KR20230161825A - An artificial intelligence-based hierarchical question and answer operation method through natural language processing - Google Patents

Abstract

The present invention discloses a hierarchical query and question-and-answer operation method through artificial intelligence-based natural language processing. This method includes (a) a medical staff terminal receiving patient data and transmitting diagnosis-related data to the hospital server; (b) the hospital server extracting keywords from the received patient data to derive a question, and providing a hierarchical query to the patient terminal based on the received data; (c) the patient terminal transmitting an answer corresponding to the hierarchical query; (d) the hospital server extracting keywords from the answers and performing artificial intelligence-based machine learning to create a learning model that derives a diagnosis; and (e) when a question that is not in the hierarchical query is received at the medical staff terminal, updating the hierarchical query; It is characterized by including.

Description

{An artificial intelligence-based hierarchical question and answer operation method through natural language processing}

The present invention relates to a hierarchical inquiry and question-and-answer operation method. In particular, in communication between patients and medical staff that occurs in a short period of time, keywords are extracted through hierarchical questions and answers between the patient terminal and the medical staff terminal to generate a learning model, Information that may be omitted due to lack of knowledge in the patient's awareness, information provided by the patient due to various medical conditions may be omitted by the relevant medical staff, or treatment in an incorrect direction due to errors in judgment may be omitted, and learning This is about a hierarchical query and question-and-answer operation method through artificial intelligence-based natural language processing that can efficiently and accurately diagnose patients by maintaining knowledge obtained through the model at all times to ensure that questions related to each situation are not missed.

In general, an artificial neural network refers to a model in which artificial neurons, which form a network through the combination of synapses, change the strength of the synapse connection through learning and have problem-solving capabilities.

Like other machine learning methods that learn from data, neural networks are typically rule-based programming and are used to solve a wide range of problems, such as difficult-to-solve image recognition or speech recognition, as well as text.

In other words, when there is data, a lot of research is being done to express it in a form that a computer can recognize (for example, in the case of images, a tool that expresses pixel information as a column vector) and apply it to learning, and these efforts are being made. As a result, various deep learning techniques such as deep neural networks, convolutional neural networks, and recurrent neural networks have been used in fields such as computer vision, speech recognition, natural language processing, and voice/signal processing. Application programs with excellent performance are being developed.

Among them, Natural Language Processing (NLP) is a technology that turns natural language, the language that people use on a daily basis, into artificial language so that computers can recognize it. This is a very important technique for data analysis.

Natural language processing consists of Morphological Analysis, Syntactic Parsing, Semantic Analysis, and Discourse Analysis, which convert existing data such as natural language into morphemes, and understand natural language through these analyses. Through this process, search terms, which are key keywords representing natural language, can be extracted.

Meanwhile, taking medical treatment as an example, it is very important to obtain patient information. What the patient says includes important and unimportant things, and the important things can be divided from the most important to the least important.

Among these, the most important part is to select and organize the important things, and the core of medical diagnosis is to dig into the patient's symptoms and signs by asking related questions about what is considered important among what the patient said.

However, the medical staff asking questions at this time may also ask related questions depending on what the patient says. In most cases, they are judged according to the situation at the time based on professional knowledge and long-term experience, but due to individual differences in medical staff, there are many possible cases. There is a possibility that important possibilities may not come to mind, leading to treatment taking a wrong direction.

In that case, there is a risk that unnecessary diagnostic tests may be prescribed or key test prescriptions may be omitted, which may result in cases where it takes a long time to confirm the patient's exact disease name or the correct diagnosis cannot be found.

In addition, patients lack medical knowledge and often do not know what is important and what is not important, so they speak based on their own experience, recommendations from those around them, or their own knowledge, and at this time, very important information is unintentionally omitted. You may.

On the other hand, on the medical staff's side, once important symptoms or signs are sorted out, diagnosis and treatment can be made more easily and safely. However, in realistic medical settings up to now, not only does it take a lot of time and effort to obtain such information, but it also requires a lot of time and effort, as well as extensive and diverse experience in the relevant medical field. There are limitations that require the sensitivity of an expert to act with knowledge.

Therefore, since it is difficult to solve these limitations in a short period of time during outpatient or hospitalization, if this questioning system is effectively systematized and applied clinically, it will reduce the time and effort required to obtain information about the patient's symptoms or signs and help make an accurate diagnosis. It is expected that it can be of great help.

In addition, if the artificial intelligence system learns based on the results already databased and the conversations between the medical staff and patients by linking the final diagnosis or results with the languages communicated between the patient and the medical staff, the artificial intelligence system will be more accurate. It is expected that it will have intelligence.

US 10769138 B2 (2020.09.08)

The problem that the present invention aims to solve is to extract keywords from the questions and answers transmitted between the patient terminal and the medical staff terminal, identify keywords from the appropriate questions of the medical staff and the corresponding patient answers, create a learning model, and create new experiences in actual treatment sites. The goal is to provide a hierarchical query and question-and-answer operation method through artificial intelligence-based natural language processing that can efficiently and accurately diagnose a patient by creating a learning model obtained during the diagnosis and treatment of a patient case.

In order to solve the above-described problem, the hierarchical query and question-and-answer operation method through artificial intelligence-based natural language processing according to one aspect of the present invention is (a) a medical staff terminal receives patient data and transmits diagnosis-related data to the hospital server. step; (b) the hospital server extracting keywords from the received patient data to derive a question, and providing a hierarchical query to the patient terminal based on the received data; (c) the patient terminal transmitting an answer corresponding to the hierarchical query; (d) the hospital server extracting keywords from the answers and performing artificial intelligence-based machine learning to create a learning model that derives a diagnosis; and (e) updating the hierarchical query when a question that is not included in the hierarchical query is received at the medical staff terminal.

Other specific details of the invention are included in the detailed description and drawings.

According to the present invention, in medical treatment, information about a patient's symptoms or signs is accurately organized through a question-and-answer process between the patient terminal and the medical staff terminal, thereby eliminating the possibility that the patient may omit important information or the medical staff being aware of it. It can reduce the omission of important information or expected possibilities.

In addition, by diagnosing and treating patients more accurately, easily and safely, not only can the time and effort required to obtain information about a patient's symptoms or signs be dramatically saved, but it can also reduce errors in judgment by medical staff and provide accurate treatment. It is possible to make a diagnostic decision and determine an accurate treatment direction.

The effects of the present invention are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description below.

Figure 1 is a block diagram of a system for implementing a hierarchical query and question-and-answer operation method through artificial intelligence-based natural language processing according to an embodiment of the present invention.
Figure 2 is a configuration diagram for explaining the overall operation of a hierarchical query and question-and-answer operation method through artificial intelligence-based natural language processing according to an embodiment of the present invention.
Figure 3 is a block diagram to explain in more detail the overall operation of the hierarchical query and question-and-answer operation method through artificial intelligence-based natural language processing according to an embodiment of the present invention.
FIG. 4A is a configuration diagram briefly explaining the overall operation in which a processor performs machine learning to create a learning model in the system shown in FIG. 1.
FIG. 4b is a configuration diagram illustrating the operation in which artificial intelligence is linked with the inquiry system and the diagnosis and treatment system while the processor performs machine learning in the system shown in FIG. 1.
FIG. 5 is a flowchart of a processor performing machine learning according to BERT in the system shown in FIG. 1.
FIG. 6 is a flowchart in which a processor determines important questions by considering the diagnosis in the system shown in FIG. 1 and performs machine learning to create a step-by-step hierarchical questionnaire based on them.
FIG. 7 is a configuration diagram illustrating a partial operation in which a processor creates a learning model by labeling medical staff questions and patient answers in the system shown in FIG. 1.
FIG. 8 is a configuration diagram illustrating a partial operation in which a processor creates a learning model by labeling patient answers and diagnoses in the system shown in FIG. 1.
Figure 9 is a flow chart to explain the overall operation of the hierarchical query and question-and-answer operation method through artificial intelligence-based natural language processing according to an embodiment of the present invention.
FIG. 10 is a flowchart for explaining the detailed operation of step S500 in the hierarchical query and question-and-answer operation method shown in FIG. 9.
FIG. 11 is a ladder chart to explain data and operations transmitted and received by each component of the system shown in FIG. 1.
Each of the drawings above illustrates medical-related content as an example to clarify the description of the invention to be described later, but is not limited to medical-related content.

The advantages and features of the present invention and methods for achieving them will become clear by referring to the embodiments described in detail below along with the accompanying drawings.

Here, each of the accompanying drawings below illustrates medical-related content as an example for convenience of explanation in order to clarify the description of the invention, but is not limited to medical-related content.

The present invention is not limited to the embodiments disclosed below, but may be implemented in various different forms. The present embodiments are merely provided to ensure that the disclosure of the present invention is complete and to those skilled in the art to which the present invention pertains. It is provided to fully inform the scope of the present invention, and the present invention is defined only by the scope of the claims.

The terminology used herein is for describing embodiments and is not intended to limit the invention. As used herein, singular forms also include plural forms, unless specifically stated otherwise in the context. As used in the specification, “comprises” and/or “comprising” does not exclude the presence or addition of one or more other elements in addition to the mentioned elements. Like reference numerals refer to like elements throughout the specification, and “and/or” includes each and every combination of one or more of the referenced elements. Although “first”, “second”, etc. are used to describe various components, these components are of course not limited by these terms. These terms are merely used to distinguish one component from another. Therefore, it goes without saying that the first component mentioned below may also be a second component within the technical spirit of the present invention.

Unless otherwise defined, all terms (including technical and scientific terms) used in this specification may be used with meanings commonly understood by those skilled in the art to which the present invention pertains. Additionally, terms defined in commonly used dictionaries are not interpreted ideally or excessively unless clearly specifically defined.

Spatially relative terms such as “below”, “beneath”, “lower”, “above”, “upper”, etc. are used as a single term as shown in the drawing. It can be used to easily describe the correlation between a component and other components. Spatially relative terms should be understood as terms that include different directions of components during use or operation in addition to the directions shown in the drawings. For example, if a component shown in a drawing is flipped over, a component described as “below” or “beneath” another component will be placed “above” the other component. You can. Accordingly, the illustrative term “down” may include both downward and upward directions. Components can also be oriented in other directions, so spatially relative terms can be interpreted according to orientation.

In the present invention, the hospital server 300 is a component that processes information by communicating with external devices, and may include an application server, computing server, file server, mail server, proxy server, and web server.

In the present invention, the communication unit 310 may include one or more components that enable communication with an external device, and may include, for example, at least one of a wired communication module, a wireless communication module, and a short-distance communication module.

In the present invention, the processor 320 is a controller, microcontroller, microprocessor, microcomputer, etc., and overall controls the organic operation of the communication unit 310 and the database 320. It refers to components that perform various decisions and calculations, and may be implemented by hardware, firmware, software, or a combination of these.

Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings.

Figure 1 is a block diagram of a system for implementing a hierarchical query and question-and-answer operation method through artificial intelligence-based natural language processing according to an embodiment of the present invention, which includes a patient terminal 100, a medical staff terminal 200, and a hospital server ( 300), and the hospital server 300 includes a communication unit 310, a processor 320, and a database 330.

Figure 2 is a configuration diagram for explaining the overall operation of a hierarchical query and question-and-answer operation method through artificial intelligence-based natural language processing according to an embodiment of the present invention.

Figure 3 is a block diagram to explain in more detail the overall operation of the hierarchical query and question-and-answer operation method through artificial intelligence-based natural language processing according to an embodiment of the present invention, which includes a patient terminal 100 and a patient file storage space ( 250) and a hospital server 300.

FIG. 4A is a configuration diagram briefly explaining the overall operation in which a processor performs machine learning to create a learning model in the system shown in FIG. 1.

FIG. 4b is a configuration diagram illustrating the operation in which artificial intelligence is linked with the inquiry system and the diagnosis and treatment system while the processor performs machine learning in the system shown in FIG. 1.

FIG. 5 is a flowchart of a processor performing machine learning according to BERT in the system shown in FIG. 1.

FIG. 6 is a flowchart in which a processor determines important questions by considering the diagnosis in the system shown in FIG. 1 and performs machine learning to create a step-by-step hierarchical questionnaire based on them.

FIG. 7 is a configuration diagram illustrating a partial operation in which a processor creates a learning model by labeling medical staff questions and patient answers in the system shown in FIG. 1.

FIG. 8 is a configuration diagram illustrating a partial operation in which a processor creates a learning model by labeling patient answers and diagnoses in the system shown in FIG. 1.

Figure 9 is a flow chart to explain the overall operation of the hierarchical query and question-and-answer operation method through artificial intelligence-based natural language processing according to an embodiment of the present invention.

FIG. 10 is a flowchart for explaining the detailed operation of step S500 in the hierarchical query and question-and-answer operation method shown in FIG. 9.

FIG. 11 is a ladder chart to explain data and operations transmitted and received by each component of the system shown in FIG. 1.

With reference to FIGS. 1 to 11, the operation of the hierarchical query and question-and-answer operation method through artificial intelligence-based natural language processing according to an embodiment of the present invention is briefly described as follows.

First, the medical staff terminal 200 receives the patient's data and transmits the diagnosis-related data to the hospital server 300 (S100).

In other words, the patient's symptoms are temporarily stored in the patient file storage space 250 in the form of an electronic file, and when the patient visits the hospital and the temporarily stored electronic file is activated in the medical office, the patient's obligation is already stored in the hospital server 300. It is saved in the record.

Here, the electronic file includes a document file recording the patient's symptoms in text, an audio file recording the patient's voice, a document file converted to text using a voice-to-text conversion program, Video files or photo files of the patient's symptoms or signs; It may include one or more of the document files that answered the pre-prepared questionnaire.

Additionally, the patient's data may include one or more of images or videos taken by the patient or the patient's guardian, voice recordings of the patient's voice, and text.

Next, the hospital server 300 extracts keywords from the received patient data to derive a question (S200) and provides a hierarchical query to the patient terminal 100 based on the received data (S300).

Here, the hierarchical query may be a diagnostic question protocol in which a plurality of questions are formed in a hierarchical tree structure to derive a diagnosis.

Next, the patient terminal 100 transmits an answer corresponding to the hierarchical query (S400).

The hospital server 300 extracts keywords from the answers and performs artificial intelligence-based machine learning to create a learning model that derives a diagnosis (S500).

That is, the patient's answer corresponding to the hierarchical query is learned (S510), and the patient's diagnosis is derived using keywords included in the patient's answer (S520).

In step S510, the processor 320 generates preprocessed data labeling the hierarchical query and the patient's answer, and uses this to generate learning data, so that when the medical staff terminal 200 inputs a question later, the processor 320 generates preprocessed data labeling the hierarchical query and the patient's answer. Using the data, the expected patient response is output.

In step S510, the processor 320 labels keywords and diagnosis names included in the patient's answers and generates words extracted from the hierarchical query as learning data to diagnose the patient.

In addition, the core question among the hierarchical queries is determined (S530), and the order and diagnosis name of the hierarchical query are labeled and matched (S540).

Meanwhile, the hospital server 300 stores the patient's answer in the database 330, and key items are displayed to determine the relationship between the checked part and the derived diagnosis through artificial intelligence to diagnose the patient's future symptoms. You can utilize it.

If a question that is not included in the hierarchical query is received at the medical staff terminal 200, the processor 320 updates the hierarchical query.

1 to 11, the organic operation of the hierarchical query and question-and-answer operation method through artificial intelligence-based natural language processing according to an embodiment of the present invention will be described in detail by way of example as follows.

The present invention relates to a method of obtaining accurate information essential for accurate diagnosis and treatment of patients in communication between patients and medical staff.

That is, the present invention generates a learning model by extracting keywords through hierarchical questions and answers between the patient terminal and the medical staff terminal in short-term communication between the patient and the medical staff.

For example, very important information may be omitted in the explanation of a patient who complains in a short period of time, and medical staff also listen to the patient's explanation or review the data and ask necessary questions based on their expertise, but in some cases, the patient's explanation may be omitted. You may not be aware of something important, or you may not have considered one of the many possible diagnoses.

In this case, there is a possibility that the medical staff's judgment may be wrong and the treatment may proceed in an inaccurate direction, so the present invention provides knowledge obtained through a learning model to prevent this in advance, and is always provided to ensure that questions related to each situation are not missed. By maintaining this, it is possible to efficiently and accurately diagnose the patient.

First, assume that the patient is exchanging various questions and answers with the medical staff clinician using the hierarchical question and answer operation method through artificial intelligence-based natural language processing of the present invention.

A diagnostic question protocol in which a clinician asks a plurality of questions, for example, from the first question (①) to the seventh question (⑦), through the medical staff terminal 200, as shown in FIG. 4B, in order to accurately diagnose a patient. Use .

In other words, the diagnostic question protocol consists of a query system in which multiple questions are formed in a hierarchical tree structure, artificial intelligence, and a diagnosis and treatment system.

Artificial intelligence receives key questions from the query system and performs machine learning based on keywords to create a learning model.

For example, as shown in Figure 4b, based on the patient's statement, the query system generates a plurality of queries in the first step, and among them, the key first question (①) is used to ask another question in the second step. Create multiple other queries.

Likewise, the query system uses key questions from the previous step to generate multiple queries up to the fourth step, forming a hierarchical query tree structure.

In this embodiment, the 1st to 4th steps and the 1st to 7th questions (① to ⑦) are illustrated for ease of understanding, but the 5th step and higher and the 8th question and higher can also be created.

The key questions generated in this way are input into artificial intelligence, and artificial intelligence uses them to enable the diagnosis and treatment system to derive an accurate diagnosis for the patient.

In addition, in the query system, questions are categorized into important symptoms, the next subcategory question corresponding to the major category question is asked, and a hierarchical classification system is created in which more detailed questions are asked according to the subcategory question. In some cases, new questions that did not exist in the query system are created. There are cases where it is done.

Therefore, in preparation for such cases, when a new question arises, it is important to update it by reflecting it in the inquiry system. Even if the patient feels the same symptoms, there are cases where the patient's residential area or dialect is different and it is expressed differently, so each time. An update is needed.

Additionally, because the overall inquiry system may be too vast, it is necessary to operate only a portion of the entire system depending on the medical staff's specialty and immediately reflect it in the entire system whenever an update is made.

To this end, as shown in Figure 4b, artificial intelligence can add questions related to the questions of the second stage among the updated contents to the second stage.

On the other hand, in situations where it is difficult for the patient to answer hierarchical questions, for example, when only elderly patients visit the hospital, there may be cases where the patient is unable to explain logically.

In this case, when the patient's guardian records the patient's symptoms or signs in a freelance manner on a user terminal such as a mobile phone or a memory medium such as a USB, the patient's guardian visits the hospital and stores the contents on the hospital's PC, which is the medical staff terminal 200. Additionally, a program equipped with the hierarchical inquiry and question-and-answer operation method of the present invention can extract important keywords, automatically enter them, and guide the patient or the patient's guardian to ask additional questions or answers.

In addition, the patient's symptoms can be recorded in advance as text in an electronic file at home, or, as shown in Figure 3, recorded as a voice file for a limited time and converted to text using a voice-to-text (voice to test) conversion program. , the contents of answers to pre-prepared questionnaires, etc., are temporarily stored in the patient file storage space 250, so that they are valid only when the patient is activated within the hospital's medical room, and are stored in the patient's medical record previously stored in the hospital server 300. It may be possible.

Then, the processor 320 in the hospital server 300 extracts words from the contents of the files in the first step and answers questions in the second step about the keywords focusing on key keywords, and answers these answers. Based on these questions, the third stage of questions is created and the patient is asked to answer them.

The answers may be in text format in the form of multiple choice or subjective answers, or they may be in text format converted by a voice-to-text conversion program.

The processor 320 integrates and organizes the answers in the various formats to complete the questionnaire for the patient.

Meanwhile, artificial intelligence allows the diagnosis and treatment system to perform diagnosis based on the patient's answers to various questions from the medical staff in the first to fourth steps.

Examples of the medical staff's questions (Q1~Q4) and the patient's answers (A1~A4) are as follows.

As an example, assume that a patient's liver function suddenly worsens while taking tuberculosis medication, and the medical staff questions the patient to find the cause.

Q1: Has the patient recently taken any health food or medicine that affects the body?

A1: Absolutely not.

Q2: What about herbal medicine?

A2: Absolutely not.

Q3: What about samgyetang?

A3: Oh, you ate that?

Q4: Is the ginseng in samgyetang also an herbal medicine? Have you ever eaten anything else?

A4: Absolutely not.

Q5: Have you ever cooked anything?

A5: I boiled bellflower root with sesame root and ate it.

Q6: That’s herbal medicine, right?

A6: Oh, really? I didn't know that bellflower root and sesame root were herbal medicine.

In the above example, since the patient's liver function has worsened at first, a questionnaire is provided for the patient or the patient to answer. When the patient is asked whether he has ever taken herbal medicine, he will indicate that he has not taken it.

However, when the medical staff asks if you ever eat samgyetang and you say you ate samgyetang, they conclude that the cause is the ginseng contained in samgyetang.

If there were no appropriate related questions as above, the information would have been recorded as if the patient had never taken any other medicines, such as herbal medicine, and ultimately the medical staff's judgment would have been incorrect.

Accordingly, if liver function increases during the treatment of another patient, previously the only list of answers to the questionnaire was herbal medicine, but after the interview with the patient, a list of answers other than herbal medicine and samgyetang is added to the questionnaire.

Therefore, when the medical staff has experienced the same medical case, the questionnaire received by the patient has a list of answers called samgyetang, so the medical staff does not have to forget to ask whether they ate samgyetang or not, because the questionnaire has already been updated and modified, so it can be used in the future for the same medical case. The patient is asked to answer the questions in the questionnaire using samgyetang.

In other words, the best questionnaire is created by minimizing errors caused by humans depending on the situation (for example, human error in whether or not a question is included in each situation, etc.) and by automatically updating the questionnaire on a regular basis.

As in the case of the patient in the above example, The operation of the hospital server 300 to extract keywords from the patient's answers and perform machine learning using artificial intelligence is as follows.

When the medical staff terminal 200 receives the patient's data and transmits the diagnosis-related data to the hospital server 300, as shown in Figure 4b, the inquiry system extracts keywords from the received patient data and generates a plurality of questions. do.

In addition, as shown in FIG. 11, hierarchical queries are provided to the patient terminal 100 based on data received from the medical staff terminal 200.

At this time, the sequence of hierarchical queries provided to the patient terminal 100 is learned by BERT (Bidirectional Encoder Representation from Transformer, hereinafter BERT), as shown in FIG. 5.

Here, BERT is an unsupervised dictionary learning model that solves general tasks through a deep learning model that applies a transformer structure to a large amount of widely collected natural language data. It is based on the weights generated during the learning process. It refers to a learning model that performs additional learning with a relatively small amount of natural language data produced in a specific field and elaborately revises the existing weights.

When the patient terminal 100 transmits an answer corresponding to a hierarchical query, the processor 320 generates preprocessed data labeling the medical staff question and the corresponding patient answer and reflects it in the learning model, as shown in FIG. 7. .

The processor 320 generates learning data using preprocessed data, and later, when the medical staff terminal 200 inputs a question, it can output the patient's expected answer using the learning data.

Next, the processor 320 in the hospital server 300 extracts keywords from the patient's answers and creates a learning model that derives a diagnosis.

That is, as shown in FIG. 8, the processor 320 generates data labeling the patient's answers and the corresponding diagnosis and reflects them in the learning model.

In addition, learning data for diagnosing a patient is generated using words extracted from the diagnostic question protocol shown in FIG. 4B in which a plurality of questions are formed in a hierarchical tree structure.

Next, the processor 320 within the hospital server 300 determines key questions among the hierarchical queries shown in FIG. 4B and matches the order number of the hierarchical queries with the diagnosis name.

For example, if the patient's diagnosis is 'pneumonia', keywords included in the patient's answers corresponding to the first question (①), third question (③), and fifth question (⑤) shown in Figure 4b If it is easy to diagnose pneumonia, preprocessing data is generated by labeling it as [Pneumonia-questions 1, 3, and 5] and reflected in the learning model.

In addition, when the latest medical technology related to the disease 'pneumonia' is disclosed, the hospital server 300 may update the database 330 by communicating with an external server through the communication unit 310.

In addition, the patient's response data to the medical staff's hierarchical inquiries are stored in the database 330 within the hospital server 300, including items marked as important and core by the medical staff and checked, and items that were later confirmed to have the disease. Artificial intelligence can determine the relevance of the patient's diagnosis and use it to diagnose the patient's symptoms later.

Therefore, in the present invention, the content learned through artificial intelligence constantly updates the questionnaire, and regardless of the question asked by the medical staff or the updated questionnaire, all of the patient's answers and the two final confirmations that are the conclusion of the medical staff are confirmed. By matching, a conclusion is drawn that which word was most important and that such a diagnosis should be made when that word (or combination of words) appears, creating a learning model that the question must be included.

In this way, the present invention extracts keywords from the questions and answers transmitted between the patient terminal and the medical staff terminal, identifies keywords among the medical staff's appropriate questions and the corresponding patient answers, and creates a learning model in a short time, like an actual clinical situation. In the case of possibilities omitted from the patient's explanation or based on the medical staff's judgment, the possibility of mistakes being omitted from the list or overlooked even when the patient is explaining is reduced, and an improved questionnaire is consistently applied on a regular basis. , Provides a hierarchical query and question-and-answer operation method through artificial intelligence-based natural language processing that can efficiently and accurately diagnose patients.

Through this, the present invention accurately organizes information about the patient's symptoms or signs in medical treatment through a question-and-answer process between the patient terminal and the medical staff terminal, making it easier and safer to diagnose and treat the patient. The time and effort required to obtain information about symptoms or signs can be dramatically saved.

The steps of the method or algorithm described in connection with embodiments of the present invention may be implemented directly in hardware, implemented as a software module executed by hardware, or a combination thereof. The software module may be RAM (Random Access Memory), ROM (Read Only Memory), EPROM (Erasable Programmable ROM), EEPROM (Electrically Erasable Programmable ROM), Flash Memory, hard disk, removable disk, CD-ROM, or It may reside on any type of computer-readable recording medium well known in the art to which the present invention pertains.

Above, embodiments of the present invention have been described with reference to the attached drawings, but those skilled in the art will understand that the present invention can be implemented in other specific forms without changing its technical idea or essential features. You will be able to understand it. Therefore, the embodiments described above should be understood in all respects as illustrative and not restrictive.

For example, in the above-described embodiments of the present invention, questions and answers between patients and medical staff are exemplified, but in social fields where various question-and-answer processes, including criminal investigations, are conducted, and various question-and-answer processes, including state inspections and hearings, In the field of politics, etc., the present invention can be used in various cases where questions and answers are operated through natural language processing based on artificial intelligence.

100: Patient terminal 200: Medical staff terminal
250: Patient file storage space 300: Hospital server
310: Communication unit 320: Processor
330: database

Claims (10)

(a) a medical staff terminal receiving patient data and transmitting diagnosis-related data to the hospital server;
(b) the hospital server extracting keywords from the received patient data to derive a question, and providing a hierarchical query to the patient terminal based on the received data;
(c) the patient terminal transmitting an answer corresponding to the hierarchical query;
(d) the hospital server extracting keywords from the answers and performing artificial intelligence-based machine learning to create a learning model that derives a diagnosis; and
(e) when a question that is not in the hierarchical query is received at the medical staff terminal, updating the hierarchical query;
Characterized in that it includes,
Hierarchical question and answer operation method using artificial intelligence-based natural language processing. According to claim 1,
In step (a),
Temporarily storing the patient's symptoms in the form of an electronic file in a patient file storage space; and
When the patient visits the hospital and the temporarily stored electronic file is activated in the medical office, storing it in the patient's medical record previously stored on the hospital server;
Characterized in that it includes,
Hierarchical question and answer operation method using artificial intelligence-based natural language processing. According to claim 2,
The electronic file is,
A document file recording the patient's symptoms in text, an audio file recording the patient's voice, a document file converted to text using a voice-to-text conversion program, a video file or photo file about the patient's symptoms or signs, Characterized by including one or more of the document files that answer a pre-prepared questionnaire,
Hierarchical question and answer operation method using artificial intelligence-based natural language processing. According to claim 1,
In step (a) above,
The patient's data is,
Characterized in that it includes at least one of an image or video taken by the patient or the patient's guardian, a voice recording of the patient's voice, and text,
Hierarchical question and answer operation method using artificial intelligence-based natural language processing. According to claim 1,
In step (b) above,
The above hierarchical query is
Characterized in that it is a diagnostic question protocol in which a plurality of questions are formed in a hierarchical tree structure to derive the diagnosis.
Hierarchical question and answer operation method using artificial intelligence-based natural language processing. According to claim 1,
In step (d),
(d-1) learning about the patient's answer corresponding to the hierarchical query;
(d-2) deriving the patient's diagnosis using keywords included in the patient's answer; and
(d-3) determining a key question among the hierarchical queries and labeling and matching the diagnostic name with the order of the hierarchical queries;
Characterized in that it includes,
Hierarchical question and answer operation method using artificial intelligence-based natural language processing. According to claim 6,
The step (d-1) is
Generating preprocessed data labeling the hierarchical query and the patient's answer;
Generating learning data using the preprocessed data; and
When the medical staff terminal inputs a question later, outputting the patient's expected answer using the learning data;
Characterized in that it includes,
Hierarchical question and answer operation method using artificial intelligence-based natural language processing. According to claim 7,
The step (d-2) is
labeling keywords included in the patient's answer and the diagnosis; and
generating words extracted from the hierarchical query as learning data for diagnosing the patient;
Characterized in that it includes,
Hierarchical question and answer operation method using artificial intelligence-based natural language processing. According to claim 8,
In step (d),
The hospital server stores the patient's answers in a database, and uses artificial intelligence to determine the relationship between the checked part and the derived diagnosis by displaying key items to diagnose the patients' symptoms. Characterized by,
Hierarchical question and answer operation method using artificial intelligence-based natural language processing. A computer program combined with a hardware computer and stored in a computer-readable recording medium to perform the hierarchical query and question-and-answer operation method through artificial intelligence-based natural language processing according to any one of claims 1 to 9.

Images