforementioned individual dialogue data, electronic medical record data, data on a data warehouse, or referral letter data, and outputs additional individual utterance examples. The execution of the speech generation program by the processor realizes the function of outputting the additional individual speech examples, A medical conversation support system characterized by taking the aforementioned additional individual utterance examples as input, searching or matching the aforementioned utterance examples based on the matching or similarity of the medical professional utterance database to obtain search results or matching results, and outputting one or more of the aforementioned individual utterance examples, search results, or matching results to the individual patient.

3. A medical conversation support system characterized in that the speech examples registered in the medical professional speech database described in Claim 1 or Claim 2 are linked to one or more of the following situations: initial consultation, follow-up consultation, admission, during hospitalization, or admission / discharge support center, and the situation is entered when performing the search or matching.

4. A medical conversation support system according to Claim 2, The aforementioned utterance examples are linked to contradictory situations, The memory holds a conversation inconsistency evaluation program that has the function of evaluating inconsistencies in conversations, The additional utterance examples and one or more of the individual key point data or individual dialogue data are input into the conversation contradiction evaluation program. A medical conversation support system characterized by detecting or evaluating contradictions between two parties.

5. A medical conversation support system according to Claim 2, The speech examples registered in the medical professional speech database are linked to a flag indicating whether or not they contain information about cancer diagnosis. The aforementioned processor, A medical conversation support system characterized in that, when it is detected that policy information indicating not to disclose cancer is stored in one or more of the individual key point data or the individual dialogue data, the system controls the search of the medical professional speech database so that conversations containing information about disclosing cancer are not searched.

6. A medical conversation support system according to claim 1 or claim 2, The aforementioned medical professional speech database contains the content of the examination explanations registered in the database. The processor controls the inclusion of the pre-consent examination explanation content in the medical conversation. A medical conversation support system characterized by recording as a log that the spoken content of the aforementioned examination explanation was completed.

7. A medical conversation support system according to Claim 2, The system associates and stores speech examples registered in the aforementioned medical professional speech database with one or more of the following: electronic medical profile information, electronic medical record template information or task information, checklist information, or sorting information. When the processor searches or matches the medical professional speech database based on the individual dialogue data, The method for displaying conversation items to be presented according to one or more of the following: items in the associated electronic medical profile field, items in the electronic medical record template, task information, checklist information, or sorting information, is managed. Furthermore, the medical conversation support system is characterized by improving the visibility of which conversation items have been completed by saving the patient's responses in association with the conversation items.

8. A medical conversation support system according to Claim 1 or Claim 2, The aforementioned medical professional speech database is registered with one or more associated functions, either for inserting long explanatory text or for inserting documents to replace hospital names, department names, or other medical institution-specific information. The medical conversation support system is characterized in that the processor controls the inclusion of customized information in medical conversations by replacing it with medical institution-specific information using the function of performing mail merge.

9. A medical conversation support system according to claim 1 or claim 2, A medical conversation support system characterized by having a function to create and register speech examples using the speech creation program from either electronic medical record information or case reports.

Images