Medical question and answer method and device based on large language model

Abstract

The invention discloses a medical question and answer method and device based on a large language model, and is applied to the technical field of medical question and answer. Performing static inquiry based on the first chief complaint information to obtain static dialogue content; wherein the static dialogue content is associated with the basic information and the chief complaint of the user; performing multiple rounds of dynamic inquiry based on the first chief complaint information and the static dialogue content to obtain dynamic dialogue content; wherein the dynamic dialogue content is associated with the symptom and the medical examination condition of the user; and based on the first chief complaint information, the static dialogue content and the dynamic dialogue content, obtaining medical suggestion information associated with the first chief complaint information. Accurate medical question and answer processing can be realized, and the precision of medical question and answer and the accuracy of medical suggestions are effectively improved.

Description

Technical Field

[0001] This application relates to the technical field of medical Q&A, and in particular, to a medical Q&A method and device based on a large language model. Background Art

[0002] With the rapid development of medical informatization, medical Q&A systems, as a tool for assisting medical diagnosis and user consultation, have gradually received extensive attention. Through medical Q&A systems, users' questions can be answered promptly and quickly, and reasonable medical advice can be provided to users. Currently, most medical Q&A systems are decision tree-based medical Q&A systems, which are also known as expert systems and clinical decision support systems. Such systems often use fixed question templates to collect users' chief complaint information through multiple rounds of questioning, and accordingly provide corresponding medical advice to users. For example, existing medical Q&A systems first ask users whether they have dizziness. If the user answers yes, the medical Q&A system then asks about the user's blood pressure. If the user answers no, the medical Q&A system then asks about the user's diarrhea. These questions are usually fixed and have a certain order. The advantages of such systems are that the algorithms are clear and transparent, it is convenient to modify the question templates, and the accuracy of their medical advice is relatively considerable, which can be applied to most medical Q&A scenarios. However, the actual disease conditions of users are often complex and diverse, and fixed question templates are difficult to comprehensively cover the details of users' conditions, resulting in a decrease in the accuracy of the medical advice finally output by such systems. Summary of the Invention

[0003] Embodiments of this application provide a medical Q&A method and device based on a large language model to improve the accuracy of medical advice.

[0004] On the one hand, embodiments of this application provide a medical Q&A method based on a large language model, including the following steps: Obtain the first chief complaint information; Conduct static questioning based on the first chief complaint information to obtain static conversation content; wherein, the static conversation content is associated with the user's basic information and chief complaint; Conduct multi-round dynamic questioning based on the first chief complaint information and the static conversation content to obtain dynamic conversation content; wherein, the dynamic conversation content is associated with the user's symptoms and medical examination conditions; Based on the first chief complaint information, the static conversation content, and the dynamic conversation content, obtain medical advice information associated with the first chief complaint information.

[0005] On the other hand, embodiments of this application provide a medical Q&A device based on a large language model, including: An acquisition module, configured to obtain the first chief complaint information; A static inquiry module, configured to perform static inquiry based on the first chief complaint information to obtain static conversation content; wherein, the static conversation content is associated with the user's basic information and chief complaint; A dynamic inquiry module, configured to perform multi-round dynamic inquiry based on the first chief complaint information and the static conversation content to obtain dynamic conversation content; wherein, the dynamic conversation content is associated with the user's symptoms and medical examination conditions; A medical record processing module, configured to obtain medical advice information associated with the first chief complaint information based on the first chief complaint information, the static conversation content, and the dynamic conversation content.

[0006] According to a medical Q&A method and device based on a large language model provided by the present application, first, obtain the first chief complaint information; then, perform static inquiry based on the first chief complaint information to obtain static conversation content; wherein, the static conversation content is associated with the user's basic information and chief complaint; afterwards, perform multi-round dynamic inquiry based on the first chief complaint information and the static conversation content to obtain dynamic conversation content; wherein, the dynamic conversation content is associated with the user's symptoms and medical examination conditions; finally, obtain medical advice information associated with the first chief complaint information based on the first chief complaint information, the static conversation content, and the dynamic conversation content. According to the technical solution of the present application, considering the shallow disease details associated with the user's basic information, chief complaint, symptoms, and medical examination conditions, as well as the deep disease details associated with the user's symptoms and chief complaint, accurate medical Q&A processing is achieved, so as to effectively improve the accuracy of medical Q&A, enhance the accuracy of medical advice, and further improve the user experience in medical Q&A. Description of the Drawings

[0007] Figure 1 is a flowchart of a medical Q&A method based on a large language model provided by the present application; Figure 2 is an example diagram of an inquiry main interface provided by the present application; Figure 3 is an example diagram of the static inquiry stage provided by the present application; Figure 4 is another example diagram of the static inquiry stage provided by the present application; Figure 5 is an example diagram of the first-round dynamic inquiry provided by the present application; Figure 6 is an example diagram of the second-round dynamic inquiry provided by the present application; Figure 7 is an example diagram of the third-round dynamic inquiry provided by the present application; Figure 8 is another example diagram of the third-round dynamic inquiry provided by the present application; Figure 9 It is an example diagram of the medical advice provided by this application; Figure 10 It is another example diagram of the main inquiry interface provided by this application. Detailed implementation manners

[0008] In order to make the objectives, technical solutions and advantages of this application more clear and understandable, the following further details this application in combination with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not used to limit this application. The following further describes this application in combination with the drawings of the specification and specific embodiments. The described embodiments should not be regarded as a limitation of this application. All other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of this application. In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments. However, it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments and can be combined with each other without conflict. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0009] In each specific embodiment of this application, when it comes to relevant processing based on data related to the user's identity or characteristics, such as user information, user behavior data, user historical data, and user location information, the user's permission or consent will be obtained first. Moreover, the collection, use, and processing of these data will comply with the relevant laws, regulations, and standards of the relevant countries and regions. In addition, when this application embodiment needs to obtain the user's sensitive personal information, it will obtain the user's separate permission or separate consent through methods such as pop-up windows or jumping to a confirmation page. After clearly obtaining the user's separate permission or separate consent, the necessary user-related data for this application embodiment to operate normally will be obtained.

[0010] In view of the problems and defects existing in the related technologies, the embodiments of this application provide a medical Q&A method and device based on a large language model, aiming to effectively improve the accuracy of medical Q&A and enhance the accuracy of medical advice.

[0011] First, a medical Q&A method based on a large language model provided by this application will be elaborated in detail below in combination with the accompanying drawings.

[0012] A medical Q&A method based on a large language model provided by an embodiment of the present application can be applied to a terminal, or to a server, or can also be software running on a terminal or a server, etc. The terminal can be a tablet computer, a laptop computer, a desktop computer, etc., but is not limited thereto. The server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or can also be a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, Content Delivery Network (CDN), and big data and artificial intelligence platforms. In addition, the server can also be a node server in a blockchain network, but is not limited thereto. Among them, blockchain is a new application mode of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, and encryption algorithms.

[0013] Referring to Figure 1 , the medical Q&A method based on the large language model may include but is not limited to the following steps S101 - S104.

[0014] S101, obtain the first chief complaint information.

[0015] It should be noted that the first chief complaint information refers to the descriptive information about at least one of the user's health condition, medical history, or symptoms, and the first chief complaint information reflects the superficial disease details associated with the user's chief complaint.

[0016] In this step, during the medical Q&A process, the first chief complaint information is obtained first. Through the first chief complaint information, the user's condition can be initially understood, so as to conduct static inquiries and multi-round dynamic inquiries based on the first chief complaint information in the subsequent steps.

[0017] Optionally, the first chief complaint information can be set according to the actual situation, and the embodiment of the present application does not make specific limitations thereto. Exemplarily, the first chief complaint information can be "I have been feeling chest tightness for the past week"; or, the first chief complaint information can be "diarrhea", but is not limited thereto.

[0018] S102, conduct static inquiries based on the first chief complaint information to obtain static conversation content; wherein, the static conversation content is associated with the user's basic information and chief complaint.

[0019] It should be noted that the static inquiry stage is used to determine the user's basic information and further explore the user's main complaint, while the static conversation content refers to the result of the static inquiry stage, which is used to indicate the questions raised in the static inquiry stage and the responses to the questions, and is associated with the user's basic information and main complaint, that is, it reflects the superficial disease details associated with the user's basic information and the deep disease details associated with the user's main complaint. It is worth noting that the questions involved in the static inquiry stage are all fixed question templates.

[0020] In this step, after obtaining the first main complaint information, the static inquiry stage will be entered. In the static inquiry stage, based on the first main complaint information, one or more questions associated with the user's basic information and main complaint are called and provided to the user for answering, and then these questions and their responses are integrated into the static conversation content.

[0021] S103, conduct multi-round dynamic inquiries based on the first main complaint information and the static conversation content to obtain the dynamic conversation content; among them, the dynamic conversation content is associated with the user's symptoms and medical examination conditions.

[0022] It should be noted that the dynamic inquiry stage is used to deeply explore the user's symptoms and determine the medical examination conditions, while the dynamic conversation content refers to the result of the dynamic inquiry stage, which is used to indicate the questions raised in the dynamic inquiry stage and the responses to the questions, and is associated with the user's symptoms and medical examination conditions, that is, it reflects the superficial disease details and deep disease details associated with the user's symptoms and the superficial disease details associated with the user's medical examination conditions. It is worth noting that the questions involved in the dynamic inquiry stage are all questions generated in real time dynamically.

[0023] In this step, after the static inquiry stage ends, the dynamic inquiry stage will be entered. The dynamic inquiry stage is configured with multi-round dynamic inquiries. In each round of dynamic inquiry, based on the first main complaint information, the static conversation content, and the conversation content of the previous round of dynamic inquiry, one or more questions belonging to this round of dynamic inquiry and associated with the user's symptoms or medical examination conditions are dynamically generated and provided to the user for answering, and then these questions and their responses are integrated into the conversation content of this round of dynamic inquiry. Through multi-round dynamic inquiries, the conversation content of multi-round dynamic inquiries can be obtained, and then the dynamic conversation content can be obtained.

[0024] S104, obtain medical advice information associated with the first main complaint information based on the first main complaint information, the static conversation content, and the dynamic conversation content.

[0025] In this step, after the dynamic inquiry stage ends, the medical advice generation stage will be entered. In the medical advice generation stage, based on the first chief complaint information, the results of the static inquiry stage, and the results of the dynamic inquiry stage, the final medical advice information is generated, and the medical advice information is associated with the first chief complaint information, so as to implement medical Q&A processing.

[0026] It can be seen that in the embodiment of the present application, first, the first chief complaint information is obtained. The first chief complaint information reflects the superficial disease details associated with the user's chief complaint. Then, based on the first chief complaint information, a static inquiry is performed to obtain static conversation content. Through the multi-round Q&A in this static inquiry stage, the user's basic information can be accurately determined and the user's chief complaint can be further explored. The superficial disease details associated with the user's basic information and the deep disease details associated with the user's chief complaint can be effectively explored. After that, based on the first chief complaint information and the static conversation content, multi-round dynamic inquiries are performed to obtain dynamic conversation content. Through different rounds of dynamic inquiries, the user's symptoms and medical examination conditions can be understood from multiple perspectives. Further, the superficial disease details and deep disease details associated with the user's symptoms and the superficial disease details associated with the user's medical examination conditions can be explored, which is beneficial to comprehensively grasp the actual disease condition of the user. Finally, based on the first chief complaint information, the static conversation content, and the dynamic conversation content, medical advice information associated with the first chief complaint information is obtained, so as to implement accurate medical Q&A processing while fully considering the superficial disease details associated with the user's basic information, chief complaint, symptoms, and medical examination conditions, as well as the deep disease details associated with the user's symptoms and chief complaint. In this way, the accuracy of medical Q&A can be effectively improved, the accuracy of medical advice can be enhanced, and thus the user experience in medical Q&A can be improved.

[0027] In some embodiments, the implementation process of the above step S101 may include the following steps: Display an inquiry main interface; wherein, the inquiry main interface may include, but is not limited to, a conversation display area, a send control, and a conversation input area; in response to an input instruction in the conversation input area, obtain the chief complaint information corresponding to the input instruction in the conversation input area and display it in the conversation input area; in response to a trigger instruction for the send control, determine the chief complaint information corresponding to the input instruction in the conversation input area as the first chief complaint information and display it in the conversation display area.

[0028] In this embodiment, taking the terminal as an example, such as Figure 2As shown in the figure, first, the inquiry main interface is displayed on the terminal. Among them, the inquiry main interface may include, but is not limited to, a dialogue display area 100, a sending control 200, and a dialogue input area 300. The dialogue display area 100 is used to display the first chief complaint information, static dialogue content, dynamic dialogue content, etc.; the dialogue input area 300 is used to provide the user with the input of the chief complaint information or the editing of the existing chief complaint information; the sending control 200 is used to determine the chief complaint information input by the user as the first chief complaint information and display it in the dialogue display area 100 when triggered. Then, the user can input the chief complaint information in the dialogue input area 300. The terminal responds to the input instruction of the dialogue input area 300, obtains the chief complaint information corresponding to the input instruction of the dialogue input area 300, and synchronously displays the chief complaint information in the dialogue input area 300. It should be noted that if the user inputs the chief complaint information in text form, the terminal can directly obtain the chief complaint information corresponding to the input instruction of the dialogue input area 300; if the user inputs the chief complaint information in voice form, the terminal needs to convert the chief complaint information corresponding to the input instruction of the dialogue input area 300 into text form through a pre-trained speech recognition model. After the user inputs or edits the first chief complaint information, the user can trigger the sending control 200 to make the terminal determine the final first chief complaint information. When the sending control 200 is triggered, the terminal responds to its trigger instruction, determines the chief complaint information input by the user as the final first chief complaint information, and synchronously displays it in the dialogue display area 100, such as displaying the first chief complaint information of "stomachache". In this way, this embodiment realizes the acquisition of the first chief complaint information through the interactive display technology, realizes intuitive, transparent and traceable, what you see is what you get, and thus is conducive to improving the acquisition efficiency of the first chief complaint information.

[0029] Optionally, the triggering method of the sending control 200 may include, but is not limited to, clicking. The sending control 200 may be a click-to-determine button, or other scroll / slide selectors, drop-down menu selection units, etc., but is not limited thereto.

[0030] Optionally, after the inquiry main interface is displayed, preset medical Q&A prompt information is displayed in the dialogue content area to guide the user to accurately input the first chief complaint information, thereby helping to improve the accuracy of medical Q&A. Among them, the medical Q&A prompt information can be flexibly set according to the actual situation. Exemplarily, as Figure 2 shown, the medical Q&A prompt information may be "I am your robot doctor, familiar with more than 1,200 diseases. I can: 1. Help you judge the severity of the condition and whether you need to go to the hospital; 2. Make an appointment for the on-vehicle AI hospital to come to your door for blood and urine tests; 3. Contact a human doctor to see a doctor and deliver medicine to your door; 4. Scientifically refute rumors. Enter your question to start the experience!", but is not limited thereto.

[0031] In some embodiments, the above static conversation content may include, but is not limited to, a plurality of static question information and reply information for each static question information; the implementation process of step S102 may include the following steps S201 - S203.

[0032] S201, in response to the first chief complaint information, obtain a plurality of static question information from a preset question database.

[0033] It should be noted that the question database pre - stores a number of static question information, and these static question information are associated with the user's basic information and chief complaint. That is, the plurality of static question information may include at least one basic question information and at least one chief complaint question information. Among them, the basic question information is used to indicate questions related to the user's basic information in the static inquiry stage, and the chief complaint question information is used to indicate questions for asking at least one of the user's health status, medical history or symptoms in the static inquiry stage.

[0034] In this step, taking the terminal as an example, if the first chief complaint information is detected, it enters the static inquiry stage. In the static inquiry stage, the terminal first obtains a plurality of static question information from the preset question database, including at least one basic question information and at least one chief complaint question information.

[0035] Optionally, the specific content of the basic question information can be set according to the actual situation, and this embodiment does not make specific limitations. Exemplarily, the basic question information may include questions for asking the user's basic information such as name, gender, occupation, location, age, past medical history, past medication history, family history, current season, etc. For example, as Figure 3 shown, the basic question information may include "Please select your gender", "Name" and "Age", etc., but is not limited thereto.

[0036] Optionally, the specific content of the chief complaint question information can be set according to the actual situation, and this embodiment does not make specific limitations. For example, as Figure 4 shown, the chief complaint question information may be "Sleep", "Defecation", "Urination", "Do you have a runny nose, cough or expectoration?" or "Do you have polydipsia, polyphagia, polyuria, etc.?" etc. Of course, in addition to this, it can also be other types of chief complaint question information, such as "Do you have dysuria, hematuria or other uncomfortable urinary conditions?", "Do you have a rapid heartbeat, chest pain, shortness of breath or fainting?", "Do you have pale skin or lips, gum bleeding, bruises or swollen lymph nodes?", "Do you have a crooked mouth, fainting, convulsions, sudden weakness or numbness in hands and feet?", "Do you have joint swelling, pain or deformation?", etc., but is not limited thereto.

[0037] Optionally, in some embodiments, the static question information obtained from the question database can be applicable to all diseases, that is, the static question information has universality. Exemplarily, a first preset number of basic question information and a second preset number of chief complaint question information are randomly selected from the question database as the static question information for this static inquiry. For example, two basic question information, namely "Please select your gender" and "Name", are randomly selected from the question database, and one chief complaint question information, namely "Sleep", is randomly selected from the question database, and these question information are used as the static question information for this static inquiry.

[0038] In addition, in some other embodiments, the static question information obtained from the question database can also be only applicable to the first chief complaint information, that is, the static question information is targeted. Exemplarily, the static question information corresponding to the first chief complaint information is selected from the question database as the static question information for this static inquiry. For example, when the first chief complaint information is "stomachache", three basic question information corresponding to "diarrhea", namely "Please select your gender", "Name", "Previous medical history", and "Current season", are selected from the question database, and three chief complaint question information corresponding to "diarrhea", namely "Feces", "Urine", and "Do you have polydipsia, polyphagia, polyuria, etc.?" are selected from the question database, and these question information are used as the static question information for this static inquiry. It can be understood that a mapping relationship between multiple preset first chief complaint information and multiple preset static question information is pre-constructed in the question database, so that the static question information corresponding to the first chief complaint information can be accurately extracted from the question database during the static inquiry stage.

[0039] Of course, in some other embodiments, the basic question information obtained from the question database can also be applicable to all diseases, that is, the basic question information has universality. For the chief complaint question information, it is necessary to selectively call the chief complaint question information corresponding to the reply information of the foregoing basic question information from the question database based on the reply information of the foregoing basic question information. For example, the terminal can determine that the user is male through the reply information of the basic question information, and the terminal will call the chief complaint question information related to the prostate from the question database. The principle of this selective call is the symptoms involved in the high-incidence diseases of specific gender, occupation, region, age, previous medical history, previous medication history, family history, and current season, but it is not limited to this.

[0040] S202, display each static question information in the dialogue display area of the inquiry main interface. Each static question information is configured with a corresponding first response control in the dialogue display area. In response to the trigger instruction of the first response control of each static question information, obtain the reply information of each static question information and display it in the dialogue display area.

[0041] It should be noted that corresponding first response controls are configured for each static question information in the dialogue display area, and the first response control is used to obtain the reply information of the static question information when triggered.

[0042] In this step, taking the terminal as an example, as Figure 3 and Figure 4 shown, after obtaining each static question information, the terminal displays each static question information in the dialogue display area 100 of the inquiry main interface; subsequently, for each static question information, the user can select or input the reply information by triggering the first response control of the current static question information. When the first response control of the current static question information is triggered, the terminal will obtain the reply information of the current static question information and synchronously display it in the dialogue display area 100.

[0043] Optionally, the triggering method of the first response control may include but is not limited to clicking or inputting, etc. The first response control may be a click-to-confirm button or a character input box, or other scrolling / sliding selectors, dropdown menu selection units, etc., but is not limited thereto. In addition, there is at least one first response control configured for each static question information, and its specific quantity can be flexibly set according to the actual situation, and its specific examples will be elaborated in the subsequent embodiments.

[0044] Optionally, the display method of each static question information can be set according to the actual situation, and this embodiment does not make specific limitations on this. For example, the terminal may determine the first static question information as the current static question information, display the current static question information in the dialogue display area of the inquiry main interface, then in response to the trigger instruction of the first response control of the current static question information, obtain the reply information of the current static question information and display it in the dialogue display area, and then obtain the next static question information as the current static question information, and return to the step of displaying the current static question information in the dialogue display area of the inquiry main interface until there is corresponding reply information for all static question information. Another example is that the terminal can display all static question information in the dialogue display area of the inquiry main interface at the same time. For each static question information, in response to the trigger instruction of the first response control of the current static question information, obtain the reply information of the current static question information and display it in the dialogue display area, so that the reply information of multiple static question information can be obtained.

[0045] S203. Determine each static question information and the reply information of each static question information as static dialogue content.

[0046] In this step, taking the terminal as an example, after obtaining the reply information of each static question information, the terminal integrates each static question information and its reply question into static dialogue content, thereby realizing static inquiry.

[0047] It can be seen that in the static inquiry stage, in this embodiment, multiple static question information is first extracted from a preset question database. Subsequently, for each piece of static question information, in response to the trigger instruction of its first response control, the reply information of the current static question information is obtained and displayed in the dialogue display area. Finally, all the static question information and their reply questions are integrated into static dialogue content, thus completing the static inquiry. In this way, through multiple pieces of static question information, this embodiment can accurately determine the user's basic information from multiple perspectives and deeply explore the user's main complaints, effectively capture more shallow disease details associated with the user's basic information and deep disease details associated with the user's main complaints, and reduce the interference of potential diseases unrelated to the user's actual illness on medical Q&A, thereby helping to improve the accuracy of medical Q&A and the accuracy of medical advice.

[0048] In some embodiments, the implementation process of obtaining the reply information of each piece of static question information in response to the trigger instruction of the first response control of each piece of static question information and displaying it in the dialogue display area may include the following steps A01 - A02.

[0049] A01, in response to the trigger instruction of the first response control of each piece of basic question information, obtain the user's basic information corresponding to the trigger instruction of the first response control of each piece of basic question information as the reply information of each piece of basic question information, and display it in the dialogue display area. It should be noted that the user's basic information is used to indicate information associated with the user's basic information. The user's basic information may include, but is not limited to, the user's name, gender, occupation, location, age, past medical history, past medication history, family history, current season, etc.

[0050] In this step, taking the terminal as an example, as Figure 3 shown, after obtaining each piece of static question information, the terminal displays each piece of static question information in the dialogue display area 100 of the inquiry main interface, which includes basic question information. For each piece of basic question information, the user can select or input the reply information by triggering the first response control of the current basic question information. The reply information is the user's basic information. When the first response control of the current basic question information is triggered, the terminal will obtain the user's basic information corresponding to the trigger instruction of the first response control of the current basic question information, and use this user's basic information as the reply information of the current basic question information, and simultaneously display it synchronously in the dialogue display area 100.

[0051] Exemplarily, as Figure 3As shown, the multiple basic question information in this round of static inquiry includes "Name", "Please select your gender", and "Age". Among them: If the basic question information is "Name" and "Age", the first response controls configured for these two pieces of basic question information can both be one, for the user to enter their name and age, and the reply information for these two pieces of basic question information is the name and age entered by the user respectively; If the basic question information is "Please select your gender", the first response control configured for this piece of basic question information can be two, namely the first response control for indicating "Male" and the first response control for indicating "Female", for the user to select their gender. For example, if the user clicks the first response control for indicating "Female", then the reply information for this piece of basic question information is "Female".

[0052] Optionally, the display method of each piece of basic question information can be set according to the actual situation, and this embodiment does not make specific limitations in this regard. It can be understood that the examples of the display method of the basic question information here can follow the examples of the display method of the static question information in the foregoing embodiments, and will not be elaborated here.

[0053] A02. In response to the trigger instruction for the first response control of each main complaint question information, obtain the second main complaint information corresponding to the trigger instruction of the first response control of each main complaint question information as the reply information for each main complaint question information, and display it in the dialogue display area.

[0054] It should be noted that the second main complaint information refers to the descriptive information about at least one of the user's health status, medical history, or symptoms. Among them, the second main complaint information can include but is not limited to digestive system symptom information, respiratory system symptom information, etc. It should be noted that compared with the first main complaint information, the second main complaint information is more in line with the user's actual illness situation. Exemplarily, taking the digestive system symptom information as an example, the second main complaint information can include but is not limited to the information indicating the sleep situation, the information indicating the stool situation, the information indicating the urine situation, the information indicating whether there is urethral pain, hematuria, or other uncomfortable urine conditions, the information indicating whether there is polydipsia, polyphagia, or polyuria, but is not limited thereto. Another example is, taking the respiratory system symptom information as an example, the second main complaint information can include but is not limited to the information indicating whether there is a runny nose, cough, or expectoration, the information indicating whether the heart beats fast, chest pain, dyspnea, or fainting, the information indicating whether the skin and lips are pale, tooth bleeding, skin bruising, or lymph node enlargement, but is not limited thereto.

[0055] In this step, taking the terminal as an example, such as Figure 4As shown, after obtaining each piece of static problem information, the terminal displays each piece of static problem information in the conversation display area 100 of the inquiry main interface, which includes the main complaint problem information. For each piece of main complaint problem information, the user can select or input a reply message by triggering the first response control of the current main complaint problem information. The reply message is the second main complaint information. When the first response control of the current main complaint problem information is triggered, the terminal will obtain the second main complaint information corresponding to the trigger instruction of the first response control of the current main complaint problem information, and use this second main complaint information as the reply message of the current main complaint problem information, and simultaneously display it synchronously in the conversation display area 100.

[0056] Exemplarily, as Figure 4 shown, when the first main complaint information is "stomachache", the multiple main complaint problem information in this round of static inquiry includes "sleep", "bowel movement", "urination", "Do you have a runny nose, cough or expectoration?" and "Do you have polydipsia, polyphagia, polyuria, etc.?" Taking some of the main complaint problem information as examples for illustration: If the main complaint problem information is "sleep", the first response controls configured for this main complaint problem information can be three, namely the first response control for indicating "excellent", the first response control for indicating "good", and the first response control for indicating "poor", for the user to feedback their sleep situation. For example, if the user clicks the first response control for indicating "good", then the reply message for this main complaint problem information is "good", which is used to indicate that the user's sleep situation is average; If the main complaint problem information is "Do you have polydipsia, polyphagia, polyuria, etc.?" The first response controls configured for this main complaint problem information can be two, namely the first response control for indicating "yes" and the first response control for indicating "no". For example, if the user clicks the first response control for indicating "yes", then the reply message for this main complaint problem information is "yes", which is used to indicate that the user has dysuria, hematuria or other uncomfortable urination conditions.

[0057] Optionally, the display method of each piece of main complaint problem information can be set according to the actual situation, and this embodiment does not make specific limitations on this. It can be understood that the examples of the display method of the main complaint problem information here can follow the examples of the display method of the static problem information in the foregoing embodiments, and will not be elaborated here.

[0058] Thus, this embodiment realizes the preliminary excavation of the details of the user's condition from two dimensions: the user's basic information and the user's main complaint, effectively captures the shallow condition details associated with the user's basic information and the deep condition details associated with the user's main complaint, and further excludes potential diseases irrelevant to the user's actual illness, reducing its interference with medical questions and answers, thereby effectively improving the accuracy of medical questions and answers and the accuracy of medical advice.

[0059] In some embodiments, the above dynamic conversation content may include, but is not limited to, the first dynamic conversation content, the second dynamic conversation content, and the third dynamic conversation content; the implementation process of the above step S103 may include the following steps S301 - S303.

[0060] S301, conduct the first round of dynamic inquiry based on the first chief complaint information and the static conversation content to obtain the first dynamic conversation content.

[0061] It should be noted that the first round of dynamic inquiry is used to initially explore the user's symptoms based on the first chief complaint information and the static conversation content. The first dynamic conversation content refers to the result of the first round of dynamic inquiry. The first dynamic conversation content is used to indicate the questions mentioned in the first round of dynamic inquiry and the responses to the questions, which reflects the superficial disease details associated with the user's symptoms.

[0062] In this step, taking the terminal as an example, the terminal can obtain the static conversation content through the previous static inquiry stage. Based on the first chief complaint information and the static conversation content, the terminal will conduct the first round of dynamic inquiry, aiming to initially explore the user's symptoms based on the previous conversation content, and then obtain the first dynamic conversation content.

[0063] S302, conduct the second round of dynamic inquiry based on the first chief complaint information, the static conversation content, and the first dynamic conversation content to obtain the second dynamic conversation content.

[0064] It should be noted that the second round of dynamic inquiry is used to further explore the user's symptoms based on the first chief complaint information, the static conversation content, and the result of the first round of dynamic inquiry. The second dynamic conversation content refers to the result of the second round of dynamic inquiry. The second dynamic conversation content is used to indicate the questions mentioned in the second round of dynamic inquiry and the responses to the questions, which reflects the deep disease details associated with the user's symptoms.

[0065] In this step, taking the terminal as an example, the terminal can obtain the first dynamic conversation content through the first round of dynamic inquiry. Based on the first chief complaint information, the static conversation content, and the result of the first round of dynamic inquiry, the terminal will conduct the second round of dynamic inquiry, aiming to further explore the user's symptoms based on the previous conversation content, and then obtain the second dynamic conversation content.

[0066] S303, conduct the third round of dynamic inquiry based on the first chief complaint information, the static conversation content, the first dynamic conversation content, and the second dynamic conversation content to obtain the third dynamic conversation content.

[0067] It should be noted that the third round of dynamic inquiry is used to explore the user's medical examination situation based on the first chief complaint information, static conversation content, and the first two rounds of dynamic inquiry. The third dynamic conversation content refers to the result of the third round of dynamic inquiry. The third dynamic conversation content is used to indicate the questions mentioned in the third round of dynamic inquiry and the responses to the questions, which reflects the superficial disease details associated with the user's medical examination situation.

[0068] In this step, taking the terminal as an example, through the first two rounds of dynamic inquiry, the terminal can obtain the first dynamic conversation content and the second dynamic conversation content. Based on the first chief complaint information, static conversation content, and the results of the first two rounds of dynamic inquiry, the terminal will conduct the third round of dynamic inquiry, aiming to explore the user's medical examination situation based on the foregoing conversation content, and then obtain the third dynamic conversation content.

[0069] Thus, in this embodiment, the dynamic inquiry stage is divided into three levels of dynamic inquiry rounds. The first round of dynamic inquiry is carried out based on the first chief complaint information and static conversation content, and the latter two rounds of dynamic inquiry are both carried out based on the first chief complaint information, static conversation content, and the results of the previous rounds of dynamic inquiry. Through the first two rounds of dynamic inquiry, the superficial disease details and deep disease details associated with the user's symptoms can be gradually explored, and potential diseases unrelated to the user's actual disease condition can be further excluded, reducing its interference with medical Q&A, which plays a great role in narrowing the scope of the third round of dynamic inquiry. Through the last round of dynamic inquiry, the user's medical examination situation can be accurately determined, so as to explore more superficial disease details associated with the user's medical examination situation. Through this progressive dynamic inquiry method, the disease details of the user can be comprehensively covered, thereby effectively improving the accuracy of medical Q&A and the accuracy of medical advice.

[0070] In some embodiments, the above-mentioned first dynamic conversation content may include, but is not limited to, at least one first dynamic question information and the response information of each first dynamic question information, and the above-mentioned first dynamic question information is associated with the user's symptoms; the implementation process of the above step S301 may include the following steps B01 - B03.

[0071] B01, based on the first chief complaint information, static conversation content, and preset first prompt information, combined with a pre-trained large language model, obtain at least one first dynamic question information.

[0072] It should be noted that the first prompt information refers to the prompt information used to guide the large language model to generate the first dynamic question information in the first round of dynamic inquiry, and the first prompt information is associated with the user's symptoms and the diseases suffered. It is worth noting that the first dynamic question information refers to the questions associated with the user's symptoms in the first round of dynamic inquiry. Different from the main complaint question information in the static inquiry stage (the main complaint question information can also be used to inquire about the user's symptoms to a certain extent), the first dynamic question information is generated based on the first main complaint information, the static conversation content, and the large language model. The first dynamic question information has a certain degree of pertinence and accuracy for inquiring about the user's symptoms.

[0073] In this step, taking the terminal as an example, after the static inquiry stage is completed, the terminal inputs the first main complaint information, the static conversation content, and the preset first prompt information into the pre-trained large language model. In this large language model, symptom information with differential diagnostic value is first generated based on the first main complaint information, the static conversation content, and the first prompt information, and then at least one first dynamic question information is generated based on this symptom information to facilitate the subsequent first round of dynamic inquiry.

[0074] It can be understood that the large language model is trained based on a general text dataset; among them, the general text dataset can include, but is not limited to, various types of corpora in the medical field of Chinese and English types. An existing general text dataset can be used, or a required general text dataset can be constructed according to the actual situation. This embodiment does not make specific limitations on this. In addition, the type of the large language model can be flexibly set according to the actual situation, and this embodiment does not make specific limitations on this. For example, the large language model can be large language models such as GPT-3, Instruct-GPT, GLM, etc., but is not limited thereto.

[0075] Optionally, in some embodiments, the large language model can be configured in a preset server. During each round of dynamic inquiry, the terminal can upload the first main complaint information, the static conversation content, and the preset first prompt information to the server. The server outputs at least one first dynamic question information through the large language model configured locally and sends it to the terminal. Of course, in other embodiments, the large language model can also be configured in the terminal. During each round of dynamic inquiry, the terminal can directly input the first main complaint information, the static conversation content, and the preset first prompt information into the large language model configured locally, and output at least one first dynamic question information through the large language model, but is not limited thereto.

[0076] Optionally, to ensure the accuracy of the first round of dynamic inquiries, after generating at least one piece of first dynamic question information, for each piece of first dynamic question information, the terminal can determine whether the current first dynamic question information belongs to the main complaint question information in the static question information. If so, it means that the current first dynamic question information repeats a certain main complaint question information in the static question information. In this case, the current first dynamic question information will be deleted. Otherwise, it means that the current first dynamic question information does not repeat any of the main complaint question information in the static question information. In this case, the current first dynamic question information will be retained. For example, if the current first dynamic question information is "Do you have a fever?", and there is also a main complaint question information of "Do you have a fever?" in the static question information, this means that the question "Do you have a fever?" has been asked to the user during the static inquiry stage. In this case, the current first dynamic question information will be deleted.

[0077] Optionally, the first prompt message can be set according to the actual situation, and this embodiment does not make specific limitations in this regard. For example, the first prompt message can be "You are familiar with clinical guidelines and expert consensus. There is currently a user with the situation described above. What diseases might this user have? Which diseases need to be differentiated from these diseases? Please give symptoms with differential diagnostic value. Please note the distinction between symptoms and signs. Based on this, please only give questions asking about symptoms and do not answer other questions. Keep it simple and do not explain. For example, if fever has differential significance, output 'Do you have a fever?'", but it is not limited to this.

[0078] Optionally, the specific content of the first dynamic question information can be generated by the large language model according to the actual situation, and this embodiment does not make specific limitations in this regard. For example, referring to Figure 5 , when the first main complaint information is "stomachache", the first dynamic question information can be "Diarrhea situation?", "When did it start?", "Number of times per day?", or "Approximately how many milliliters each time?", but it is not limited to this.

[0079] Optionally, in addition to including questions, the first dynamic question information can also include a prompt message, which is used to guide the user to answer the first dynamic question information. The prompt message can be flexibly set according to the actual situation. For example, the prompt message can be "First round - Please select the corresponding symptom (diarrhea) below", but it is not limited to this.

[0080] B02, display each piece of first dynamic question information in the dialogue display area of the inquiry main interface. Each piece of first dynamic question information is configured with a corresponding second response control in the dialogue display area. In response to the trigger instruction for the second response control of each piece of first dynamic question information, obtain the first user symptom information corresponding to the trigger instruction for the second response control of each piece of first dynamic question information as the reply information for each piece of first dynamic question information, and display it in the dialogue display area.

[0081] It should be noted that each first dynamic problem information is configured with a corresponding second response control in the dialogue display area. The second response control is used to obtain the reply information of the first dynamic problem information when triggered, that is, the first user symptom information.

[0082] It can be understood that the first user symptom information is used to indicate the symptom information associated with the user's first chief complaint information, and it is the reply information of the first dynamic problem information. Among them, the first user symptom information may include, but is not limited to, digestive system symptom information, respiratory system symptom information, etc. Exemplarily, when the first chief complaint information is "stomachache", the first user symptom information may include, but is not limited to, information indicating whether there is diarrhea, the start time of diarrhea, the number of diarrhea times, and the degree of diarrhea (characterized by how many milliliters per diarrhea), but is not limited thereto. It should be noted that different from the second chief complaint information in the static inquiry stage (the second chief complaint information can also be used to reflect the user's symptoms to some extent), since the first user symptom information is the reply information of the first dynamic problem information, the first user symptom information is more targeted and accurate in describing the user's symptoms.

[0083] In this step, taking the terminal as an example, as Figure 5 shown, after obtaining each first dynamic problem information, the terminal displays each first dynamic problem information in the dialogue display area 100 of the inquiry main interface; subsequently, for each first dynamic problem information, the user can select or input the reply information by triggering the second response control of the current first dynamic problem information. The reply information is the first user symptom information. When the second response control of the current first dynamic problem information is triggered, the terminal will obtain the first user symptom information corresponding to the trigger instruction of the second response control of the current first dynamic problem information, and use this first user symptom information as the reply information of the current first dynamic problem information, and simultaneously display it synchronously in the dialogue display area 100.

[0084] Optionally, the triggering method of the second response control may include, but is not limited to, clicking or inputting, etc. The second response control may be a click-to-confirm button or a character input box, or other scroll / slide selectors, drop-down menu selection units, etc., but is not limited thereto. In addition, there is at least one second response control configured for each first dynamic problem information, and its specific quantity can be flexibly set according to the actual situation, and this embodiment does not make specific limitations on this.

[0085] Exemplarily, referring to Figure 5, in the case where the first chief complaint information is "stomachache", the multiple first dynamic question information in the first round of dynamic inquiry may include "Diarrhea situation?", "When did it start?", "Frequency per day?", and "Approximately how many milliliters each time?". Among them: If the first dynamic question information is "Diarrhea situation?", the second response controls configured for the first dynamic question information may be two, namely the second response control for indicating "Yes" and the second response control for indicating "No". For example, if the user clicks the first response control for indicating "Yes", the reply information for this first dynamic question information is "Yes", which is used to indicate that the user has diarrhea; If the first dynamic question information is "When did it start?", "Frequency per day?", and "Approximately how many milliliters each time?", the second response control for the first dynamic question information may be one, for the user to input the diarrhea start time, diarrhea frequency, and diarrhea degree (characterized by how many milliliters of diarrhea each time). At this time, the reply information for these first dynamic question information is the diarrhea start time, diarrhea frequency, and diarrhea degree (characterized by how many milliliters of diarrhea each time).

[0086] Optionally, the display manner of each first dynamic question information may be set according to the actual situation, and this embodiment does not make specific limitations in this regard. It can be understood that the examples of the display manner of the first dynamic question information here may follow the examples of the display manner of the static question information in the foregoing embodiments, and will not be elaborated here.

[0087] B03, determine each first dynamic question information and the reply information of each first dynamic question information as the first dynamic conversation content.

[0088] In this step, taking the terminal as an example, after obtaining the reply information of each first dynamic question information, the terminal integrates each first dynamic question information and its reply information into the first dynamic conversation content, thereby realizing the first round of dynamic inquiry.

[0089] It can be seen that in the first round of dynamic inquiry, in this embodiment, at least one first dynamic question information is first generated based on the first chief complaint information and the static conversation content in combination with the large language model, so as to effectively improve the question accuracy of the first round of dynamic inquiry, and make the first dynamic question information have a certain pertinence and accuracy for asking about the user's symptoms. Then, for each first dynamic question information, obtain its reply information, that is, the first user symptom information, by responding to the trigger instruction of its second response control. Finally, integrate each first dynamic question information and its reply information into the first dynamic conversation content for the subsequent second round of dynamic inquiry. In this way, it is possible to initially mine the user's symptoms based on the foregoing conversation content, effectively capture more shallow disease details associated with the user's symptoms, thereby helping to improve the accuracy of medical questions and answers and the accuracy of medical advice.

[0090] In some embodiments, the above-mentioned second dynamic conversation content may include, but is not limited to, at least one second dynamic question information and the reply information of each second dynamic question information, and the above-mentioned second dynamic question information is associated with the user's symptoms; the implementation process of the above-mentioned step S302 may include the following steps C01-C03.

[0091] C01, based on the first chief complaint information, static conversation content, first dynamic conversation content and preset second prompt information, and combined with a pre-trained large language model, obtain at least one second dynamic question information.

[0092] It should be noted that the second prompt information refers to the prompt information used to guide the large language model to generate the second dynamic question information in the second round of dynamic inquiry, and the second prompt information is associated with the user's symptoms and the diseases suffered. It is worth noting that the second dynamic question information refers to the question associated with the user's symptoms in the second round of dynamic inquiry. Different from the first dynamic question information in the first round of dynamic inquiry, the second dynamic question information is generated based on the first chief complaint information, static conversation content and first dynamic conversation content in combination with the large language model, and the second dynamic question information is more targeted and accurate for asking about the user's symptoms.

[0093] In this step, taking the terminal as an example, after the first round of dynamic inquiry is completed, the terminal inputs the first chief complaint information, static conversation content, first dynamic conversation content and second prompt information into the pre-trained large language model. In the large language model, first generate symptom information with differential diagnosis value based on the first chief complaint information, static conversation content, first dynamic conversation content and second prompt information, and then generate at least one second dynamic question information based on the symptom information.

[0094] It can be understood that the description of the large language model can follow the description in the foregoing embodiments and will not be elaborated here.

[0095] Optionally, to ensure the accuracy of the second-round dynamic inquiry, after generating at least one piece of second dynamic question information, for each piece of second dynamic question information, the terminal can determine whether the current second dynamic question information belongs to the main complaint question information among multiple pieces of first dynamic question information and static question information. If so, it means that the current second dynamic question information repeats a main complaint question information in a certain piece of first dynamic question information or static question information. In this case, the current second dynamic question information will be deleted. Otherwise, it means that the current second dynamic question information does not repeat any of the main complaint question information in all the first dynamic question information and static question information. In this case, the current second dynamic question information will be retained. For example, if the current second dynamic question information is "Do you have a fever?", and there is also a main complaint question information of "Do you have a fever?" in the static question information or the first dynamic question information, this means that the question "Do you have a fever?" has been asked to the user during the static inquiry stage or the first-round dynamic inquiry. In this case, the current second dynamic question information will be deleted.

[0096] Optionally, the second prompt information can be the same as the first prompt information or different from the first prompt information; the second prompt information can be set according to the actual situation, and this embodiment does not make specific limitations on this.

[0097] Optionally, the specific content of the second dynamic question information can be generated by the large language model according to the actual situation, and this embodiment does not make specific limitations on this. For example, referring to Figure 6 , when the first main complaint information is "stomachache", the second dynamic question information can be "Mass nodule?", "Acid reflux and heartburn?", "Abnormal stool?", or "Pain?"; of course, in addition to this, it can also be other types of second dynamic question information, such as "Nausea and vomiting?", "Constipation?", "Abnormal appetite?", or "Fatigue?", etc., but not limited to this.

[0098] Optionally, in addition to including questions, the second dynamic question information can also include a prompt message, which is used to guide the user to answer the second dynamic question information. The prompt message can be flexibly set according to the actual situation. For example, for the second dynamic question information of "Mass nodule?", the prompt message can be "Second round - Please select the following corresponding symptoms (mass nodule):", but not limited to this.

[0099] C02, display each piece of second dynamic question information in the dialogue display area of the inquiry main interface. Each piece of second dynamic question information is configured with a corresponding third response control in the dialogue display area. In response to the trigger instruction for the third response control of each piece of second dynamic question information, obtain the second user symptom information corresponding to the trigger instruction for the third response control of each piece of second dynamic question information as the reply information for each piece of second dynamic question information, and display it in the dialogue display area.

[0100] It should be noted that each second dynamic question information is configured with a corresponding third response control in the dialogue display area. The third response control is used to obtain the reply information of the second dynamic question information when triggered, that is, the second user symptom information.

[0101] It can be understood that the second user symptom information is used to indicate the symptom information associated with the user's first chief complaint information. The second user symptom information may include, but is not limited to, digestive system symptom information, respiratory system symptom information, etc. Exemplarily, when the first chief complaint information is "stomachache", the second user symptom information may include, but is not limited to, information indicating whether there are mass nodules, information indicating whether there is acid reflux and heartburn, information indicating whether the stool is abnormal, information indicating whether there is pain, information indicating whether there is nausea and vomiting, information indicating whether there is constipation, information indicating whether the appetite is abnormal, and information indicating whether there is fatigue, etc., but not limited thereto. It should be noted that different from the first user symptom information in the first-round dynamic inquiry, since the second user symptom information is the reply information of the second dynamic question information, the second user symptom information is more targeted and accurate in describing the user's symptoms.

[0102] In this step, taking the terminal as an example, as Figure 6 shown, after obtaining each second dynamic question information, the terminal displays each second dynamic question information in the dialogue display area 100 of the inquiry main interface; subsequently, for each second dynamic question information, the user can select or input the reply information by triggering the third response control of the current second dynamic question information. The reply information is the second user symptom information. When the third response control of the current second dynamic question information is triggered, the terminal will obtain the second user symptom information corresponding to the trigger instruction of the third response control of the current second dynamic question information, and use this second user symptom information as the reply information of the current second dynamic question information, and simultaneously display it synchronously in the dialogue display area 100.

[0103] Optionally, the triggering method of the third response control may include, but is not limited to, clicking or inputting, etc. The third response control may be a click-to-confirm button or a character input box, or other scroll / slide selectors, drop-down menu selection units, etc., but not limited thereto. In addition, there is at least one third response control configured for each second dynamic question information, and its specific quantity can be flexibly set according to the actual situation, and this embodiment does not make a specific limitation on this.

[0104] Exemplarily, referring to Figure 6, when the first chief complaint information is "stomachache", the multiple second dynamic question information in the second round of dynamic inquiry can include "lump or nodule?", "acid reflux or heartburn?", "abnormal stool?", and "pain?". The third response controls configured for these second dynamic question information can both be two, namely the second response control for indicating "yes" and the second response control for indicating "no". If the user clicks the first response control for indicating "yes", then the reply information for this second dynamic question information is "yes"; otherwise, the reply information for this second dynamic question information is "no". For example, for the second dynamic question information of "acid reflux or heartburn?", if the user clicks the first response control for indicating "yes", then the reply information for this second dynamic question information is "yes", indicating that the user has acid reflux or heartburn.

[0105] Optionally, the display mode of each second dynamic question information can be set according to the actual situation, and this embodiment does not make specific limitations on this. It can be understood that the examples of the display mode of the second dynamic question information here can follow the examples of the display mode of the static question information in the foregoing embodiments, and will not be elaborated here.

[0106] C03, determine each second dynamic question information and the reply information of each second dynamic question information as the second dynamic conversation content.

[0107] In this step, taking the terminal as an example, after obtaining the reply information of each second dynamic question information, the terminal integrates each second dynamic question information and its reply information into the second dynamic conversation content, thereby realizing the second round of dynamic inquiry.

[0108] It can be seen that in practical applications, due to defects such as hallucinations and short reasoning lengths in large language models, the information they can extract about the details of the user's condition (especially the user's symptoms) in the first-round dynamic inquiry is too little, resulting in difficulty in differentiating which information has high availability in the first-round dynamic inquiry. Therefore, it is difficult to comprehensively excavate the deep disease details associated with the user's symptoms only through the first-round dynamic inquiry. In response to this, this embodiment introduces a second-round dynamic inquiry after the first-round dynamic inquiry ends. First, based on the first chief complaint information, static conversation content, and first dynamic conversation content, at least one second dynamic question information is generated in combination with the large language model. In this way, potential diseases unrelated to the user's actual illness can be excluded through the first-round dynamic inquiry, reducing their interference with medical Q&A and narrowing the scope of the second-round dynamic inquiry. At the same time, compared with the first-round dynamic inquiry, the large language model can accurately capture information with high availability based on the first chief complaint information, static conversation content, and first dynamic conversation content, thereby making the generated second dynamic question information more targeted and accurate for asking about the user's symptoms, effectively improving the question accuracy of the second-round dynamic inquiry. Then, for each second dynamic question information, its response information, that is, the second user symptom information, is obtained by responding to the trigger instruction of its third response control. Finally, each second dynamic question information and its response information are integrated into the second dynamic conversation content for the last-round dynamic inquiry. In this way, the user's symptoms can be excavated deeper based on the foregoing conversation content, and the deep disease details associated with the user's symptoms can be comprehensively captured, thus helping to improve the accuracy of medical Q&A and the accuracy of medical advice.

[0109] In some embodiments, the above third dynamic conversation content may include, but is not limited to, at least one third dynamic question information and the response information of each third dynamic question information, and the above third dynamic question information is associated with the user's medical examination situation; the implementation process of the above step S303 may include the following steps D01-D03.

[0110] D01, based on the first chief complaint information, static conversation content, first dynamic conversation content, second dynamic conversation content, and preset third prompt information, in combination with a pre-trained large language model, obtain at least one third dynamic question information.

[0111] It should be noted that the third prompt information refers to the prompt information used to guide the large language model to generate the third dynamic question information in the third round of dynamic inquiry. The third prompt information is associated with the user's medical examination, imaging examination and other medical examination conditions. It is worth noting that the third dynamic question information refers to the questions associated with the user's medical examination conditions in the third round of dynamic inquiry. It is generated by combining the large language model based on the first chief complaint information, static conversation content and the results of the first two rounds of dynamic inquiry. The third dynamic question information can provide inquiries about the user's medical examination conditions based on the user's basic information and symptoms, so as to cover the disease details in the aspect of the user's medical examination conditions.

[0112] In this step, taking the terminal as an example, after the second round of dynamic inquiry is completed, the terminal inputs the first chief complaint information, static conversation content, first dynamic conversation content, second dynamic conversation content and third prompt information into the pre-trained large language model. In this large language model, medical test items and imaging examination items with differential diagnostic value are generated based on the first chief complaint information, static conversation content, first dynamic conversation content, second dynamic conversation content and third prompt information. Subsequently, at least one third dynamic question information is generated based on the medical test items and imaging examination items.

[0113] It can be understood that the description of the large language model can follow the description in the foregoing embodiments and will not be elaborated here.

[0114] Optionally, the third prompt information can be set according to the actual situation, and this embodiment does not make specific limitations on this. Exemplarily, the third prompt information can be "You are familiar with clinical guidelines and expert consensus. There is currently a patient with the situation described above. What diseases might this patient have? Which diseases need to be differentially diagnosed from these diseases? Please give medical test items and imaging examination items with differential diagnostic value. You only need to give questions asking for the names of the examination items and do not need to answer other questions. Answer simply without explanation. For example, if a blood routine test has differential significance, output 'Have you had a blood routine test and what are the values?'", but it is not limited to this.

[0115] Optionally, the specific content of the third dynamic question information can be generated by the large language model according to the actual situation, and this embodiment does not make specific limitations on this. For example, referring to Figure 7 , when the first chief complaint information is "stomachache", the third dynamic question information can be "Have you had the following examination items?", but it is not limited to this.

[0116] D02. Display each third dynamic question information in the dialogue display area of the inquiry main interface. Each third dynamic question information is configured with a corresponding fourth response control in the dialogue display area. In response to the trigger instruction for the fourth response control of each third dynamic question information, obtain the medical examination information corresponding to the trigger instruction for the fourth response control of each third dynamic question information as the reply information for each third dynamic question information, and display it in the dialogue display area.

[0117] It should be noted that each third dynamic question information is configured with a corresponding fourth response control in the dialogue display area. The fourth response control is used to obtain the reply information of the third dynamic question information, that is, medical examination information, when triggered.

[0118] It can be understood that the medical examination information is used to indicate the examination information associated with the user's first chief complaint information. The medical examination information may include, but is not limited to, the examination information of medical test items and the examination information of imaging examination items. Exemplarily, the examination information may include information indicating whether to perform a routine fecal examination, information indicating whether to perform a five-classification white blood cell examination, information indicating whether to perform a urine component analysis, etc., but is not limited thereto. In addition, exemplarily, the examination information may include the examination reports of medical test items and imaging examination items. It should be noted that the medical examination information helps to further master the details of the user's condition.

[0119] In this step, taking the terminal as an example, as Figure 7 shown, after obtaining each third dynamic question information, the terminal displays each third dynamic question information in the dialogue display area 100 of the inquiry main interface; subsequently, for each third dynamic question information, the user can enter the reply information by triggering the fourth response control of the current third dynamic question information. The reply information is the medical examination information. When the fourth response control of the current third dynamic question information is triggered, the terminal will obtain the medical examination information corresponding to the trigger instruction of the fourth response control of the current third dynamic question information, and use this medical examination information as the reply information of the current third dynamic question information, and at the same time synchronously display it in the dialogue display area 100.

[0120] Optionally, the triggering method of the fourth response control may include, but is not limited to, clicking or inputting, etc. The fourth response control may be a click-to-confirm button or a character input box, or other scrolling / sliding selectors, dropdown menu selection units, etc., but is not limited thereto. In addition, there is at least one fourth response control configured for each third dynamic question information, and its specific quantity can be flexibly set according to the actual situation, and this embodiment does not make a specific limitation on this.

[0121] Exemplarily, referring to Figure 7, when the first chief complaint information is "stomachache", the third dynamic question information can be "Have you undergone the following examination items?", and this third dynamic question information can be configured with six fourth response controls, namely, the fourth response control for indicating whether to undergo fecal routine examination, the fourth response control for indicating whether to undergo white blood cell five-classification examination, the fourth response control for indicating whether to undergo urine component analysis, the fourth response control for indicating whether to undergo urine dry chemistry analysis, the fourth response control for indicating whether to undergo hepatobiliary, spleen, and pancreas ultrasound examination, and the fourth response control for indicating that none of the above examinations have been performed (i.e., "none"). For example, if the user clicks the fourth response control for indicating whether to undergo fecal routine examination, then the reply information of this third dynamic question information can include the information for indicating to undergo fecal routine examination. Another example, if the user clicks the fourth response control for indicating whether to undergo white blood cell five-classification examination and the fourth response control for indicating whether to undergo urine component analysis, then the reply information of this third dynamic question information can include the information for indicating to undergo white blood cell five-classification examination and the information for indicating to undergo urine component analysis.

[0122] Optionally, in the above example, as Figure 7 shown, if the user clicks the fourth response control for indicating that none of the above examinations have been performed (i.e., "none"), then AI test appointment information is displayed in the dialogue display area 100. The AI test appointment information is used to indicate medical test items and imaging examination items with differential diagnostic value, which are generated by a large language model, such as fecal routine, white blood cell five-classification, urine component analysis, urine dry chemistry analysis, hepatobiliary, spleen, and pancreas ultrasound examination, etc. These medical test items and imaging examination items are all configured with corresponding sub-response controls. The terminal, in response to a trigger instruction for any one of the sub-response controls, determines the item corresponding to the trigger instruction of the sub-response control as the item to be tested and reports it to an authorized medical institution so that the medical institution can provide on-site examination services for the user. For example, if the user clicks the sub-response controls for fecal routine and urine dry chemistry analysis, the terminal, in response to its trigger instruction, determines fecal routine and urine dry chemistry analysis as the items to be tested by the user and reports them to an authorized medical institution. In addition, the AI test appointment information can also include a prompt for the AI test, such as "Based on the data you submitted, we recommend the following examinations. If you need to do them, please click on the corresponding examination name. If you don't need to do them, please click no. The platform provides an in-vehicle AI hospital and you can make an appointment for an on-site examination.", but it is not limited to this.

[0123] Another example, referring to Figure 8, when the first chief complaint information is "stomachache", the third dynamic question information can be "Have you undergone the following examination items?", and this third dynamic question information can be configured with six fourth response controls, namely the fourth response control for indicating whether a routine fecal examination has been performed, the fourth response control for indicating whether a five-classification white blood cell examination has been performed, the fourth response control for indicating whether a urine component analysis has been performed, the fourth response control for indicating whether a dry chemical urine analysis has been performed, the fourth response control for indicating whether a hepatobiliary spleen and pancreas ultrasound examination has been performed, and the fourth response control for indicating that none of the above examinations have been performed (i.e., "none"). When the user triggers any of the fourth response controls other than the fourth response control for indicating that none of the above examinations have been performed, the terminal, in response to its trigger instruction, pops up an examination upload pop-up window 400 in the dialogue display area 100. This examination upload pop-up window 400 includes a prompt message and an upload control 410. For example, the prompt message can be "Please upload your examination report.", but it is not limited to this. The user can trigger the upload control 410 to upload the examination report corresponding to the item of this fourth response control, and the terminal, in response to its trigger instruction, obtains the examination report corresponding to the trigger instruction of the upload control 410 as the reply information for this third dynamic question information.

[0124] For example, if the user clicks the fourth response control for indicating whether a routine fecal examination has been performed, then the user can upload the examination report of the routine fecal examination by clicking the upload control 410 of the examination upload pop-up window 400, and the terminal takes it as the reply information for this third dynamic question information. Another example, if the user clicks the fourth response control for indicating whether a five-classification white blood cell examination has been performed and the fourth response control for indicating whether a urine component analysis has been performed, then the user can upload the examination report of the five-classification white blood cell examination and the examination report of the urine component analysis by clicking the upload control 410 of the examination upload pop-up window 400, and the terminal takes it as the reply information for this third dynamic question information.

[0125] Of course, in some embodiments, this examination upload pop-up window 400 can also include a no-report control 420. If the user does not want to upload the corresponding examination report or does not have the corresponding examination report, the user can trigger this no-report control 420. At this time, the terminal, in response to its trigger instruction, only takes the medical examination information corresponding to the trigger instruction of the fourth response control triggered by the user as the reply information for this third dynamic question information. For example, the user enters the examination upload pop-up window 400 by clicking the fourth response control for indicating whether a routine fecal examination has been performed, and then the user can click the no-report control 420. At this time, the terminal, in response to its trigger instruction, takes the information for indicating a routine fecal examination as the reply information for this third dynamic question information.

[0126] Optionally, the display manner of each third dynamic question information can be set according to the actual situation, and this embodiment does not make specific limitations thereto. It can be understood that the examples of the display manner of the third dynamic question information here can follow the examples of the display manner of the static question information in the foregoing embodiments, and will not be elaborated herein.

[0127] D03, determine each third dynamic question information and the reply information of each third dynamic question information as the third dynamic conversation content.

[0128] In this step, taking the terminal as an example, after obtaining the reply information of each third dynamic question information, the terminal integrates each third dynamic question information and its reply information into the third dynamic conversation content, thereby realizing the third round of dynamic inquiry.

[0129] It can be seen that in the third round of dynamic inquiry, in this embodiment, at first, based on the first chief complaint information, the static conversation content, and the results of the previous two rounds of dynamic inquiry, at least one third dynamic question information is generated by combining with the large language model. Subsequently, for each third dynamic question information, its reply information, that is, the second user symptom information, is obtained by responding to the trigger instruction of its fourth response control. Finally, each third dynamic question information and its reply information are integrated into the third dynamic conversation content, thereby realizing the third round of dynamic inquiry. In this way, since the previous two rounds of dynamic inquiry further exclude potential diseases irrelevant to the user's actual disease condition, reduce their interference with medical Q&A, and narrow the scope of the third round of dynamic inquiry, therefore, the third round of dynamic inquiry is carried out by combining with the large language model based on the first chief complaint information, the static conversation content, and the results of the previous two rounds of dynamic inquiry, and in the third round of dynamic inquiry, the user's medical examination situation is accurately mined based on the foregoing conversation content. This can not only effectively improve the accuracy of the third dynamic question information and ensure the question accuracy of the third round of dynamic inquiry, but also effectively capture more shallow disease details associated with the user's medical examination situation. This shallow disease detail plays a crucial role in mining the user's actual disease condition, thereby helping to improve the accuracy of medical Q&A and the accuracy of medical advice.

[0130] In some embodiments, the above-mentioned dynamic conversation content may include, but is not limited to, the first dynamic conversation content, the second dynamic conversation content, and the third dynamic conversation content; the implementation process of the above-mentioned step S104 may include, but is not limited to, the following step S401: S401, based on the first chief complaint information, the static conversation content, the first dynamic conversation content, the second dynamic conversation content, the third dynamic conversation content, and the preset fourth prompt information, combine with the pre-trained large language model to obtain medical advice information associated with the first chief complaint information and display it in the conversation display area of the inquiry main interface.

[0131] It should be noted that the fourth prompt information is used to indicate the prompt information for guiding the large language model to generate medical advice information during the medical advice generation stage, and the fourth prompt information is associated with the user's disease and treatment drugs.

[0132] In this embodiment, taking the terminal as an example, as Figure 9 shown, the terminal can accurately capture the shallow disease details associated with the user's basic information and chief complaint, that is, the static conversation content, through the static inquiry stage, and can accurately capture the deep disease details associated with the user's symptoms and the shallow disease details associated with the user's medical examination situation through the progressive dynamic inquiry. Accordingly, the terminal inputs the first chief complaint information, the static conversation content, the first dynamic conversation content, the second dynamic conversation content, the third dynamic conversation content, and the fourth prompt information into the pre-trained large language model, generates medical advice information associated with the first chief complaint information through the large language model, and displays it in the conversation display area 100 of the inquiry main interface. In this way, considering the user's chief complaint information, the shallow disease details associated with the user's basic information, symptoms, and medical examination situation, and the deep disease details associated with the user's symptoms, accurate medical Q&A processing is realized, the accuracy of medical Q&A can be effectively improved, the accuracy of medical advice can be enhanced, and thus the user experience in medical Q&A can be improved.

[0133] Exemplarily, referring to Figure 9 , when the first chief complaint information is "stomachache", the finally output medical advice information can be "Based on the medical history and examination results you provided: 1. Western medicine may diagnose: acute bacillary dysentery. 2. Western medicine recommended treatment: ① General treatment: It is recommended to isolate and rest, and choose easily digestible and low-fiber foods, such as rice porridge, noodles, cooked vegetables, etc., and avoid greasy, spicy, and raw and cold foods to reduce intestinal irritation; ② Drug treatment: Drugs such as A and B can be used. 3. Traditional Chinese medicine may diagnose; damp-heat dysentery. 4. Traditional Chinese medicine recommended treatment: ① Drug treatment: Drugs such as C and D can be used.", but not limited to this. It should be emphasized that the finally output medical advice in the embodiments of the present application is not the final diagnosis result or treatment method, and the finally output medical advice is only for the user's reference. The user's final disease diagnosis result or treatment method still needs to be judged and decided by a professional doctor. That is, the embodiments of the present application do not directly determine the final diagnosis result or treatment method and do not belong to the diagnosis and treatment of diseases.

[0134] It can be understood that the description of the large language model can follow the description in the foregoing embodiments and will not be elaborated here.

[0135] Optionally, the fourth prompt message can be set according to the actual situation, and this embodiment does not make specific limitations thereto. For example, the fourth prompt message can be "You are familiar with clinical guidelines and expert consensus. There is a patient whose situation is as described above. Please judge according to the diagnostic criteria of each disease. What disease is this patient most likely to have? What are the treatment principles for this disease? According to the situation of this patient, the indications and contraindications of each drug, what is the preferred treatment drug? If the current information does not meet the diagnostic criteria of any disease, please output 'The current information does not meet the diagnostic criteria of known diseases.'", but it is not limited thereto.

[0136] Optionally, before generating and processing the medical advice information, a precondition can also be set. If the precondition is met, step S401 is executed; otherwise, a new dynamic inquiry stage is entered to achieve cyclic inquiry. Through cyclic inquiry, more comprehensive details about the user's actual illness situation can be obtained, which helps to improve the accuracy of medical Q&A. Among them, the precondition can be flexibly set according to the actual situation. Exemplarily, taking the terminal as an example, the precondition can be that the current data volume reaches a preset data volume threshold, where the current data volume refers to the sum of the information volumes of the first chief complaint information, static dialogue content, and dynamic dialogue content at the current moment. When the terminal determines that the precondition is not met, it means that due to the limited current information volume, the large language model may generate incorrect medical advice information. At this time, the terminal enters a new dynamic inquiry stage. In the Nth dynamic inquiry stage, N≥2, for each round of inquiry, the terminal will inquire based on the dynamic dialogue content, the first chief complaint information, and the static dialogue content; the dynamic dialogue content of the Nth dynamic inquiry stage can be obtained through three rounds of dynamic inquiry, and the dynamic dialogue contents of the first to the Nth dynamic inquiry stages are integrated into a new dynamic dialogue content and returned to the step of judging whether the precondition is met. When the terminal determines that the precondition is met, it means that the large language model has obtained sufficient information volume and can generate accurate medical advice information. At this time, the terminal will obtain the medical advice information based on the first chief complaint information, the static dialogue content, and the dynamic dialogue content (the dynamic dialogue content at this time includes the dynamic dialogue contents of all dynamic inquiry stages).

[0137] Optionally, as Figure 9As shown, after the medical advice information is displayed in the conversation content area, the above method may further include: displaying a home visit reminder message in the conversation content area. For example, the home visit reminder message may be "Do you need a professional doctor's diagnosis? If so, please directly reply 'Yes'. If not, no action is required.", but not limited thereto. Taking the terminal as an example, the user inputs "Yes" or "No" in the conversation input area 300. The terminal responds to the input instruction in the conversation input area 300, obtains the reply information corresponding to the input instruction in the conversation input area 300 and displays it in the conversation input area 300. Subsequently, the user triggers the send control 200. The terminal responds to the trigger instruction for the send control 200, displays the reply information corresponding to the input instruction in the conversation input area 300 in the conversation display area 100. At the same time, if the reply information is "Yes", the terminal reports to the authorized medical institution so that the authorized medical institution can provide the user with home visit or online diagnosis services. Otherwise, the terminal does nothing. Among them, the description of the send control 200 follows the description of any control in the foregoing embodiment and will not be elaborated herein.

[0138] Optionally, after the medical advice information is displayed in the conversation content area, the above method may further include: based on the first chief complaint information, static conversation content, first dynamic conversation content, second dynamic conversation content, third dynamic conversation content, medical advice and a preset fifth prompt information, combining with a pre-trained large language model to obtain the medical record information corresponding to the first chief complaint information. For example, taking the terminal as an example, the terminal inputs the first chief complaint information, static conversation content, first dynamic conversation content, second dynamic conversation content, third dynamic conversation content, medical advice and the fifth prompt information into the large language model, and generates the medical record information corresponding to the first chief complaint information through the large language model to provide it for the user's reference. In this way, the user can more intuitively understand their actual illness situation and the suggestions for their actual illness situation, thereby improving the user experience during medical Q&A. Among them, the fifth prompt information can be flexibly set according to the actual situation. For example, the fifth prompt information may be "You are familiar with the writing of clinical medical records. There is a patient with the situation as described above. Please sort out the patient information into a clinical medical record according to the writing norms of clinical medical records.", but not limited thereto.

[0139] In some embodiments, the above method may further include the following steps: Obtain the feedback information of the actual illness information and the medical advice information; wherein, the feedback information is used to indicate whether the medical advice information is accurate; if the feedback information is used to indicate that the medical advice information is inaccurate, then update the preset question database based on the actual illness information and the static conversation content to obtain the updated question database as the new question database.

[0140] In this embodiment, considering that the chief complaint problem information in the static inquiry stage is crucial for mining the deep disease details of the user's chief complaint, which will affect the accuracy of medical Q&A, this embodiment will update the question database in real time. Specifically, taking the terminal as an example, as Figure 10 shown, after the medical advice information is displayed in the conversation content area, the terminal can obtain the feedback information of the medical advice information. Specifically, the terminal displays evaluation guidance information in its conversation content area. For example, the evaluation guidance information can be "Please ask whether the current medical advice information is accurate. If so, please click 'Yes', otherwise please click 'No' and enter the disease you actually have." The evaluation guidance information is configured with two advice controls, namely the advice control for indicating 'Yes' and the advice control for indicating 'No', but it is not limited thereto; when the user triggers any one of the advice controls, the terminal responds to the trigger instruction of any one of the advice controls, obtains the feedback information corresponding to the trigger instruction of the advice control, and then the user can enter the disease they actually have in the conversation input area 300. The terminal responds to the input instruction of the conversation input area 300, obtains the disease information corresponding to the input instruction of the conversation input area 300 and displays it in the conversation input area 300. Then the user triggers the send control 200, and the terminal responds to the trigger instruction of the send control 200, determines the disease information corresponding to the input instruction of the conversation input area 300 as the actual disease information, and determines the feedback information corresponding to the trigger instruction of the first advice control as the feedback information of the medical advice information. This feedback information is used to indicate whether the medical advice information is accurate. After obtaining the feedback information, the terminal will judge whether the medical advice information is accurate based on this feedback information. If not, the terminal will update the question database used in the static inquiry stage based on the actual disease information and the static conversation content. Specifically, it is manifested as: generating new chief complaint problem information based on the actual disease information and the static conversation content in combination with the pre-trained large language model and adding it to the question database. In this way, without adjusting the large language model, the update process of the question database is realized, which can not only effectively improve the accuracy of static inquiry, thus improving the precision of medical Q&A, but also effectively reduce the update cost of the large language model in terms of questions.

[0141] Exemplarily, as Figure 10As shown, the static conversation content of a certain user includes information indicating male gender, information indicating an age of 40 years old, and information indicating dizziness. This user suffers from "cryptococcal encephalitis" and was misdiagnosed with "cryptococcal encephalitis" during the medical Q&A process because questions related to "cryptococcal encephalitis" were not asked. Based on this, the user provides feedback to the terminal, that is, the user clicks on the suggestion control indicating "no" and enters "cryptococcal encephalitis" in the conversation input area 300. At this time, the terminal will dynamically generate new chief complaint question information based on "cryptococcal encephalitis" and the user's static conversation content. For example, the newly added question is "Have you raised or come into contact with pigeons recently?" When encountering a male user aged 40 with dizziness again during the subsequent medical Q&A process, the terminal will actively ask "Have you raised or come into contact with pigeons recently?"

[0142] In addition, the embodiment of the present application also provides a medical Q&A device based on a large language model. The device may include: an acquisition module for acquiring first chief complaint information; a static inquiry module for performing static inquiries based on the first chief complaint information to obtain static conversation content, where the static conversation content is associated with the user's basic information and chief complaint; a dynamic inquiry module for performing multi-round dynamic inquiries based on the first chief complaint information and the static conversation content to obtain dynamic conversation content, where the dynamic conversation content is associated with the user's symptoms and medical examination conditions; a medical record processing module for obtaining medical advice information associated with the first chief complaint information based on the first chief complaint information, the static conversation content, and the dynamic conversation content.

[0143] The content in the above method embodiments is applicable to the device embodiments of the present application. The functions specifically implemented by the device embodiments of the present application are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.

[0144] In summary, the current medical Q&A systems based on large language models have problems such as insufficient guiding ability, the possibility of outputting two answers for the judgment of the same patient, and huge update costs. Compared with the fixed-question medical Q&A systems based solely on decision trees, the present application can actively guide users to provide valuable information, and even non-medical professionals can use this system without obstacles. Specifically, in the static inquiry stage, it helps the large language model narrow the semantic space to be retrieved, enabling the large language model used later to rule out numerous potential diseases and focus its attention within a very small range of diseases. Then, through the first two rounds of dynamic inquiries, potential diseases are further excluded, which plays a great role in narrowing the scope of the third round of dynamic inquiries (medical tests, imaging examinations). After the first round of dynamic questions, the second round of dynamic questions, the third round of dynamic questions and their answers are input, due to the huge amount of information, the large language model can accurately output medical suggestions. In addition, the present application can update the question database in the static inquiry stage without adjusting the large language model, so as to realize the update of the questions in the medical Q&A process, and can effectively reduce the update cost of the large language model in terms of questions.

[0145] Although the embodiments of the present application have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principle and spirit of the present application. The scope of the present application is defined by the claims and their equivalents.

[0146] The above is a specific description of the preferred embodiments of the present application, but the present application is not limited to the described embodiments. Those skilled in the art can also make various equivalent deformations or substitutions without departing from the spirit of the present application, and these equivalent deformations or substitutions are all included in the scope defined by the claims of the present application.

Claims

1. A medical Q&A method based on large language models, characterized in that, The method includes: Obtaining first chief complaint information; Performing static inquiry based on the first chief complaint information to obtain static conversation content; wherein, the static conversation content is associated with the user's basic information and chief complaint; Performing multi-round dynamic inquiry based on the first chief complaint information and the static conversation content to obtain dynamic conversation content; wherein, the dynamic conversation content is associated with the user's symptoms and medical examination conditions; Based on the first chief complaint information, the static conversation content, and the dynamic conversation content, obtaining medical advice information associated with the first chief complaint information.

2. The method according to claim 1, wherein The static conversation content includes multiple static question information and reply information for each of the static question information, and the multiple static question information includes at least one basic question information and at least one chief complaint question information; The performing static inquiry based on the first chief complaint information to obtain static conversation content includes: In response to the first chief complaint information, obtaining multiple pieces of the static question information from a preset question database; Displaying each piece of the static question information in a conversation display area of an inquiry main interface, and each piece of the static question information is configured with a corresponding first response control in the conversation display area. In response to a trigger instruction of the first response control for each piece of the static question information, obtaining the reply information for each piece of the static question information and displaying it in the conversation display area; Determining each piece of the static question information and the reply information for each piece of the static question information as the static conversation content.

3. The method according to claim 2, wherein The in response to a trigger instruction of the first response control for each piece of the static question information, obtaining the reply information for each piece of the static question information and displaying it in the conversation display area includes: In response to a trigger instruction of the first response control for each piece of the basic question information, obtaining the user's basic information corresponding to the trigger instruction of the first response control for each piece of the basic question information as the reply information for each piece of the basic question information, and displaying it in the conversation display area; In response to a trigger instruction of the first response control for each piece of the chief complaint question information, obtaining a second chief complaint information corresponding to the trigger instruction of the first response control for each piece of the chief complaint question information as the reply information for each piece of the chief complaint question information, and displaying it in the conversation display area.

4. The method according to claim 1, wherein The dynamic conversation content includes first dynamic conversation content, second dynamic conversation content, and third dynamic conversation content; the performing multi-round dynamic inquiry based on the first chief complaint information and the static conversation content to obtain dynamic conversation content includes: Performing a first-round dynamic inquiry based on the first chief complaint information and the static conversation content to obtain the first dynamic conversation content; Performing a second-round dynamic inquiry based on the first chief complaint information, the static conversation content, and the first dynamic conversation content to obtain the second dynamic conversation content; Performing a third-round dynamic inquiry based on the first chief complaint information, the static conversation content, the first dynamic conversation content, and the second dynamic conversation content to obtain the third dynamic conversation content.

5. The method according to claim 4, wherein The first dynamic conversation content includes at least one first dynamic question information and response information for each of the first dynamic question information, and the first dynamic question information is associated with the user's symptoms; the first-round dynamic inquiry based on the first chief complaint information and the static conversation content to obtain the first dynamic conversation content includes: Based on the first chief complaint information, the static conversation content, and the preset first prompt information, and in combination with a pre-trained large language model, obtain at least one of the first dynamic question information; Display each of the first dynamic question information in the conversation display area of the inquiry main interface. Each of the first dynamic question information is configured with a corresponding second response control in the conversation display area. In response to a trigger instruction for the second response control of each of the first dynamic question information, obtain first user symptom information corresponding to the trigger instruction for the second response control of each of the first dynamic question information as the response information for each of the first dynamic question information, and display it in the conversation display area; Determine each of the first dynamic question information and the response information for each of the first dynamic question information as the first dynamic conversation content.

6. The method according to claim 4, characterized in that, The second dynamic conversation content includes at least one second dynamic question information and response information for each of the second dynamic question information, and the second dynamic question information is associated with the user's symptoms; the second-round dynamic inquiry based on the first chief complaint information, the static conversation content, and the first dynamic conversation content to obtain the second dynamic conversation content includes: Based on the first chief complaint information, the static conversation content, the first dynamic conversation content, and the preset second prompt information, and in combination with a pre-trained large language model, obtain at least one of the second dynamic question information; Display each of the second dynamic question information in the conversation display area of the inquiry main interface. Each of the second dynamic question information is configured with a corresponding third response control in the conversation display area. In response to a trigger instruction for the third response control of each of the second dynamic question information, obtain second user symptom information corresponding to the trigger instruction for the third response control of each of the second dynamic question information as the response information for each of the second dynamic question information, and display it in the conversation display area; Determine each of the second dynamic question information and the response information for each of the second dynamic question information as the second dynamic conversation content.

7. The method according to claim 4, characterized in that, The third dynamic conversation content includes at least one third dynamic question information and response information for each of the third dynamic question information, and the third dynamic question information is associated with the user's medical examination situation; the third-round dynamic inquiry based on the first chief complaint information, the static conversation content, the first dynamic conversation content, and the second dynamic conversation content to obtain the third dynamic conversation content includes: Based on the first chief complaint information, the static conversation content, the first dynamic conversation content, the second dynamic conversation content, and the preset third prompt information, and in combination with a pre-trained large language model, obtain at least one of the third dynamic question information; Display each of the third dynamic question information in the dialogue display area of the inquiry main interface. Each of the third dynamic question information is configured with a corresponding fourth response control in the dialogue display area. In response to the trigger instruction for the fourth response control of each of the third dynamic question information, obtain the medical examination information corresponding to the trigger instruction of the fourth response control of each of the third dynamic question information as the reply information for each of the third dynamic question information, and display it in the dialogue display area; Determine each of the third dynamic question information and the reply information of each of the third dynamic question information as the third dynamic dialogue content.

8. The method according to claim 1, wherein The dynamic dialogue content includes the first dynamic dialogue content, the second dynamic dialogue content, and the third dynamic dialogue content; obtaining the medical advice information associated with the first chief complaint information based on the first chief complaint information, the static dialogue content, and the dynamic dialogue content includes: Based on the first chief complaint information, the static dialogue content, the first dynamic dialogue content, the second dynamic dialogue content, the third dynamic dialogue content, and the preset fourth prompt information, and in combination with a pre-trained large language model, obtain the medical advice information associated with the first chief complaint information and display it in the dialogue display area of the inquiry main interface.

9. The method according to claim 1, wherein The method further includes: Obtain the feedback information of the actual disease information and the medical advice information; wherein, the feedback information is used to indicate whether the medical advice information is accurate; If the feedback information is used to indicate that the medical advice information is inaccurate, update the preset question database based on the actual disease information and the static dialogue content, and obtain the updated question database as the new question database.

10. A medical Q&A device based on a large language model, characterized in that, The device includes: An acquisition module, configured to acquire the first chief complaint information; A static inquiry module, configured to perform static inquiry based on the first chief complaint information to obtain a static dialogue content; wherein, the static dialogue content is associated with the user's basic information and chief complaint; A dynamic inquiry module, configured to perform multi-round dynamic inquiry based on the first chief complaint information and the static dialogue content to obtain a dynamic dialogue content; wherein, the dynamic dialogue content is associated with the user's symptoms and medical examination conditions; A medical record processing module, configured to obtain the medical advice information associated with the first chief complaint information based on the first chief complaint information, the static dialogue content, and the dynamic dialogue content.

Images