Method and system for medical question answering using large language models

By using large language models in medical Q&A systems, combining diffusion thinking and centralized thinking, and generating medical dialogue responses, the problem of lack of holistic thinking in the existing system is solved, the accuracy and user experience of Q&A is improved, and the data is scarce High-efficiency performance in scenarios.

CN116595131BActive Publication Date: 2024-12-06INST OF AUTOMATION CHINESE ACAD OF SCI
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310295002.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-03-23
Publication Date
2024-12-06
Estimated Expiration
2043-03-23

AI Technical Summary

Technical Problem

When using large language models, existing medical Q&A systems lack holistic thinking and are difficult to effectively consider the patient's health status, medical history and current status, which affects the accuracy and user experience of Q&A responses.

Method used

By obtaining the user's medical conversation history content and inputting it with the first and second prompt instructions into a large language model, diversified sampling and decoding generate multiple responses, combining diffusion thinking and centralized thinking, generating replies to medical conversations, improving The depth and breadth of thinking.

Benefits of technology

Achieve more accurate medical Q&A responses, improve user experience, and can generate performance close to or exceed small sample inferences in zero-sample settings, showing effectiveness and flexibility in data scarcity scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116595131B_ABST
    Figure CN116595131B_ABST
Patent Text Reader

Abstract

The present invention provides a method and system for medical question and answer using a large language model, the method comprising: obtaining the medical conversation history content of the user; inputting the medical conversation history content and a first prompt instruction into the large language model, and obtaining multiple first responses based on the diversified sampling decoding of the large language model; inputting the medical conversation history content and multiple second prompt instructions into the large language model, and obtaining multiple second responses corresponding to each of the second prompt instructions; inputting the medical conversation history content, the multiple first responses, and the multiple second responses into the large language model, generating the reply content of the medical conversation and sending it to the user. The large language model is enabled to utilize holistic thinking, improve the depth and breadth of thinking, thereby being able to generate more accurate reply content and enhance the user experience.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method and system for medical question answering using a large language model. Background Art

[0002] Medical research based on artificial intelligence technology is growing rapidly. Medical conversational question answering (CQA) tasks aim to improve the efficiency of healthcare by providing patients with a range of professional medical services. CQA systems can improve patients' experience during clinical treatment by quickly responding to patients' needs and providing relevant medical information.

[0003] In recent years, using neural networks and language models to solve medical problems has become a viable solution. However, general language models often struggle to perform well in the medical field because they are only trained on small medical datasets and lack the ability to resolve rare symptoms and diseases. Recently, Large Language Models (LLMs) have demonstrated that this problem can be alleviated through contextual learning when provided with a small number of shot samples. The Chain of Thoughts (CoT) prompting method solves complex logical problems by breaking down reasoning into multiple thinking steps, which can be transferred to complex medical reasoning.

[0004] Medical CQA systems require strong medical reasoning and the ability to consider individual patient factors (think broadly and deeply). Currently, even with the use of CoT to break down complex questions, these tasks still lack the holistic thinking required for all relevant information, such as the patient's health status, medical history, and current condition, which may affect the accuracy of the Q&A response and affect the user experience. Summary of the invention

[0005] In view of the problems existing in the prior art, the present invention provides a method and system for medical question answering using a large language model.

[0006] In a first aspect, the present invention provides a method for medical question answering using a large language model, comprising:

[0007] Obtain the user's medical conversation history;

[0008] Inputting the medical conversation history content and the first prompt instruction into a large language model, and obtaining a plurality of first responses based on diversified sampling decoding of the large language model;

[0009] Inputting the medical conversation history content and a plurality of second prompt instructions into the large language model to obtain a plurality of second responses corresponding to each of the second prompt instructions;

[0010] Inputting the medical conversation history content, the plurality of first responses and the plurality of second responses into the large language model, generating reply content of the medical conversation and sending it to the user;

[0011] The large language model is trained based on a common Chinese and English text dataset.

[0012] Optionally, the content of the first prompt instruction includes:

[0013] “The doctor:” or “The doctor might wonder:”

[0014] Optionally, the content of the second prompt instruction is determined based on the record items of the medical record.

[0015] Optionally, the content of the second prompt instruction includes at least one of the following:

[0016] What is the chief complaint?

[0017] What is past history?

[0018] What is auxiliary examination?

[0019] What is the diagnosis?

[0020] What is the suggestion.

[0021] In a second aspect, the present invention further provides a system for performing medical question answering using a large language model, comprising:

[0022] The acquisition module is used to obtain the user's medical conversation history content;

[0023] A first input module, used for inputting the medical conversation history content and the first prompt instruction into a large language model, and obtaining a plurality of first responses based on diversified sampling decoding of the large language model;

[0024] A second input module, used for inputting the medical conversation history content and a plurality of second prompt instructions into the large language model, and obtaining a plurality of second responses corresponding to each of the second prompt instructions;

[0025] A reply module, configured to input the medical conversation history content, the plurality of first responses and the plurality of second responses into the large language model, generate reply content of the medical conversation and send it to the user;

[0026] The large language model is trained based on a common Chinese and English text dataset.

[0027] Optionally, the content of the first prompt instruction includes:

[0028] “The doctor:” or “The doctor might wonder:”

[0029] Optionally, the content of the second prompt instruction is determined based on the record items of the medical record.

[0030] Optionally, the content of the second prompt instruction includes at least one of the following:

[0031] What is the chief complaint?

[0032] What is past history?

[0033] What is auxiliary examination?

[0034] What is the diagnosis?

[0035] What is the suggestion.

[0036] In a third aspect, the present invention also provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for medical question-answering using a large language model as described in the first aspect above is implemented.

[0037] In a fourth aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for medical question-answering using a large language model as described in the first aspect above.

[0038] The method and system for medical question and answer using a large language model provided by the present invention input the medical conversation history content and a first prompt instruction into the large language model, and obtain multiple first responses based on the diversified sampling decoding of the large language model; input the medical conversation history content and multiple second prompt instructions into the large language model to obtain multiple second responses corresponding to each second prompt instruction; then input the medical conversation history content, multiple first responses and multiple second responses into the large language model to generate reply content for the medical conversation, so that the large language model can use holistic thinking to improve the depth and breadth of thinking, thereby being able to generate more accurate reply content and enhance the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0040] Figure 1 A flowchart of a method for performing medical question answering using a large language model provided by the present invention;

[0041] Figure 2 An example diagram of the present invention providing a medical question-answering method using a large language model through holistic thinking;

[0042] Figure 3 An example diagram of the difference between direct response generation and holistic thinking provided for the present invention;

[0043] Figure 4 A comparison chart of experimental results of holistic thinking and few-sample methods in multiple data sets provided by the present invention;

[0044] Figure 5 An example diagram showing the effects of different prompt texts provided by the present invention on holistic thinking performance;

[0045] Figure 6 A schematic diagram of the results of the HoT ablation experiment provided by the present invention;

[0046] Figure 7 A schematic diagram of the structure of a system for medical question answering using a large language model provided by the present invention;

[0047] Figure 8 This is a schematic structural diagram of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] Figure 1 A flowchart of a method for performing medical question answering using a large language model provided by the present invention is shown in FIG. Figure 1 As shown, the method comprises the following steps:

[0050] Step 100: Obtain the user's medical conversation history content.

[0051] Step 101: input the medical conversation history content and the first prompt instruction into a large language model, and obtain multiple first responses based on diversified sampling decoding of the large language model.

[0052] Step 102: Input the medical conversation history content and multiple second prompt instructions into a large language model to obtain multiple second responses corresponding to each second prompt instruction.

[0053] Step 103: Input the medical conversation history content, multiple first responses and multiple second responses into a large language model, generate the reply content of the medical conversation and send it to the user.

[0054] Among them, the large language model is trained based on a general Chinese and English text dataset.

[0055] Specifically, the executor of the method may be a system, device, equipment, etc. for medical question and answer. For ease of description, the medical question and answer system is referred to below.

[0056] The large language model used in the present invention can be an existing large language model, such as GPT-3, Instruct-GPT, GLM and other large language models.

[0057] The large language model used in the present invention is trained based on a Chinese and English general text dataset. The Chinese and English general text dataset contains various types of corpora in various fields such as Chinese and English news, novels, articles, dialogues, chats, comments, and reviews. The existing Chinese and English general text dataset can be used, or a Chinese and English general text dataset constructed as needed, without specific limitation.

[0058] When a user asks medical-related questions on an interactive platform (including but not limited to application APPs, mini-programs, web pages, software clients, etc.), the medical question and answer system can obtain the user's medical conversation history, including the user's questions and the doctor's answers.

[0059] After obtaining the user's medical conversation history content, the medical question and answer system can input the user's medical conversation history content and a first prompt instruction into a large language model. The first prompt instruction can be used to stimulate the divergent thinking of the large language model, so that the large language model can generate multiple different medical answers, i.e., first responses, based on the user's medical conversation history content through diversified sampling decoding, where the number of first responses generated can be set as needed.

[0060] Optionally, the content of the first prompt instruction may include: "Doctor:" or "Doctor may think:". Of course, the content of the first prompt instruction may also be other similar relatively simple prompt instructions, which are not specifically limited here.

[0061] After obtaining the user's medical conversation history content, the medical question-answering system can also input the user's medical conversation history content and multiple second prompt instructions into the large language model, and each second prompt instruction can be used to generate a corresponding response content, that is, a second response. The second response can be understood as the result of focused thinking of the large language model.

[0062] Optionally, the content of the second prompt instruction can be determined based on the record items of the medical record, so as to generate the patient's medical record information. For example, the prompt instruction of "What is the ..." can be used to generate the answer to each medical record item.

[0063] Optionally, the recording items of the medical record may include Chief Complaint, Past History, Auxiliary Examination, Diagnosis, Suggestion, etc.

[0064] Optionally, the content of the second prompt instruction may include at least one of the following:

[0065] (1) What is the chief complaint?

[0066] (2) What is the Past History?

[0067] (3) What is the Auxiliary Examination?

[0068] (4) What is the Diagnosis?

[0069] (5) What is the Suggestion?

[0070] After generating the first response and the second response, the medical question-and-answer system can input the user's medical conversation history content, the first response and the second response into a large language model to obtain the final reply content and return it to the user.

[0071] It should be noted that the execution order of step 101 and step 102 is not limited. For example, step 101 may be executed first and then step 102, or step 102 may be executed first and then step 101.

[0072] The present invention uses a large language model for medical question and answering. The entire process does not require any additional supervisory data, and only requires the input of medical conversation history content and related prompt instructions.

[0073] The method for medical question and answer using a large language model provided by the present invention comprises the following steps: inputting the medical conversation history content and a first prompt instruction into the large language model, and obtaining a plurality of first responses based on the diversified sampling decoding of the large language model; inputting the medical conversation history content and a plurality of second prompt instructions into the large language model, and obtaining a plurality of second responses corresponding to each second prompt instruction; and then inputting the medical conversation history content, the plurality of first responses, and the plurality of second responses into the large language model to generate reply content for the medical conversation, so that the large language model can utilize holistic thinking, improve the depth and breadth of thinking, and thereby generate more accurate reply content, thereby improving the user experience.

[0074] The present invention is further described in detail below in conjunction with specific embodiments.

[0075] In the real world, the use of holistic thinking enables doctors to consider various factors of the patient's condition during the consultation phase and provide a broad and in-depth understanding. Without this holistic thinking, incorrect medical advice may be generated and cause harm to the patient. Based on this, the present invention proposes to use a large language model to conduct medical Q&A through holistic thinking (Holistically Thought, hereinafter referred to as HoT).

[0076] Figure 2 An example diagram of using a large language model to answer medical questions through holistic thinking provided by the present invention, such as Figure 2 As shown in the figure, the whole process can be divided into three steps: diffuse thinking, focused thinking and response generation. First, use diversified decoding to obtain different thought contents; then, generate medical summaries through fixed prompts of medical records; finally, use medical dialogue history and the above two thinking results to generate responses.

[0077] In diffuse thinking, the goal is to generate diverse responses appropriate for medical conversation. Figure 3 An example diagram of the difference between direct response generation and holistic thinking provided by the present invention, such as Figure 3As shown in Figure 1, the original direct response (i.e., generated directly by a large language model) may lack careful consideration, have no detailed guidance, and sometimes have factual errors. The diverse responses generated by diffuse thinking can be regarded as candidate responses with different thought contents. Specifically, the fixed prompt "Doctor:" on the LLM can be used to generate |D| times of different content information, where the value of |D| can be set as needed. The LLM will generate different responses under diversified decoding, such as Figure 2 As shown, a fixed prompt is used to generate three different responses for this type of diffuse thinking. Each outcome may contain a potential solution to better respond to the medical CQA situation.

[0078] The purpose of focused thinking is to gather key information before generating a conversational response. This process is similar to medical conversation summary generation. Original medical conversation summary generation directly generates summary content, and key information may be missing during the generation process. The difference of focused thinking is that for a typical medical record, the manually designed prompts of LLM can be used to generate the results of each record item, which is less difficult to generate than generating a complete summary, and the generated results of each item in the medical record are shorter and more specific. Figure 2 As shown, given a typical medical record, the LLM is asked to generate answers to each item using the prompt “What is the…”, and the generated record contains some key information about the history of the medical conversation, which can be used as a thought chain for further reply generation.

[0079] Response generation combines the results of diffuse and convergent thinking, and the information generated by diffuse and convergent thinking can be used to prompt the LLM to obtain the final response. The generation results of diffuse and convergent thinking cover two direct thinking directions, which may help the LLM learn and integrate key information to generate responses. Although the generation of holistic thinking processes can be performed in a zero-shot setting, experiments show that such thinking processes can be induced by only prompting a few examples including thought chains, which does not require a large training dataset or modifying the weights of the language model.

[0080] To illustrate the effectiveness of the present invention, the present invention is experimentally explored on three different datasets, namely, MedDialog, COVID and CMDD datasets, as shown in Table 1.

[0081] Table 1 Comparison of different dataset types

[0082]

[0083] Overall performance comparison: Tables 2 and 3 show the experimental results on multiple data sets. The methods used for comparison include: large language model direct generation (Direct), CoT prompt (CoT) and a series of fine-tuning-based methods (Fine-tune). The HoT in the table is the method provided by the present invention. The experiment selected three large language models, GPT-3, Instruct-GPT and GLM, for comparison to illustrate their effectiveness. It can be found from the table that the performance of the method of the present invention is significantly better than the benchmark method, which also illustrates the effectiveness of the present invention.

[0084] Table 2 Experimental results in MedDialog and COVID datasets

[0085]

[0086] Table 3 Experimental results on the CMDD dataset

[0087]

[0088] In addition, the experiment also came to a series of conclusions:

[0089] 1. Through holistic thinking, LLM reasoning under zero-shot can achieve performance close to or even better than that of few-shot reasoning.

[0090] Contextual learning is a widely used technique for fast and large language models. It involves using a small portion of data (called a prompt) to improve the performance of a model without fine-tuning its parameters. Figure 4 This is a comparison chart of the experimental results of holistic thinking and few-sample methods in multiple data sets provided by the present invention, such as Figure 4 As shown, it can be seen that Holistically Thought in MedDialog and COVID exceeds the performance of Few-shot prompts in three evaluation metrics (BLEU-2, Metero, Nist-2) of three different LLMs. This shows that Holistic Thinking is competitive even compared to manually constructed Few-shot prompts, which shows its importance in data-scarce scenarios, such as COVID-19. In addition, Holistic Thinking is more flexible and has greater task adaptability.

[0091] 2. Simple prompts can effectively stimulate LLM's overall thinking ability. Figure 5 The following is an example diagram of the influence of different prompt texts provided by the present invention on the overall thinking performance, such as Figure 5As shown in the figure, by studying the medical answer generation ability under various prompt templates, it can be found that simple prompt templates such as "#1Doctor:" or "#2Doctor may think:" can effectively stimulate the overall thinking ability of LLM, while the performance of domain prompts #3-#5 with specific descriptions and advanced text instruction prompts #6-#8 is not much better than that of simple prompt templates.

[0092] 3. Concentrated thinking and divergent thinking are important for holistic thinking. Figure 6 The result schematic diagram of the HoT ablation experiment provided by the present invention is as follows: Figure 6 As shown, in the English datasets (MedDialog and COVID), both focused thinking and diffuse thinking alone have a positive impact compared to direct generation, however, the performance gain is significantly lower compared to the combined HoT. In general, both focused thinking and diffuse thinking are beneficial to holistic thinking. In addition, the experiments show that by leveraging holistic thinking, LLM is able to consider the depth and breadth of thinking, thereby generating more accurate answers.

[0093] The following is a description of the system for medical question answering using a large language model provided by the present invention. The system for medical question answering using a large language model described below and the method for medical question answering using a large language model described above can be referenced to each other.

[0094] Figure 7 A schematic diagram of the structure of the system for medical question answering using a large language model provided by the present invention, such as Figure 7 As shown, the system includes:

[0095] An acquisition module 700 is used to acquire the medical conversation history content of the user;

[0096] A first input module 710 is used to input the medical conversation history content and the first prompt instruction into the large language model, and obtain multiple first responses based on the diversified sampling decoding of the large language model;

[0097] The second input module 720 is used to input the medical conversation history content and multiple second prompt instructions into the large language model to obtain multiple second responses corresponding to each second prompt instruction;

[0098] A reply module 730, for inputting the medical conversation history content, the plurality of first responses and the plurality of second responses into the large language model, generating a reply content of the medical conversation and sending it to the user;

[0099] Among them, the large language model is trained based on a general Chinese and English text dataset.

[0100] Optionally, the content of the first prompt instruction includes:

[0101] “The doctor:” or “The doctor might wonder:”

[0102] Optionally, the content of the second prompt instruction is determined based on the record items of the medical record.

[0103] Optionally, the content of the second prompt instruction includes at least one of the following:

[0104] What is the chief complaint?

[0105] What is past history?

[0106] What is auxiliary examination?

[0107] What is the diagnosis?

[0108] What is the suggestion.

[0109] It should be noted here that the above-mentioned system provided by the present invention can implement all the method steps implemented by the above-mentioned method embodiment, and can achieve the same technical effect. The parts and beneficial effects that are the same as the method embodiment in this embodiment will not be described in detail here.

[0110] Figure 8 A schematic diagram of the structure of an electronic device provided by the present invention, such as Figure 8 As shown, the electronic device may include: a processor 810, a communication interface 820, a memory 830 and a communication bus 840, wherein the processor 810, the communication interface 820 and the memory 830 communicate with each other via the communication bus 840. The processor 810 may call the logic instructions in the memory 830 to execute any of the methods for performing medical question-answering using a large language model provided in the above embodiments.

[0111] In addition, the logic instructions in the above-mentioned memory 830 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0112] It should be noted here that the electronic device provided by the present invention can implement all the method steps implemented by the above-mentioned method embodiment, and can achieve the same technical effects. The parts and beneficial effects of this embodiment that are the same as the method embodiment will not be described in detail here.

[0113] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute any of the methods for medical question-answering using a large language model provided in the above-mentioned embodiments.

[0114] It should be noted here that the non-transitory computer-readable storage medium provided by the present invention can implement all the method steps implemented by the above-mentioned method embodiment, and can achieve the same technical effect. The parts and beneficial effects that are the same as the method embodiment in this embodiment will not be described in detail here.

[0115] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0116] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0117] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for medical question answering using a large language model, It is characterized in that include: Obtain the user's medical conversation history; Inputting the medical conversation history content and the first prompt instruction into a large language model, and obtaining a plurality of first responses based on diversified sampling decoding of the large language model; the first prompt instruction is used to stimulate the divergent thinking of the large language model, so that the large language model can generate a plurality of different medical answers, namely, the first responses, based on the medical conversation history content through diversified sampling decoding; Inputting the medical conversation history content and a plurality of second prompt instructions into the large language model, obtaining a plurality of second responses corresponding to each of the second prompt instructions; each of the second prompt instructions is used to generate a corresponding response content, namely, the second response, which is the result of focused thinking by the large language model; the content of the second prompt instruction is determined based on the record items of the medical record, so as to generate the medical record information of the patient; Inputting the medical conversation history content, the plurality of first responses and the plurality of second responses into the large language model, generating reply content of the medical conversation and sending it to the user; The large language model is trained based on a common Chinese and English text dataset.

2. The method for medical question answering using a large language model according to claim 1, It is characterized in that The content of the first prompt instruction includes: "The doctor:" or "The doctor might wonder:".

3. The method for medical question answering using a large language model according to claim 1, It is characterized in that The content of the second prompt instruction includes at least one of the following: What is the chief complaint? What is past history? What is auxiliary examination? What is the diagnosis? What is the suggestion.

4. A system for medical question answering using a large language model, It is characterized in that include: The acquisition module is used to obtain the user's medical conversation history content; A first input module is used to input the medical conversation history content and the first prompt instruction into a large language model, and obtain a plurality of first responses based on the diversified sampling decoding of the large language model; the first prompt instruction is used to stimulate the divergent thinking of the large language model, so that the large language model can generate a plurality of different medical answers, namely, the first responses, based on the medical conversation history content through diversified sampling decoding; a second input module, configured to input the medical conversation history content and a plurality of second prompt instructions into the large language model, and obtain a plurality of second responses corresponding to each of the second prompt instructions; each of the second prompt instructions is configured to generate a corresponding response content, namely, the second response, and the second response is the result of focused thinking by the large language model; the content of the second prompt instruction is determined based on the record items of the medical record, so as to generate the medical record information of the patient; A reply module, configured to input the medical conversation history content, the plurality of first responses and the plurality of second responses into the large language model, generate reply content of the medical conversation and send it to the user; The large language model is trained based on a common Chinese and English text dataset.

5. The system for medical question answering using a large language model according to claim 4, It is characterized in that The content of the first prompt instruction includes: "The doctor:" or "The doctor might wonder:".

6. The system for medical question answering using a large language model according to claim 4, It is characterized in that The content of the second prompt instruction includes at least one of the following: What is the chief complaint? What is past history? What is auxiliary examination? What is the diagnosis? What is the suggestion.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the program, the method for performing medical question answering using a large language model as described in any one of claims 1 to 3 is implemented.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the method for performing medical question answering using a large language model as described in any one of claims 1 to 3 is implemented.