Intelligent teaching method and device based on large language model and chat robot

Abstract

The invention relates to an intelligent teaching method and device based on a big language model and a chat robot. The method comprises the following steps: acquiring a medical problem input by a user through a chat robot of an instant messaging platform; inputting the medical question into a medical teaching model integrated by the chat robot to obtain a medical answer corresponding to the medical question, and outputting the medical answer through the chat robot; the medical teaching model is a large language model obtained through training according to corpora in the medical field; generating a user portrait of the user according to the medical problems, the medical answers and the learning behavior data of the user on the instant messaging platform in a preset historical time period under the condition that the timed teaching content generation event is detected; and pushing the teaching content corresponding to the user portrait to the user through the chat robot. By adopting the method, the intelligent degree of medical teaching can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of medical education information technology, in particular to an intelligent teaching method and device based on a large language model and a chat robot, a computer device, a computer readable storage medium and a computer program product. BACKGROUND

[0002] In recent years, with the gradual improvement of orthopedic training system, orthopedic specialized teaching system has made significant progress in curriculum setting, training process, examination mechanism, etc. However, clinical teaching still faces some bottlenecks: high-quality teaching resources are mainly concentrated in a few high-level medical institutions, teachers have heavy teaching burden and large individual capacity differences, traditional teaching mode has poor interaction and feedback lag, and it is difficult to provide differentiated teaching guidance for different students. Especially in the orthopedic teaching, which is highly dependent on clinical experience accumulation and skill training, the limitations of traditional teaching mode in efficiency and effectiveness are more prominent.

[0003] At present, the research and practice of intelligent teaching system are gradually developing, and some teaching platforms or systems have tried to introduce artificial intelligence technology into the education scene to realize personalized recommendation and intelligent management of teaching content. In the "smart classroom" and "artificial intelligence tutor" systems that have emerged in recent years, some platforms have begun to explore the use of natural language processing technology to realize knowledge pushing and auxiliary teaching functions based on question and answer systems. However, the current teaching method usually bases on keyword matching and rule engine for teaching, resulting in serious template and static problems of teaching content, so there is a problem of low intelligent degree of medical teaching. SUMMARY

[0004] Therefore, it is necessary to provide an intelligent teaching method, device, computer equipment, computer readable storage medium and computer program product based on a large language model and a chat robot, which can improve the intelligent degree of medical teaching.

[0005] In a first aspect, the present application provides an intelligent teaching method based on a large language model and a chat robot, comprising:

[0006] obtaining a medical question input by a user through a chat robot of an instant messaging platform;

[0007] inputting the medical question into a medical teaching model integrated with the chat robot to obtain a medical answer corresponding to the medical question, and outputting the medical answer through the chat robot; the medical teaching model is a large language model trained according to medical field corpus;

[0008] In a case where a timing teaching content generation event is detected, a user portrait of the user is generated according to the medical question, the medical answer, and learning behavior data of the user on the instant messaging platform within a preset historical time period;

[0009] The teaching content corresponding to the user portrait is pushed to the user by the chat robot.

[0010] In one of the embodiments, the medical question is input into the medical teaching model integrated with the chat robot to obtain a medical answer corresponding to the medical question, which includes:

[0011] A medical keyword in the medical question is recognized, and a standard answer corresponding to the medical keyword is queried in a pre-constructed medical knowledge graph;

[0012] The standard answer, the medical question, and identity information of the user are input into the medical teaching model integrated with the chat robot, and a customized answer matching the identity information is generated by the medical teaching model based on the standard answer as the medical answer corresponding to the medical question.

[0013] In one of the embodiments, before the step of generating the user portrait of the user according to the medical question, the medical answer, and the learning behavior data of the user on the instant messaging platform within a preset historical time period, the method further includes:

[0014] The medical question, the medical answer, and the learning behavior data of the user on the instant messaging platform within a preset historical time period are called from a database.

[0015] In one of the embodiments, the teaching content corresponding to the user portrait is pushed to the user by the chat robot, which includes:

[0016] According to the user portrait, a learning stage matching the user is determined;

[0017] According to the graphic and text data and the link data corresponding to the learning stage, the teaching content is generated;

[0018] The teaching content is pushed to the user by the chat robot.

[0019] In one of the embodiments, the learning behavior data includes at least one of the frequency of the user inputting the medical question, the learning frequency of the user on the teaching content, and the learning feedback result of the user on the teaching content.

[0020] In one of the embodiments, the chat robot accesses the instant messaging platform through an open source framework; the chat robot through the instant messaging platform acquires a medical question input by a user, including:

[0021] In a case where the message listening interface built through the open source framework listens to message content sent by the user to the chat robot of the instant messaging platform, the message content is structured to obtain the medical question input by the user.

[0022] In a second aspect, the application further provides an intelligent teaching device based on a large language model and a chat robot, including:

[0023] An acquisition module is configured to acquire a medical question input by a user through a chat robot of an instant messaging platform;

[0024] An answering module is configured to input the medical question into a medical teaching model integrated with the chat robot, to obtain a medical answer corresponding to the medical question, and to output the medical answer through the chat robot; the medical teaching model is a large language model trained according to a medical field corpus;

[0025] A generation module is configured to, in a case where a timing teaching content generation event is detected, generate a user portrait of the user according to the medical question, the medical answer, and learning behavior data of the user on the instant messaging platform within a preset historical time period;

[0026] A pushing module is configured to push teaching content corresponding to the user portrait to the user through the chat robot.

[0027] In a third aspect, the application further provides a computer device including a memory and a processor, the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0028] In a fourth aspect, the application further provides a computer readable storage medium having a computer program stored thereon, and the computer program implements the steps of the above method when executed by a processor.

[0029] In a fifth aspect, the application further provides a computer program product including a computer program, and the computer program implements the steps of the above method when executed by a processor.

[0030] The intelligent teaching method, device, computer device, computer readable storage medium and computer program product based on the large language model and the chat robot obtain a medical question input by a user through a chat robot of an instant messaging platform; input the medical question into a medical teaching model integrated with the chat robot to obtain a medical answer corresponding to the medical question, and output the medical answer through the chat robot, wherein the medical teaching model is a large language model trained according to a medical field corpus; in the case of detecting a timed teaching content generation event, a user portrait of the user is established based on medical questions, medical answers and learning behavior data of the user on the instant messaging platform in a preset historical time period; and the chat robot pushes teaching content corresponding to the user portrait to the user. The medical large language model and the instant messaging platform are deeply integrated, the chat robot for the medical scene is developed, the all-weather, low-threshold, semantic understanding driven teaching interaction and real-time question and answer feedback are realized; meanwhile, through intelligent data analysis and learning behavior modeling, the learning path intelligent recommendation and personalized resource pushing based on the user portrait can be realized, the bottleneck of the traditional rule type robot in semantic understanding, professional adaptation and interaction efficiency is broken, the closed-loop intelligent teaching platform integrating real-time semantic question and answer and personalized knowledge pushing is constructed, and the intelligent degree and effectiveness of medical teaching are significantly improved. BRIEF DESCRIPTION OF DRAWINGS

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related art, the drawings needed to be used in the description of the embodiments of the present application or the related art will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other related drawings can be obtained by those skilled in the art without creative labor.

[0032] Figure 1 An application environment diagram of an intelligent teaching method based on a large language model and a chat robot in an embodiment;

[0033] Figure 2 A flowchart of an intelligent teaching method based on a large language model and a chat robot in an embodiment;

[0034] Figure 3 A flowchart of an intelligent teaching method based on a large language model and a chat robot in another embodiment;

[0035] Figure 4 A structural block diagram of an intelligent teaching device based on a large language model and a chat robot in an embodiment;

[0036] Figure 5 An internal structure diagram of a computer device in an embodiment. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application and not to limit the present application.

[0038] In the related art, although the existing intelligent teaching system has been applied in some education scenarios, it still has obvious technical shortcomings in system architecture, functional adaptability, platform integration, and teaching interaction depth in the medical field (especially in orthopedic professional training), which cannot meet the complex needs of intelligent, instant, and personalized orthopedic teaching.

[0039] Firstly, the existing intelligent teaching system has poor adaptability to the WeChat platform, complex system access, and lack of instant interaction features. The existing platform-based intelligent teaching system is mostly based on Web-side independent deployment or relies on a closed teaching platform, which is not deeply integrated with WeChat, the most widely used instant messaging tool in China. This results in a high actual application threshold for the system, and medical students need to use additional terminals, Apps, or web pages for operation, which does not meet the needs of fragmented and mobile learning by orthopedic training physicians during clinical rotation. The lack of native access to WeChat also restricts the timeliness and response speed of the system in information pushing, knowledge questioning, and interactive feedback.

[0040] Secondly, the existing teaching system lacks targeted modeling of the knowledge system and task flow of orthopedic physician training, and has insufficient semantic understanding capability, which cannot meet the contextual analysis needs of professional clinical problems. Traditional question-and-answer type intelligent teaching tools mostly rely on keyword matching or pre-set logic templates, and when faced with complex structured and semantically rich case analysis, diagnosis and treatment processes, or operation step problems raised by medical students, they often cannot accurately understand the problem semantics and are difficult to provide clinically practical answer content. In addition, the current general-purpose large language model is stronger than the professional one, and lacks a professional corpus fine-tuning mechanism for orthopedics and other key departments, limiting its precise response capability in orthopedic teaching.

[0041] Thirdly, the existing system lacks a dynamic personalized pushing and intelligent feedback mechanism based on students' historical learning data, and the teaching resource allocation and teaching strategy adjustment lack data-driven support. Medical teaching has high individualization and stage characteristics, and there are significant differences in learning pace, ability level, and knowledge gaps among different orthopedic training students. Traditional teaching platforms cannot dynamically adjust teaching content and form according to students' learning paths and behavior trajectories, resulting in inaccurate teaching resource pushing to meet students' needs, which seriously affects learning efficiency and effectiveness. At the same time, teaching evaluation still mostly relies on manual sorting and periodic examination, and lacks continuous collection and analysis of process data of teaching activities, affecting real-time optimization of teaching quality.

[0042] In addition, the large language model itself is complex to deploy and consumes high resources, and there is currently a lack of large language model integration solutions suitable for medical teaching scenarios and stable operation under medium computing power conditions. Orthopedic teaching activities are high-frequency and interactive, and without a lightweight model deployment and optimization mechanism, it will put a lot of pressure on the server's concurrent processing capability. The existing solutions fail to propose effective system resource scheduling and edge computing adaptation strategies, and face performance bottlenecks and cost constraints in the widespread promotion of teaching.

[0043] In summary, the current technology has not solved the following core technical problems: (1) lack of native embedding of instant messaging platforms, instant response intelligent teaching tools; (2) insufficient semantic analysis ability of large language models and medical professional task adaptability; (3) lack of personalized learning resource pushing and feedback mechanism based on dynamic learning data; (4) the system architecture fails to effectively balance performance and deployment resource constraints of medical teaching scenarios. In summary, the most important technical bottleneck is the inability to implement an integrated intelligent teaching system that supports medical professional semantic understanding, personalized interaction, and teaching process closed-loop on instant messaging platforms.

[0044] The intelligent teaching method based on a large language model and a chat robot provided by the embodiments of the present application can be applied to an application environment as shown in Figure 1 . The terminal 102 communicates with the server 104 through a network. The data storage system can store data required to be processed by the server 104. The data storage system can be integrated on the server 104, or placed on the cloud or other network servers. The terminal 102 obtains a medical question input by a user through a chat robot of an instant messaging platform; the terminal 102 inputs the medical question to a medical teaching model integrated by the chat robot to obtain a medical answer corresponding to the medical question, and outputs the medical answer through the chat robot; the medical teaching model is a large language model trained according to medical field corpus; the terminal 102 generates a user portrait of the user according to medical questions, medical answers and learning behavior data of the user on the instant messaging platform in a preset historical time period, in a case where a timed teaching content generation event is detected; and the terminal 102 pushes teaching content corresponding to the user portrait to the user through the chat robot. The terminal 102 can be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The server 104 can be a standalone physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0045] In an exemplary embodiment, as shown in Figure 2 , an intelligent teaching method based on a large language model and a chat robot is provided. Taking the terminal 102 in Figure 1 as an example, the method includes the following steps:

[0046] In step S202, the terminal acquires the medical question input by the user through the chat robot of the instant messaging platform.

[0047] The instant messaging platform can refer to a social software or enterprise communication system supporting real-time messaging, can provide an open messaging interface, support multi-modal input (text, voice, picture, file), and have user identity authentication and session management capabilities.

[0048] The chat robot can refer to a medical teaching dedicated dialogue agent program deployed on the instant messaging platform, and can realize two-way communication based on the message protocol of the instant messaging platform. For example, the chat robot can listen to and preprocess the medical question input by the user in real time through a multi-protocol adaptation middleware, integrate an asynchronous non-blocking architecture, dynamic message filtering rules, and a multi-modal semantic analysis module, to provide high-quality input for subsequent large language model reasoning and user portrait construction.

[0049] The medical question can be a medical-related question, such as a question related to orthopedic training: "How to handle the discovery of a femoral comminuted fracture during surgery?" "Treatment method for femoral neck fracture", etc.

[0050] For example, the terminal can acquire the medical question input by the user through the chat robot of the instant messaging platform, which can include: acquiring the message content input by the user through the chat robot of the instant messaging platform, and matching the medical keywords (such as "fracture", "surgery", "medication") in the message content through regular expressions, and taking the message content containing the medical keywords as the medical question. Alternatively, the message content input by the user can be acquired through the chat robot of the instant messaging platform, and the relevance score of the message content to the medical field keywords can be calculated through the TF-IDF algorithm (term frequency-inverse document frequency), so as to filter the message content that is irrelevant content (such as chatting, advertising) according to the relevance score, and obtain the medical question.

[0051] For example, the terminal can acquire the medical question input by the user through the chat robot of the instant messaging platform, which can include: acquiring the message content input by the user through the chat robot of the instant messaging platform, and matching the medical keywords (such as "fracture", "surgery", "medication") in the message content through regular expressions, and taking the message content containing the medical keywords as the medical question. Alternatively, the message content input by the user can be acquired through the chat robot of the instant messaging platform, and the relevance score of the message content to the medical field keywords can be calculated through the TF-IDF algorithm (term frequency-inverse document frequency), so as to filter the message content that is irrelevant content (such as chatting, advertising) according to the relevance score, and obtain the medical question.

[0052] Step S204, input the medical question to the medical teaching model integrated with the chat robot, obtain the medical answer corresponding to the medical question, and output the medical answer through the chat robot.

[0053] The medical teaching model is a large language model trained according to medical field corpus.

[0054] The large language model (LLM) is a deep learning model trained using a large amount of text data, with strong language understanding and generation capabilities, capable of generating reasonable answers or continuing text based on input text, and performing well in various natural language processing tasks. For example, the large language model can include GPT architecture, BERT architecture, BART architecture, Llama architecture, etc. For example, the large language model can be ChatGLM, Baichuan, MiniCPM, Claude or LLaMA2, etc. have similar semantic understanding and natural language generation capabilities, or for resource-constrained scenarios, Alpaca, TinyLLaMA, etc. lightweight language model to support edge devices or private deployment.

[0055] In specific implementation, the medical teaching model is a special model based on the large language model architecture and fine-tuned with medical field corpus, capable of efficiently understanding and generating medical-related language content, and applied to medical education scenarios. The medical field corpus can include basic medical knowledge, clinical practice data, teaching interaction data, etc. For example, the medical field corpus can include publicly published materials such as "Practical Orthopedics" and "Surgery", as well as teaching materials compiled by the project team, supplemented by PubMed orthopedic field abstract texts as English academic support, thereby improving the model's semantic recognition and generation capabilities for orthopedic professional terminology, disease pathways, and operation specifications. In addition, the medical field corpus can also include medical examination databases and expert-annotated case dialogue data to improve the model's medical question and answer interaction capabilities.

[0056] Optionally, the medical teaching model can use a lightweight parameter adjustment strategy (such as Low-Rank Adaptation, LoRA) to significantly improve the model's understanding of Chinese medical context while considering resource consumption. The medical teaching model can be built through a FastAPI (a modern and fast web framework for building APIs) service interface, supporting concurrent processing and asynchronous response to ensure system stability and response efficiency in high-concurrency question scenarios.

[0057] For example, if the user inputs a medical question such as "treatment method for femoral neck fracture", the terminal can call the medical teaching model for semantic understanding and content generation, and return structured reply content including treatment suggestions, relevant guideline links and precautions as the medical answer output by the chat robot.

[0058] In one of the embodiments, the knowledge graph can be embedded into the medical teaching model to enhance the model's understanding of medical entity relationships (such as the triple association of disease-symptom-treatment method), and the medical question input to the medical teaching model integrated with the chat robot can include:

[0059] The medical entities in the medical question are identified, the relevant subgraph is retrieved from the pre-constructed medical knowledge graph, and the structured knowledge (such as classification criteria, surgical indications) obtained from the subgraph is converted into a natural language prompt (Prompt) and spliced into the original input medical question, and the medical question spliced with the natural language prompt is input to the medical teaching model integrated with the chat robot to obtain the medical answer corresponding to the medical question.

[0060] It should be noted that the subgraph in the knowledge graph refers to a local structure extracted from the entire graph, which has a specific theme or association relationship, contains a group of related nodes and their connection edges, for example, in the orthopedics knowledge graph, the "fracture treatment" subgraph can cover the nodes of "femoral neck fracture", "internal fixation", "rehabilitation training" and their "treatment plan", "surgical steps", "postoperative care" relationships; the node represents the basic entity or concept in the graph, each node has a unique identifier and attributes (such as disease name, pathological characteristics, ICD code), and is connected to other nodes through semantic relationship edges (such as "belongs to", "causes", "contraindication") to form a structured medical knowledge network, providing a traceable reasoning path for clinical decision-making.

[0061] For example, the original input medical question is "What is the Garden classification of femoral neck fracture?", and the medical question spliced with the natural language prompt is "What is the Garden classification of femoral neck fracture? [Knowledge background] Garden classification is divided into four types according to the degree of fracture displacement: type I is incomplete fracture……". By introducing the knowledge graph, the coverage of professional terms and the latest guidelines by the model can be significantly improved.

[0062] In one of the embodiments, the medical question is input into the medical teaching model integrated with the chat robot to obtain a medical answer corresponding to the medical question. The medical question and the identity information of the user can also be input into the medical teaching model integrated with the chat robot, and a customized answer matched with the identity information is generated by the medical teaching model as the medical answer corresponding to the medical question. Optionally, the identity information can include a learning stage label (such as “Orthopedic Training Physician-1st Year”) in the user portrait, and the medical teaching model can dynamically adjust the degree of detail and brevity of the generated medical answer. For example, in the case where the identity information represents that the user is a junior physician, the core conclusion can be retained and the complex mechanism can be omitted in the customized answer matched with the identity information; in the case where the identity information represents that the user is a senior physician, the customized answer matched with the identity information can be supplemented with literature references and analysis of controversial points.

[0063] In step S206, in the case where the timing teaching content generation event is detected, a user portrait of the user is generated according to the medical questions, medical answers and learning behavior data of the user on the instant messaging platform in a preset historical time period.

[0064] The timing teaching content generation event is used to trigger the generation of the user portrait and the pushing of the teaching content, and the preset historical time period can be 24 hours or 48 hours, which is not limited. For example, the timing teaching content generation event can be periodically detected by a timing task scheduler (such as performing portrait updating at 2:00 am every day), and all user portraits can be updated in full every 24 hours.

[0065] Optionally, the timing teaching content generation event can be triggered in advance when it is detected that the amount of daily newly added interaction data of the user exceeds a threshold (such as the number of medical questions asked is greater than a number threshold).

[0066] The user portrait can refer to a learning feature and knowledge state of the user described by a structured label system. For example, the user portrait can be a comprehensive description of the learning state and knowledge mastery of a student generated based on learning behavior data, test scores, interaction information and the like of the student. The user portrait can be used to evaluate the learning progress and weak links of the orthopedic training physician in real time, so as to improve the recommendation accuracy of the personalized teaching content.

[0067] For example, the user portrait can include the following dimensions: learning stage, knowledge mastery, learning preference, ability short board, and behavior pattern, wherein the learning stage can refer to the year of orthopedic training or the level of orthopedic physician; the knowledge mastery can refer to a dynamic score based on the question content and the answer accuracy rate (such as an orthopedic surgery step mastery score), the learning preference can refer to the preference identified according to the interaction behavior (such as a tendency of text materials or video tutorials), the ability short board can refer to weak knowledge points found through error question clustering analysis (such as fracture classification confusion points), and the behavior pattern can refer to time sequence characteristics such as active period, learning time, and interruption frequency.

[0068] For example, the learning behavior data can include, but is not limited to: user identity, question content, response record, knowledge calling frequency, message sending frequency, resource download record, simulation test participation times, message reading time, answer detail and sketch click difference, and multi-round dialogue backtracking times.

[0069] As an example, the terminal can establish a learning portrait engine based on rule matching and behavior analysis, and based on the learning portrait engine, generate a user portrait of the user according to the medical question, the medical answer, and the learning behavior data of the user on the instant messaging platform, to realize personalized teaching content recommendation.

[0070] For example, the rule matching can be based on knowledge point coverage rules, behavior trigger rules, and the like, wherein the knowledge point coverage rules can define orthopedic core knowledge points (such as fracture classification, surgery steps), and associate standard terms (such as CT terms), if the medical question asked by the user contains “Garden classification” and the medical answer involves “IV type surgery indications”, mark “femoral neck fracture classification mastery +1”; the behavior trigger rules can be used to define abnormal behavior patterns (such as high-frequency repeated questions about the same knowledge point), if the same knowledge point is asked more than 3 times a week, “knowledge weakness warning” can be triggered.

[0071] For another example, the behavior analysis can be to count the number of valid operations of the user within a time (such as the number of questions per hour), to construct an activity index, or to analyze the pause / playback position of the user's teaching video to identify learning difficulties.

[0072] As another example, the user portrait and the teaching content corresponding to the user portrait can be generated by using high-order recommendation algorithm models such as collaborative filtering, reinforcement learning, and graph neural network, so as to realize more complex teaching content scheduling and dynamic path generation functions, and improve the intelligent level of medical teaching.

[0073] Step S208, pushing the teaching content corresponding to the user portrait to the user through the chat robot.

[0074] Among them, the teaching content can include but is not limited to knowledge strengthening type content such as picture-text guide, 3D operation animation, case analysis document, virtual operation simulator link, operation specification video and the like skill training type content, adaptive test questions, clinical decision practice questions and the like evaluation feedback type content.

[0075] In a specific implementation, a teaching plan scheduling and content automatic pushing function can be introduced. The chat robot can automatically send personalized learning materials including teaching video links, case analysis questions, operation specification documents and the like teaching content through a scheduled task (using the apscheduler module in Python) every day. The pushed teaching content is generated based on the user portrait, which can match the learning needs of physicians at different stages. Optionally, user grouping can be determined by the label field in MySQL, that is, the user portrait can be identified by the label field defined in the database. The chat robot can execute the content generator at 0 o'clock every morning. The teaching content can be sent to the instant messaging platform of the target user in the form of rich text (picture-text + link), solving the problem of insufficient personalization in traditional teaching, and realizing learning path recommendation and dynamic content pushing for individuals. By continuously collecting user learning behavior data (such as question content, question frequency, knowledge point clicks, task completion time, etc.) on the WeChat end, a multi-dimensional learning portrait is constructed, and based on the reasoning ability of the large language model and the label classification mechanism, learning resources, simulation cases or evaluation tasks are intelligently matched for physicians at different stages of orthopedic training. This mechanism effectively breaks the low adaptability problem caused by the distribution of unified teaching content, and realizes individualized teaching and dynamic optimization.

[0076] For example, through the analysis of learning behavior data, the user portrait obtained can dynamically evaluate the learning state and weak links, automatically recommend subsequent learning content, simulation exercises or typical case resources, and realize the generation of a stage evaluation report, providing intelligent feedback support for teachers and students. This mechanism forms a teaching closed loop from "questioning-recommendation-feedback-optimization", which can improve the accuracy of teaching and the efficiency of resource matching. For example, in a real test scene in an orthopedic training base, a student user raises a medical question "What are the surgical methods for old humeral shaft fracture combined with malunion?", and the terminal can return complete answer content including treatment principles, surgical method selection, and perioperative precautions within less than 1.5 seconds through the chat robot, and also provide relevant chapter indexes in "Practical Orthopedics", significantly improving the semantic accuracy and logical completeness of medical answers.

[0077] In the related art, the teaching content pushing of the traditional intelligent teaching method has problems such as low degree of personalization, poor interactivity, and inability to automatically adjust learning strategies. For example, daily questions are pushed in a group form, lacking user state perception and interactive feedback mechanism, and unable to realize adaptive adjustment of teaching content and have multi-round dialogue capability. The method of the above embodiment can dynamically adjust the teaching strategy according to the orthopedic training physician's questioning behavior in the instant messaging platform, learning feedback, and usage frequency to push adaptive teaching content and support instant question answering through a medical teaching model, significantly improving teaching efficiency and participation.

[0078] In the above intelligent teaching method based on large language model and chat robot, the chat robot of the instant messaging platform obtains the medical question input by the user; the medical question is input into the medical teaching model integrated by the chat robot to obtain the medical answer corresponding to the medical question, wherein the medical teaching model is a large language model trained according to medical field corpus; a user portrait of the user is established based on the medical question, the medical answer, and the learning behavior data of the user in the instant messaging platform; and the chat robot pushes the teaching content corresponding to the user portrait. The medical large language model and the instant messaging platform are deeply integrated, the chat robot for the medical scene is developed, the all-weather, low-threshold, and semantic understanding driven teaching interaction and real-time question and answer feedback are realized; at the same time, through intelligent data analysis and learning behavior modeling, the learning path intelligent recommendation and personalized resource pushing based on the user portrait can be realized, breaking through the bottleneck of traditional rule-based robots in semantic understanding, professional adaptation, and interactive efficiency, and constructing a closed-loop intelligent teaching platform integrating real-time semantic question and answer and personalized knowledge pushing, which significantly improves the intelligent degree and effectiveness of medical teaching.

[0079] In another embodiment, the intelligent teaching method based on large language model and chat robot can be applied to a terminal where an intelligent teaching system is built. The intelligent teaching system adopts a technical architecture of "front-end communication module + back-end large language model + middle platform intelligent teaching control module", is designed especially for physician training, especially for the scene needs of orthopedic professional teaching, aims to break through the technical limitations of existing intelligent teaching systems in platform adaptability, professional semantic understanding, personalized pushing, and teaching feedback mechanism, and realizes a truly deployable, scalable, and optimized intelligent teaching solution.

[0080] The front-end communication module can be written in Python or JavaScript language and accessed to the instant messaging platform such as the WeChat platform by using the WeChat open source framework. The WeChat open source framework is an open source WeChat robot development framework that supports message transmission, user interaction and other operations through a programming interface with WeChat. The WeChat open source framework can be used for integration with the WeChat platform, enabling orthopedic training physicians to interact intelligently through WeChat for intelligent teaching, and supporting multi-language environments (such as Python, JavaScript), providing cross-platform WeChat robot development functions. The WeChat open source framework is a cross-platform WeChat automation interface framework that can stably obtain WeChat message events, process user input, and achieve two-way communication. The system registers a message event listener to achieve real-time message reception on the user side, and transmits the message content to the back-end large language model after structuring. Compared with traditional public number or enterprise WeChat interface, the WeChat open source framework is more suitable for independent deployment and rapid iteration when accessing the WeChat platform, and has higher flexibility and local control, especially in a non-open interface environment, it performs stably and adapts to the high-frequency interaction and low-latency requirements of the current teaching scenario. For example, the chat robot can be a WeChat robot based on a large language model, which relies on Python language development, combines the WeChat open source framework as middleware to access the WeChat platform, and realizes automatic message interaction by calling the WeChat Web API. In the specific development process, a system framework based on a modular architecture can be designed, including a natural language understanding module, a dialogue management module, a knowledge graph calling module, a teaching content generation module, and a user portrait analysis module. All modules are integrated into the WeChat dialogue interaction channel to realize intelligent management and real-time service response of the whole process of orthopedic training physician teaching.

[0081] The back-end large language model can be based on the ChatGPT, Gemini, Kimi architecture, and fine-tuned with medical special corpus to improve its semantic recognition and generation ability for orthopedic professional terms, disease path, operation specification, etc. The fine-tuning corpus comes from publicly published materials such as Practical Orthopedics and General Surgery, as well as teaching materials compiled by the project team, supplemented by PubMed abstracts in the field of orthopedics as English academic support. Using a lightweight parameter adjustment strategy (such as LoRA), the model significantly improves its understanding of Chinese medical context while considering resource consumption. The model uses FastAPI to build a service interface that supports concurrent processing and asynchronous response, ensuring the stability and response efficiency of the system in a high-concurrency question scenario.

[0082] The middle platform intelligent teaching control module can be used to undertake the management and data logic processing of the teaching process, is the core module of the intelligent teaching system, is responsible for managing the allocation of teaching tasks, collecting learning data, generating personalized recommendations, and evaluating teaching effectiveness, integrates large language models, knowledge graphs, and learning behavior analysis technologies, ensures the intelligentization and personalization of teaching content and services, and adjusts teaching strategies in real time. The intelligent teaching system can design a structured teaching behavior database, store user identity, question content, response records, knowledge call frequency, and other key learning behavior data using MySQL, and establish a learning profile engine based on rule matching and behavior analysis. Through the analysis of user interaction data, the system can dynamically evaluate the learning state and weak links of the user, automatically recommend subsequent learning content, simulated exercises, or typical case resources, and generate a stage evaluation report, providing intelligent feedback support for teachers and students. This mechanism forms a teaching closed loop from "questioning-recommendation-feedback-optimization", which is the core of improving teaching accuracy and resource matching efficiency. In summary, the intelligent teaching system is based on Python and the WeChat open source framework as the core architecture, through model fine-tuning, semantic analysis, multi-dimensional data feedback, and deep integration of the original interaction capabilities of the WeChat platform, an intelligent teaching system suitable for orthopedic teaching actual scenarios is constructed, realizing instant access to medical knowledge, personalized recommendation of learning paths, and whole-process closed-loop management of teaching data, effectively solving the technical pain points of current orthopedic teaching in "low efficiency", "incoherence", and "no feedback", and having practical landing promotion value and expansion potential.

[0083] For example, using Python as the programming language, implementing WeChat access based on the WeChat open source framework, and calling a large language model for teaching question and answer interaction, the intelligent teaching system can be deployed in an Ubuntu 20.04 server environment, using open source Nginx for reverse proxy, Gunicorn as a Python application service container, and MySQL database for data storage and management of user question records, interaction logs, and feedback tags.

[0084] This embodiment demonstrates the basic process of WeChat native interface, large language model calling, and teaching data structured storage, which is suitable for rapid prototyping and small-scale orthopedic training scene testing of the intelligent teaching system. Subsequent user profile analysis and teaching resource matching modules can be added based on this. Through the modular and extensible Python development structure, combined with the stable WeChat access capability realized by the WeChat open source framework, and the integration of key technologies such as large model question and answer, knowledge graph reasoning, and multi-dimensional user profile, a highly automated, interactive, and scenario-adaptive intelligent teaching system is constructed, which can significantly improve the efficiency and experience of clinical teaching, fully meeting the needs of personalized and efficient orthopedic teaching.

[0085] In another embodiment, the learning behavior data includes at least one of a frequency of user input medical questions, a learning frequency of the user on the teaching content, and a learning feedback result of the user on the teaching content.

[0086] The frequency of input medical questions can include a number of medical questions submitted by the user to the chat robot per unit of time (e.g., daily / weekly); the learning frequency of the teaching content can include a number of times and interval times of the user accessing the teaching resources (text, video, test questions); and the learning feedback result of the user on the teaching content can include an explicit score (e.g., 1-5 stars) and an implicit behavior (e.g., video viewing completion, test question accuracy) of the user on the pushed teaching content.

[0087] The technical solution of the above embodiment breaks through the limitation of a single dimension through multi-dimensional behavior fusion modeling, improves the accuracy of user portrait description, and enhances the adaptability of the portrait according to the learning behavior data of the user, thereby improving the accuracy of intelligent teaching.

[0088] In another embodiment, the chat robot accesses the instant messaging platform through an open source framework; the chat robot of the instant messaging platform obtains the medical question input by the user, including: in the case that the message listening interface constructed through the open source framework listens to the message content sent by the user to the chat robot of the instant messaging platform, the message content is structured to obtain the medical question input by the user.

[0089] The open source framework can refer to a robot development framework supporting cross-platform access, providing standardized interfaces to realize message listening, protocol adaptation, session management, etc. Optionally, the open source framework can be determined according to the development language and the deployment platform, for example, the open source framework can be an automatic interface library of the instant messaging platform developed based on the Python language.

[0090] The message listening interface is a protocol adaptation middleware constructed based on the open source framework, supports multiple communication protocols, and can realize the following functions: real-time capture of original message content (text, picture, voice, file) sent by the user, and unified conversion of message formats of different platforms into a standard data structure.

[0091] The structured processing of the message content through the message listening interface can be the conversion of the original message content (including non-text data) into a medical question semantic representation understandable by the machine, as a medical question. For example, for text message content, medical entities (disease, symptom, surgery) and intent labels (consultation, test, learning) can be extracted as medical questions; for picture / voice message content, structured text can be generated through OCR (optical character recognition) and ASR (speech-to-text) as medical questions.

[0092] The technical solutions of the above embodiments greatly improve the interactive response efficiency and platform access convenience. Through the native integration of the chat robot and the instant messaging platform, the dependence on high threshold interfaces is avoided, and the deployment and use complexity is reduced. Users can directly interact with the system through a common instant messaging platform account without the need for additional application installation or Web platform login, greatly improving the participation rate of teaching activities and the actual use rate of the system, and meeting the real-time interaction needs of mobile end fragmented learning scenarios.

[0093] In another embodiment, a medical question is input into a chat robot integrated medical teaching model to obtain a medical answer corresponding to the medical question, including: identifying a medical keyword in the medical question, and querying a corresponding standard answer in a pre-built medical knowledge graph according to the medical keyword; inputting the standard answer, the medical question and the user's identity information into the chat robot integrated medical teaching model, and generating a customized answer matching the identity information based on the standard answer through the medical teaching model as the medical answer corresponding to the medical question.

[0094] The knowledge graph is a way of representing knowledge through a graph structure, where nodes represent entities (such as people, things, concepts, etc.), and edges represent the relationships between entities. The pre-built medical knowledge graph is used to organize and store knowledge in the medical field, and through a graph database for efficient querying and reasoning, it helps large language models for semantic analysis and content generation, improving the model's understanding of medical professional problems.

[0095] In a specific implementation, to improve semantic understanding in complex teaching scenarios, entity recognition and intent classification based on the medical knowledge graph can be introduced. Optionally, the pre-built medical knowledge graph can be based on Neo4j as the core database, and common diseases, surgical steps and clinical pathways involved in orthopedic physician training can be pre-built. After identifying the user's medical question, the corresponding node can be queried through SPARQL to obtain a standard answer, and then a large language model can be linked to generate a more targeted customized answer as a medical question and answer.

[0096] For example, the user's input medical question is "How to handle the femoral comminuted fracture found during surgery?", the medical keyword "femoral comminuted fracture" can be identified, the corresponding node in the pre-built medical knowledge graph can be matched to obtain a standard answer, and the medical answer "recommend intramedullary nail fixation or artificial joint replacement" can be returned according to the standard treatment path contained in the standard answer, and operation videos can be pushed according to the user's identity information.

[0097] The user's identity information can include but is not limited to the user's learning stage, the user's title information, the user's work experience information, the user's position level information, etc.

[0098] The above embodiments significantly improve the semantic analysis capability and professional adaptability of the teaching system by knowledge graph assisted decision-making and customized question and answer strategy, and improve the accuracy and efficiency of question and answer in intelligent teaching. By deeply combining a large language model with orthopedic exclusive medical corpus (including orthopedic teaching materials, clinical guidelines, typical case sets, etc.), cooperating with a local knowledge graph entity relationship reasoning mechanism, complex medical questions raised by orthopedic training doctors can be accurately identified and answered with clear semantic levels and accurate professional terms. The problem of one-sided content and understanding deviation of existing robot answers is effectively solved, and technical support is provided for high-quality question and answer interaction in a teaching context.

[0099] In another embodiment, before the step of generating the user portrait of the user according to the medical questions, medical answers and learning behavior data of the user on the instant messaging platform in the preset historical time period, it further includes: calling the medical questions, medical answers and learning behavior data of the user on the instant messaging platform in the preset historical time period from the database.

[0100] Optionally, the database can be a mainstream relational or document type database such as MySQL, PostgreSQL, MongoDB, SQLite, etc. to adapt to different project sizes, concurrent demands and maintenance strategies. The database can be used for structured storage of learning behavior data of the user and teaching resources.

[0101] The database can be used for structured storage of medical questions, medical answers and learning behavior data of the user on the instant messaging platform, realizing a structured data collection and evaluation analysis closed loop of the whole teaching process. For example, medical questions can be stored in a questions table (fields: user_id, question_text, timestamp) in the database, medical answers can be stored in an answers table (fields: answer_id, question_id, content), and learning behavior data can be stored in a behavior_logs table (fields: user_id, action_type, resource_id, duration). Optionally, a behavior record module can be embedded in teaching content pushing, interactive question and answer, evaluation test, etc. All interaction data enters the database for structured management, and visual teaching analysis reports can be generated for teachers, teaching research groups and teaching management platforms, providing objective data basis for subsequent teaching decision-making, student evaluation and resource optimization.

[0102] The technical solutions of the above embodiments propose a portrait dynamic updating mechanism based on timing event triggering, realize precise adaptation of teaching strategies, and through structured storage of key data in the database, help to realize dynamic generation and precise optimization of the user portrait, improve data utilization efficiency and real-time performance of the portrait, and thus improve the accuracy and efficiency of intelligent teaching.

[0103] In another embodiment, the teaching content corresponding to the user portrait is pushed to the user through the chat robot, including: determining a learning stage matched with the user according to the user portrait; generating the teaching content according to the graphic data and the link data corresponding to the learning stage; and pushing the teaching content to the user through the chat robot.

[0104] For example, if the user's query frequency of basic content is higher than a certain threshold, or the user's identity is displayed as a resident physician, the learning stage can be determined as primary, and the corresponding teaching content such as basic theory reinforcement and standard operation process learning can be pushed; if the user's query frequency of advanced content is higher than a certain threshold, or the user's query frequency of basic content is less than a certain threshold, or the user's identity is displayed as a middle-level physician, the learning stage can be determined as middle-level, and the corresponding teaching content such as case analysis and differential diagnosis ability improvement learning can be pushed.

[0105] The graphic data can be knowledge points extracted from a teaching resource library that meet the current stage, and an interactive summary (such as a folding graphic card) is generated through a large language model. The link data can be links adapted based on the user device type (mobile phone / PC), such as HTML interactive pages that are preferentially pushed to mobile terminals.

[0106] The technical solutions of the above embodiments propose a learning stage adaptive matching mechanism, dynamically divide the user learning stage, and generate adaptive teaching content, thereby realizing precise alignment of teaching resources and user ability.

[0107] In summary, the above-mentioned intelligent teaching method based on large language model and chat robot is summarized as follows: by integrating large language model, instant platform communication framework, knowledge graph auxiliary module and personalized learning path generation mechanism, an intelligent teaching system with clear structure, efficient response, strong semantic understanding ability and high application landing is constructed. Compared with existing teaching technology, the system realizes multi-dimensional teaching efficiency improvement and interaction quality optimization in the scene of physician training, and produces significant technical effect. In view of the problems existing in the training of orthopedic physicians, such as uneven distribution of teaching resources, low teaching efficiency, lack of personalized guidance and learning feedback mechanism, a closed-loop intelligent teaching platform integrating "knowledge pushing-semantic question answering-learning recommendation-process evaluation-review feedback" is constructed. Compared with the prior art, the significant innovation of the present scheme lies in the deep integration of medical large language model and instant messaging platform, the development of chat robot for orthopedic training scene, and the realization of all-weather, low threshold and semantic understanding driven teaching interaction. At the same time, through intelligent data analysis and learning behavior modeling, the intelligent recommendation of learning path, personalized resource pushing, real-time question and answer feedback and multi-dimensional teaching effect evaluation can be realized, which breaks through the bottleneck of traditional rule-based robots in semantic understanding, professional adaptation and interaction efficiency, significantly improves the intelligence and effectiveness of orthopedic teaching. Not only does it break through the technical bottlenecks of existing technologies in semantic analysis, platform adaptation, interaction efficiency, content pushing and teaching closed loop, but also achieves significant improvement in teaching efficiency, strengthens the ability of students, and improves the quality of teaching feedback in actual orthopedic teaching, which has obvious practicality and generalizability.

[0108] In another embodiment, as shown in Figure 3 , an intelligent teaching method based on large language model and chat robot is provided. Taking the terminal 102 in Figure 1 as an example, the method includes the following steps:

[0109] Step S302, when the user sends a message content to the chat robot of the instant messaging platform is listened to by the message listening interface built by the open source framework, the message content is structured and processed to obtain the medical problem input by the user.

[0110] Among them, the chat robot accesses the instant messaging platform through the open source framework.

[0111] Step S304, identifying the medical keywords in the medical question, and querying the corresponding standard answer in the pre-constructed medical knowledge graph according to the medical keywords.

[0112] Step S306, inputting the standard answer, medical question and user's identity information into the medical teaching model integrated by the chat robot, generating a customized answer matched with the identity information based on the standard answer through the medical teaching model, as the medical answer corresponding to the medical question.

[0113] Step S308, in the case of detecting the timing teaching content generation event, generating a user portrait of the user according to the medical question, the medical answer and the learning behavior data of the user on the instant messaging platform within a preset historical time period.

[0114] In one of the embodiments, before the step of generating the user portrait of the user according to the medical question, the medical answer and the learning behavior data of the user on the instant messaging platform within a preset historical time period, the method further comprises: calling the medical question, the medical answer and the learning behavior data of the user on the instant messaging platform within a preset historical time period from a database.

[0115] In one of the embodiments, the learning behavior data comprises at least one of the frequency of inputting the medical question by the user, the learning frequency of the teaching content by the user and the learning feedback result of the teaching content by the user.

[0116] Step S310, pushing the teaching content corresponding to the user portrait to the user through the chat robot.

[0117] In one of the embodiments, the pushing of the teaching content corresponding to the user portrait through the chat robot can comprise: determining a learning stage matched with the user according to the user portrait; generating the teaching content according to the graphic data and the link data corresponding to the learning stage; and pushing the teaching content to the user through the chat robot.

[0118] It should be noted that the specific limitations of the above steps can refer to the specific limitations of the above-mentioned intelligent teaching method based on a large language model and a chat robot.

[0119] It should be understood that, although each step in the flowchart involved in each embodiment as described above is displayed in sequence according to the arrow, these steps are not necessarily executed in sequence according to the arrow. Unless otherwise specified herein, the execution of these steps has no strict sequence limitation, and these steps can be executed in other orders. Moreover, at least part of the steps in the flowchart involved in each embodiment as described above can include multiple steps or stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence of these steps or stages is not necessarily sequential, but can be executed in rotation or alternation with at least part of other steps or steps or stages in other steps.

[0120] Based on the same inventive concept, the present application also provides a large language model and chat robot-based intelligent teaching device for implementing the above-mentioned large language model and chat robot-based intelligent teaching method. The problem-solving implementation scheme provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more large language model and chat robot-based intelligent teaching device embodiments provided below can refer to the limitations of the large language model and chat robot-based intelligent teaching method described above, which will not be repeated here.

[0121] In one exemplary embodiment, as shown in Figure 4 a large language model and chat robot-based intelligent teaching device is provided, comprising:

[0122] The acquisition module 410 is configured to acquire a medical question input by a user through a chat robot of an instant messaging platform.

[0123] The answering module 420 is configured to input the medical question into a medical teaching model integrated with the chat robot to obtain a medical answer corresponding to the medical question, and output the medical answer through the chat robot; the medical teaching model is a large language model trained according to a medical field corpus.

[0124] The generation module 430 is configured to, in the case of detecting a timing teaching content generation event, generate a user portrait of the user according to the medical question, the medical answer, and learning behavior data of the user on the instant messaging platform within a preset historical time period.

[0125] The pushing module 440 is configured to push teaching content corresponding to the user portrait to the user through the chat robot.

[0126] In one embodiment, the answering module 420 is specifically configured to identify a medical keyword in the medical question, and query a corresponding standard answer in a pre-constructed medical knowledge graph according to the medical keyword; input the standard answer, the medical question, and identity information of the user into the medical teaching model integrated with the chat robot, generate a customized answer matching the identity information based on the standard answer through the medical teaching model, and use the customized answer as the medical answer corresponding to the medical question.

[0127] In one embodiment, the generation module 430 is specifically configured to call the medical question, the medical answer, and the learning behavior data of the user on the instant messaging platform within a preset historical time period from a database.

[0128] In one of the embodiments, the pushing module 440 is specifically configured to determine a learning stage matched with the user according to the user portrait, generate the teaching content according to the corresponding image-text data and link data of the learning stage, and push the teaching content to the user through the chat robot.

[0129] In one of the embodiments, the learning behavior data includes at least one of the frequency of the user inputting medical questions, the learning frequency of the user on the teaching content, and the learning feedback result of the user on the teaching content.

[0130] In one of the embodiments, the obtaining module 410 is specifically configured to, in the case of listening to the user sending message content to the chat robot of the instant messaging platform through the message listening interface built by the open source framework, perform structural processing on the message content to obtain the medical question input by the user.

[0131] The above various modules in the intelligent teaching device based on the large language model and the chat robot can be all or partially realized by software, hardware, and combinations thereof. The above various modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory in the computer device in software form, so as to be called and executed by the processor to perform the operations corresponding to the above various modules.

[0132] In an exemplary embodiment, a computer device, which can be a terminal, is provided, and an internal structure diagram of the computer device can be as shown in Figure 5As shown in the figure. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. Among them, the processor, the memory and the input / output interface are connected through the system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capability. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be realized through WIFI, mobile cellular network, near field communication (NFC) or other technologies. The computer program is executed by the processor to realize an intelligent teaching method based on a large language model and a chat robot. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0133] Those skilled in the art can understand that, Figure 5 The skilled in the art can understand that,

[0134] In one exemplary embodiment, a computer device is provided, including a memory and a processor, the memory stores a computer program, and the processor executes the computer program to realize the steps in each of the above method embodiments.

[0135] In one embodiment, a computer readable storage medium is provided, which stores a computer program, and the computer program is executed by a processor to realize the steps in each of the above method embodiments.

[0136] In one embodiment, a computer program product is provided, including a computer program, and the computer program is executed by a processor to realize the steps in each of the above method embodiments.

[0137] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant regulations.

[0138] It can be understood by those skilled in the art that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing related hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments of each method. In the embodiments provided in the present application, any reference to memory, database or other medium can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., without being limited thereto.

[0139] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0140] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. An intelligent teaching method based on a large language model and a chat robot, characterized in that: The method comprises: Obtaining medical questions inputted by users through chatbots on instant messaging platforms; The medical question is input into the medical teaching model integrated with the chatbot to obtain a medical answer corresponding to the medical question, and the chatbot outputs the medical answer; the medical teaching model is a large language model trained based on medical field corpus; When a scheduled teaching content generation event is detected, a user profile of the user is generated based on the medical questions, the medical answers, and the user's learning behavior data on the instant messaging platform within a preset historical time period; The teaching content corresponding to the user portrait is pushed to the user through the chat robot.

2. The method according to claim 1, characterized in that Inputting the medical question into the medical teaching model integrated with the chatbot to obtain a medical answer corresponding to the medical question includes: Identify medical keywords in the medical question and query corresponding standard answers in a pre-built medical knowledge graph based on the medical keywords; The standard answer, the medical question and the user's identity information are input into the medical teaching model integrated with the chatbot, and the medical teaching model generates a customized answer matching the identity information based on the standard answer as the medical answer corresponding to the medical question.

3. The method according to claim 1, characterized in that Before the step of generating a user profile of the user based on the medical questions, the medical answers, and the user's learning behavior data on the instant messaging platform within a preset historical time period, the method further includes: The medical questions, the medical answers and the user's learning behavior data on the instant messaging platform within a preset historical time period are retrieved from a database.

4. The method according to claim 1, wherein The pushing of teaching content corresponding to the user portrait to the user by the chat robot includes: Determining a learning stage that matches the user based on the user profile; Generating the teaching content according to the graphic data and link data corresponding to the learning stage; The teaching content is pushed to the user through the chat robot.

5. The method according to any one of claims 1 to 4, characterized in that The learning behavior data includes at least one of the frequency of the user inputting medical questions, the frequency of the user learning the teaching content, and the learning feedback result of the user on the teaching content.

6. The method according to claim 1, characterized in that The chatbot is connected to the instant messaging platform through an open source framework; the chatbot obtains the medical questions input by the user through the instant messaging platform, including: When the message monitoring interface constructed by the open source framework monitors the message content sent by the user to the chat robot of the instant messaging platform, the message content is structured to obtain the medical problem input by the user.

7. An intelligent teaching device based on a large language model and a chat robot, characterized in that: The device comprises: An acquisition module, used to acquire medical questions input by users through a chatbot on an instant messaging platform; an answer module, configured to input the medical question into the medical teaching model integrated with the chatbot, obtain a medical answer corresponding to the medical question, and output the medical answer through the chatbot; the medical teaching model is a large language model trained based on medical field corpus; a generating module configured to generate a user profile of the user based on the medical questions, the medical answers, and the user's learning behavior data on the instant messaging platform within a preset historical time period when a scheduled teaching content generation event is detected; A push module is used to push the teaching content corresponding to the user portrait to the user through the chat robot.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images