Personalized teaching methods, devices, computer equipment, and software products
By acquiring multimodal interaction data and using network models for analysis, key points in the learning process are identified, and personalized teaching content is generated. This solves the problem of fixed teaching content in existing technologies, enables dynamic adjustment and optimization of learning paths, and improves teaching efficiency and user experience.
Patent Information
- Application Number
- CN202610320363.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-03-16
- Publication Date
- 2026-11-17
AI Technical Summary
Existing online education technologies cannot achieve personalized learning. There is a significant gap between teaching content and students' learning abilities, which makes it impossible to dynamically adjust the teaching content based on students' real-time feedback, resulting in ineffective and repetitive learning.
By acquiring multimodal interaction data of the learners, the system analyzes the state change information of the mastery sequence using a target network model, identifies key sequence points, and generates personalized updated teaching content, including content explanation resources, example demonstration resources, and exercise resources. The system also uses a digital human rendering engine to generate teaching video streams.
It enables dynamic monitoring and temporal feature extraction of the learning process, accurately locates key points in the learning path, generates customized teaching resources, optimizes the learning path, improves teaching efficiency and user experience, and realizes online personalized education.
Smart Images

Figure CN123115508A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of computer technology, and in particular to a personalized teaching method, apparatus, computer equipment, computer-readable storage medium, and computer program product. Background Technology
[0002] With the development of computer technology, digital education technology has emerged. Digital online education conducts teaching online via the internet, and can record and store a large amount of student class information. Furthermore, students can start learning anytime, anywhere via the internet.
[0003] However, current online education typically uses standardized teaching content to instruct a group of students. After class, students also complete standardized assignments. This fails to achieve personalized learning, resulting in a significant gap between the teaching content and students' learning abilities. Summary of the Invention
[0004] Therefore, it is necessary to address the aforementioned technical problems by providing a personalized teaching method, device, computer equipment, computer-readable storage medium, and computer program product that can improve the fit between teaching content and the learning ability of students, realize personalized teaching, and improve the quality of teaching content.
[0005] Firstly, this application provides a personalized teaching method, including:
[0006] Acquire multimodal interaction data of learners regarding the teaching content; the teaching content includes at least one teaching content node;
[0007] Based on multimodal interactive data, the mastery level of each teaching content node is obtained, and based on the mastery level of each node, a mastery level sequence corresponding to the teaching content is obtained.
[0008] The teaching content and mastery sequence are input into the target network model. The target network model analyzes the state change information of the mastery sequence, identifies key sequence points in the mastery sequence based on the state change information, and generates updated teaching content based on the key sequence points.
[0009] The updated teaching content will be sent to the target terminal corresponding to the teaching object so that the updated teaching content can be displayed on the target terminal.
[0010] In one embodiment, the multimodal interaction data includes node interaction data corresponding to each teaching content node. The system analyzes the state change information of the mastery level sequence through a target network model, and identifies key sequence points in the mastery level sequence based on the state change information. This includes: analyzing the state change information corresponding to the mastery level sequence through a temporal feature extraction layer in the target network model; identifying target change sequence segments in the state change information; for each teaching content node corresponding to the target change sequence segment, fusing the node interaction data corresponding to the teaching content node through a multimodal fusion layer in the target network model to obtain a modal conflict reference value corresponding to the teaching content node; determining interactive conflict nodes in the teaching content based on the target modal conflict reference value; ensuring the target modal conflict reference value is greater than a preset modal conflict reference value; acquiring a knowledge graph corresponding to the teaching content; determining the preceding content nodes associated with the interactive conflict nodes based on the knowledge graph; and determining key sequence points in the mastery level sequence based on the preceding content nodes.
[0011] In one embodiment, the mastery level sequence is analyzed by a temporal feature extraction layer in the target network model to identify the state change information corresponding to the mastery level sequence; the target change sequence segment in the state change information is identified, including: converting the mastery level sequence into a feature vector sequence, inputting the feature vector sequence into the temporal feature extraction layer, performing correlation analysis on each vector in the feature vector sequence through the temporal feature extraction layer to obtain temporal state features, which represent the state change information of each teaching content node in the time dimension; calculating the state change gradient between adjacent teaching content nodes based on the temporal state features; determining the target change sequence segment based on the target state change gradient; and ensuring that the target state change gradient is greater than a preset gradient threshold.
[0012] In one embodiment, generating updated teaching content based on key sequence points includes: obtaining knowledge point attributes of key content nodes corresponding to key sequence points; knowledge point attributes include at least conceptual understanding, logical deduction, and computational skills; obtaining key features corresponding to key sequence points based on mastery level sequences; querying a preset teaching strategy mapping table based on knowledge point attributes and key features to obtain a target teaching strategy; querying a teaching content knowledge base corresponding to teaching knowledge points that match the teaching content; the teaching content knowledge base stores at least one type of teaching content corresponding to each teaching knowledge point; and obtaining updated teaching content from the teaching content knowledge base based on the target teaching strategy and knowledge point attributes.
[0013] In one embodiment, updating the teaching content includes content explanation resources, example demonstration resources, and exercise resources; generating updated teaching content based on key sequence points includes: determining the target teaching stage based on the key mastery level corresponding to the key sequence points, wherein the target teaching stage is any one of the content explanation stage, example demonstration stage, or exercise practice stage; when the target teaching stage is the content explanation stage, generating explanation content containing the knowledge point breakdown logic; when the target teaching stage is the example demonstration stage, retrieving target example questions from a preset question bank that match the knowledge point attributes corresponding to the key sequence points, and generating demonstration content of the solution steps for the target example questions; when the target teaching stage is the exercise practice stage, determining the target difficulty level based on the key mastery level, and generating exercise questions that meet the target difficulty level.
[0014] In one embodiment, the updated teaching content is sent to the target terminal corresponding to the learner, so that the updated teaching content can be displayed on the target terminal. This includes: calling a preset digital human rendering engine to obtain target digital human driving data that matches the type of the updated teaching content; the driving data includes facial expression data, body movement data, and voice broadcast data; fusing and rendering the updated teaching content with the target digital human driving data to generate a digital human teaching video stream; and sending the digital human teaching video stream to the target terminal so that the target terminal can perform interactive questioning actions with the learner through the digital human image while playing the updated teaching content.
[0015] In one embodiment, the multimodal interaction data includes at least one of screen touch data, voice response data, and facial image data.
[0016] Secondly, this application also provides a personalized teaching device, including:
[0017] The acquisition module is used to acquire multimodal interaction data of the learners in response to the teaching content; the teaching content includes at least one teaching content node;
[0018] The sequence module is used to obtain the mastery level of each teaching content node based on multimodal interactive data, and to obtain the mastery level sequence corresponding to the teaching content based on the mastery level of each node.
[0019] The processing module is used to input the teaching content and mastery level sequence into the target network model, analyze the state change information of the mastery level sequence through the target network model, identify key sequence points in the mastery level sequence based on the state change information, and generate updated teaching content based on the key sequence points.
[0020] The communication module is used to send updated teaching content to the target terminal corresponding to the teaching object so that the updated teaching content can be displayed on the target terminal.
[0021] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.
[0022] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the above-described method embodiments.
[0023] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.
[0024] The aforementioned personalized teaching methods, devices, computer equipment, computer-readable storage media, and computer program products, by acquiring multimodal interactive data encompassing multiple dimensions such as facial expressions, voice, and actions, can overcome the shortcomings of traditional teaching assessments, which rely solely on single-response data or simple behavioral logs, resulting in a singular assessment dimension and one-sided results. The introduction of multimodal data can capture subtle changes in the learner's emotional cognition and behavioral operations, thereby more realistically and objectively quantifying their mastery of each teaching content node. Secondly, by constructing a mastery degree sequence and using a target network model to analyze its state change information, dynamic monitoring of the learning process and extraction of temporal features can be achieved. By identifying the fluctuation trend and abrupt change characteristics of mastery degree over time, key sequence points in the learning path (such as knowledge gaps, attention inflection points, etc.) can be accurately located. In addition, by generating and pushing updated teaching content based on key sequence points, dynamic closed-loop optimization of teaching content is achieved. Customized teaching resources are generated in real time for identified weak links or key turning points, which solves the problem of fixed teaching content and inability to dynamically adjust according to students' real-time feedback in existing online education technologies. This reduces ineffective and repetitive learning, greatly optimizes the learning path, improves overall teaching efficiency and user experience, and realizes online personalized teaching. Attached Figure Description
[0025] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0026] Figure 1 This is a diagram illustrating the application environment of a personalized teaching method in one embodiment.
[0027] Figure 2 This is a flowchart illustrating a personalized teaching method in one embodiment;
[0028] Figure 3 This is a structural block diagram of a personalized teaching device in one embodiment;
[0029] Figure 4 This is a flowchart illustrating a personalized teaching method in another embodiment;
[0030] Figure 5 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation
[0031] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0032] The personalized teaching method provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or placed on a cloud or other network server. The server acquires multimodal interactive data of the teaching subjects regarding the teaching content; the teaching content includes at least one teaching content node; based on the multimodal interactive data, it obtains the node mastery level corresponding to each teaching content node, and based on the node mastery level, it obtains the mastery level sequence corresponding to the teaching content; it inputs the teaching content and the mastery level sequence into a target network model, analyzes the state change information of the mastery level sequence through the target network model, identifies key sequence points in the mastery level sequence based on the state change information, and generates updated teaching content based on the key sequence points; it sends the updated teaching content to the terminal corresponding to the teaching subject for display on the terminal. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can be smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, projection devices, etc. Portable wearable devices can be smartwatches, smart bracelets, head-mounted devices, etc. Headset devices can be virtual reality (VR) devices, augmented reality (AR) devices, smart glasses, etc. Server 104 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.
[0033] In one exemplary embodiment, such as Figure 2 As shown, a personalized teaching method is provided, which can be applied to... Figure 1 Taking the server in the example, the explanation includes the following steps 202 to 208. Wherein:
[0034] Step 202: Obtain multimodal interaction data of the learners regarding the teaching content.
[0035] The teaching content includes at least one teaching content node.
[0036] In this context, "teaching object" refers to the subject learning in an intelligent education scenario, typically referring to students, but can also refer to trainees. "Teaching content" refers to the specific set of knowledge or course units that the teaching object is learning. It has a hierarchical structure and can be a course or a chapter. "Multimodal interactive data" refers to the set of data collected through multiple sensory modalities during the teaching process that reflects the behavior and state of the teaching object. Multimodal interactive data includes at least one of the following: screen touch data (such as mouse clicks, touchscreen swipes, key input, etc.), voice response data (such as tone of voice, voice when answering questions), and facial image data (such as facial expressions, eye movements, body movements). "Teaching content node" refers to the smallest logical unit or knowledge point into which the teaching content is divided. For example, in a math lesson on the Pythagorean theorem, "definition of a right triangle," "Pythagorean theorem formula," and "application of the formula" can all be considered independent teaching content nodes. It can be understood that if the content the teaching object is learning is identified as a teaching content node, then the teaching content is the content to which that teaching content node belongs, such as the corresponding chapter or the parent node of that teaching content node.
[0037] For example, the server can acquire multimodal interactive data based on the audio and video capture devices of the teaching terminal. Using cameras and microphones installed on teaching terminal devices (such as smart blackboards, tablets, and learning machines), video and audio streams of the learners are captured in real time. Image processing technology is used to recognize facial expressions (such as confusion or focus), head posture, and hand gestures; audio processing technology is used to recognize speech content and emotional characteristics, thereby forming raw multimodal interactive data containing visual and auditory information. For example, during teaching, the camera captures images of students' faces at a frequency of 30 frames per second, and the microphone simultaneously records students' mutterings or answers to questions. When a student views the node "definition of a right triangle," the system records the image data of their "frowning" facial expression and the speech data of "I don't quite understand this definition," which constitutes the multimodal interactive data for that node.
[0038] Optionally, embedded code can be pre-set in the terminal interface (such as an app or webpage) displaying the teaching content. When the learner performs actions such as clicking, swiping, dragging, inputting text, or drawing, the teaching terminal automatically records the timestamp, location coordinates, operation type, and corresponding page element ID. Additionally, pressure-sensing data from the touchscreen can be combined to generate multimodal interactive data based on user behavior. For example, if a student is conducting a physics experiment simulation on a tablet, the system can use embedded code to record that the student stayed on the "Gravity Acceleration" node page for 5 minutes, repeatedly dragged the slider within the simulation area to adjust parameters, and made a larger drag on the last drag. These click events, swipe trajectories, and pressure values constitute the multimodal interactive data for that node.
[0039] Step 204: Based on the multimodal interactive data, obtain the node mastery level corresponding to each teaching content node, and based on the node mastery level, obtain the mastery level sequence corresponding to the teaching content.
[0040] Among them, the level of mastery of a node refers to a quantitative indicator of the learner's understanding and mastery of a specific teaching content node (knowledge point). This can be represented by numerical values (e.g., 0.8), grades (e.g., "Excellent," "Good," "Poor"), or probability values. The mastery sequence refers to a data vector arranged according to the logical or chronological order of the teaching content nodes, representing the level of mastery of each node. This sequence not only includes the mastery status at each point in time but also implicitly reflects the evolution of the learning state over time or as knowledge progresses.
[0041] For example, to obtain the mastery level sequence, various features (such as answer accuracy, dwell time, and facial confusion) are first extracted from the multimodal interaction data, and different weights are assigned to each feature. Then, for each teaching content node, a comprehensive score is calculated using a weighted summation formula, which serves as the node's mastery level. Finally, these scores are concatenated to form a sequence according to the logical order of the teaching content. For example, for the node "Pythagorean theorem formula," the extracted data shows that the student's answer accuracy is 0.9 (weight 0.6), the percentage of focused facial expressions while watching the video is 0.8 (weight 0.2), and the number of words entered in the notes is 50 (weight 0.2). The mastery level of this node is calculated to be 0.9 × 0.6 + 0.8 × 0.2 + 0.5 × 0.2 = 0.82. Similarly, the scores for other nodes are calculated to form a mastery level sequence such as [0.6, 0.82, 0.7, 0.9].
[0042] Optionally, multimodal interaction data can be input into a pre-trained deep learning model (such as a Convolutional Neural Network (CNN) or a Long Short-Term Memory (LSTM) network). After training with a large number of samples, the model can establish a non-linear mapping between multimodal features and mastery levels, directly outputting predicted values of node mastery. Subsequently, the system arranges the predicted values of each node in order to generate a sequence. For example, after obtaining multimodal interaction data, the server inputs the video frame sequence and mouse trajectory map of the student learning the "photosynthesis" node into a trained BERT multimodal model. After analysis, the model outputs a probability value of 0.75, indicating that the student has a 75% probability of mastering the knowledge point. All nodes are processed sequentially to generate a sequence vector containing multiple probability values.
[0043] Step 206: Input the teaching content and mastery level sequence into the target network model, analyze the state change information of the mastery level sequence through the target network model, identify key sequence points in the mastery level sequence based on the state change information, and generate updated teaching content based on the key sequence points.
[0044] The target network model refers to the artificial intelligence algorithm model deployed on the server side, used to process sequence data, mine patterns of state changes, and generate decision suggestions. It can be a recurrent neural network, a Transformer model, a reinforcement learning model, or other time-series prediction models. State change information refers to the fluctuation characteristics of the values in the mastery sequence as the sequence progresses, such as trends like rising, falling, sudden changes, and oscillations, used to reflect the dynamic evolution of learning effectiveness. Key sequence points refer to nodes in the mastery sequence that have significant state change characteristics or a major impact on the learning path. For example, nodes where mastery drops sharply (knowledge gaps) or nodes that stagnate for a long time (learning bottlenecks). Updated teaching content refers to remedial teaching materials dynamically generated to address the problems reflected by the key sequence points. These can take the form of micro-lecture videos, targeted exercises, graphic explanations, or prompts, aiming to solve specific problems encountered by learners during the learning process.
[0045] For example, the target network model can calculate the difference or gradient between the values of adjacent nodes in the mastery sequence. When the difference exceeds a preset threshold (e.g., a decrease of more than 0.3), the position is identified as a critical sequence point. Subsequently, based on the specific characteristics of the point (e.g., a sharp drop), matching remedial content is retrieved from the teaching content library. For example, the input sequence is [0.8, 0.85, 0.82, 0.4, 0.5]. The target network model calculates and finds that the mastery level from the 3rd node to the 4th node has decreased by 0.42, exceeding the preset decrease threshold of 0.3. In this case, the target network model can identify the 4th node as a critical sequence point, determining that the student has encountered an obstacle in this knowledge point. It can then call upon "basic concept explanation videos" as updated teaching content and push them to the student to fill the gap.
[0046] In other embodiments, the target network model can also employ a neural network structure with an attention mechanism to globally encode the mastery sequence. The attention mechanism calculates the weight of each position in the mastery sequence, with the position having the highest weight being the key sequence point (i.e., the problem point the model is most "concerned" about). The model then combines this with the teaching content to generate customized updated content through generative algorithms (such as the GPT model) or retrieval methods. For example, during a long course, the model uses the attention mechanism to discover that the student has the highest weight score at the 7th node, and the mastery level at that node oscillates at a low level. The model identifies this point as a key sequence point and analyzes that it involves "complex logical reasoning." Therefore, the target network model can generate a set of interactive exercises with step-by-step guidance as updated teaching content to help students gradually overcome difficulties.
[0047] For example, in the process of generating updated teaching content based on key sequence points, user profile attributes of the learners can also be incorporated. It can be understood that the server can analyze the collected multimodal interaction data (including historical interaction data) to generate user profile attributes that match the learners. For example, attributes such as strong logical reasoning skills, a strong need to understand principles, and divergent thinking. Thus, when generating updated teaching content, the target network model can obtain the user profile attributes of the learners and generate updated teaching content that better matches their learning styles.
[0048] Optionally, if the key sequence points identified based on state change information are empty, meaning each sequence point in the mastery level sequence corresponds to a node with high mastery and no interactive conflicts, it indicates that the learner has a good grasp of the current teaching content. In this case, if the teaching content is knowledge itself, then updating the teaching content could be exercises. If the teaching content is exercises, then exercises with a difficulty level greater than the current difficulty can be selected from the question bank as the updated teaching content. Alternatively, when the key sequence points are empty, the next teaching content can be determined based on the teaching syllabus as the updated teaching content. (Reference) Figure 4 If the current teaching content is knowledge itself, then the next teaching content can be an example problem; if the current teaching content is an example problem, then the next teaching content can be a basic exercise; if the current teaching content is a basic exercise, then the next teaching content can be a more difficult exercise.
[0049] Step 208: Send the updated teaching content to the target terminal corresponding to the teaching object so that the updated teaching content can be displayed on the target terminal.
[0050] The aforementioned personalized teaching methods overcome the shortcomings of traditional teaching assessments, which rely solely on single-dimensional answer data or simple behavioral logs, by acquiring multimodal interactive data encompassing facial expressions, voice, and actions. This results in a singular assessment dimension and one-sided results. The introduction of multimodal data captures subtle changes in the learner's emotional cognition and behavioral operations, thereby more realistically and objectively quantifying their mastery of each teaching content node. Secondly, by constructing a mastery sequence and analyzing its state change information using a target network model, dynamic monitoring and temporal feature extraction of the learning process can be achieved. By identifying the fluctuation trends and abrupt changes in mastery over time, key sequence points (such as knowledge gaps and attention inflection points) in the learning path can be accurately located. Furthermore, by generating and pushing updated teaching content based on key sequence points, dynamic closed-loop optimization of teaching content is achieved. Customized teaching resources are generated in real time for identified weak links or key inflection points, solving the problem of fixed teaching content and the inability to dynamically adjust according to real-time student feedback in existing online education technologies. This reduces ineffective repetitive learning, greatly optimizes the learning path, improves overall teaching efficiency and user experience, and enables personalized online teaching.
[0051] In some exemplary embodiments, the multimodal interaction data includes node interaction data corresponding to each teaching content node. The analysis of state change information in the mastery sequence using a target network model, and the identification of key sequence points in the mastery sequence based on the state change information, includes: analyzing the state change information corresponding to the mastery sequence through a temporal feature extraction layer in the target network model; identifying target change sequence segments in the state change information; for each teaching content node corresponding to the target change sequence segment, fusing the node interaction data corresponding to the teaching content node through a multimodal fusion layer in the target network model to obtain a modal conflict reference value corresponding to the teaching content node; determining interactive conflict nodes in the teaching content based on the target modal conflict reference value; the target modal conflict reference value being greater than a preset modal conflict reference value; acquiring a knowledge graph corresponding to the teaching content; determining the preceding content nodes associated with the interactive conflict nodes based on the knowledge graph; and determining key sequence points in the mastery sequence based on the preceding content nodes.
[0052] Among them, node interaction data refers to the fine-grained subset of data corresponding to teaching content nodes in multimodal interaction data. The temporal feature extraction layer refers to a specific network layer in the target network model responsible for processing data with chronological or logical order. The target change sequence segment refers to a continuous segment in the mastery sequence identified by the temporal feature extraction layer that exhibits significant state change characteristics, such as a sharp decline, violent fluctuations, or sustained low mastery level, which can characterize areas where cognitive impairment or attention deficit may exist during the learning process. The multimodal fusion layer refers to the computational unit in the target network model used to integrate information from different modalities, mapping heterogeneous node interaction data (such as facial expression feature vectors and operational behavior vectors) to the same feature space for weighted fusion to uncover the correlation and consistency between different modalities. The modal conflict reference value refers to a numerical indicator used to quantify the degree of consistency between different modalities. A high modal conflict reference value typically indicates a contradiction between the learner's outward performance (such as nodding, calm expression) and the actual interaction results (such as incorrect answers, slow operation), and can be used to determine whether "hidden learning disabilities" exist. Interactive conflict nodes refer to teaching content nodes whose modal conflict reference values exceed a preset threshold. A knowledge graph is a semantic network that stores knowledge points and their interrelationships in a graph structure. It describes the logical dependencies between teaching content nodes, such as the hierarchical structure and dependencies between prerequisite and subsequent knowledge, facilitating the identification of the root causes of problems encountered by the learner. Prerequisite content nodes refer to knowledge point nodes in the logical structure of the knowledge graph that precede interactive conflict nodes and are the foundation or necessary condition for learning those conflict nodes. For example, "addition" is a prerequisite content node for "multiplication." Key sequence points are not necessarily equivalent to conflict nodes; rather, they are the root cause nodes found through knowledge graph tracing that lead to subsequent conflicts.
[0053] For example, firstly, the target network model utilizes the self-attention mechanism in the Transformer architecture as a temporal feature extraction layer to calculate the importance weights of each sequence point in the mastery sequence to the overall sequence, extracting anomalous regions with significantly concentrated weights as target change sequence segments. Subsequently, for nodes within this region, a multimodal fusion network based on an attention mechanism is used to calculate modal conflict reference values between different modal data (such as voice intonation and answer accuracy). Specifically, by calculating the cosine similarity of feature vectors from different modalities, the lower the similarity, the higher the conflict value. The system selects nodes with conflict values greater than a preset value as interactive conflict nodes. Afterward, the system acquires a knowledge graph, uses a graph neural network to analyze the semantic association strength between nodes, determines the parent node with the highest association with the interactive conflict node as the preceding content node, and identifies its corresponding sequence point as a key sequence point.
[0054] In this embodiment, by combining temporal feature extraction, multimodal conflict detection, and knowledge graph tracing technology, it is possible to effectively identify the implicit cognitive conflicts exhibited by students during the learning process and accurately locate the root cause of learning difficulties (i.e., weak prior knowledge), rather than simply intervening in the surface symptoms. This significantly improves the depth of teaching diagnosis and the relevance of updated content, achieving precise and personalized teaching from "treating the symptoms" to "treating the root cause."
[0055] In some embodiments, the mastery level sequence is analyzed by a temporal feature extraction layer in the target network model to identify the state change information corresponding to the mastery level sequence; the target change sequence segment in the state change information is identified, including: converting the mastery level sequence into a feature vector sequence, inputting the feature vector sequence into the temporal feature extraction layer, performing correlation analysis on each vector in the feature vector sequence through the temporal feature extraction layer to obtain temporal state features, which represent the state change information of each teaching content node in the time dimension; calculating the state change gradient between adjacent teaching content nodes based on the temporal state features; determining the target change sequence segment based on the target state change gradient; and ensuring that the target state change gradient is greater than a preset gradient threshold.
[0056] Among them, the feature vector sequence refers to the set of vectors mapped from the numerical data in the mastery level sequence through data preprocessing or embedding techniques. It is an input format that the target network model can directly process, containing not only numerical information about the mastery level but also encoding positional or contextual information. Temporal state features refer to the high-dimensional feature representation output after processing by the temporal feature extraction layer. They are used to characterize the dynamic changes of teaching content nodes in the time dimension, such as upward trends, downward trends, stable states, or fluctuations in mastery level. They represent a deep abstraction of the original numerical sequence. The state change gradient is a numerical indicator used to quantify the drastic change in mastery level between two adjacent teaching content nodes. Mathematically, it is usually represented by a difference value or derivative; the larger the gradient value, the more drastic the sudden change in learning state (e.g., suddenly not understanding).
[0057] For example, firstly, the server converts the mastery sequence (e.g., [0.9, 0.85, 0.4]) into a feature vector sequence. Then, this feature vector sequence is input into a temporal feature extraction layer composed of a Long Short-Term Memory (LSTM) network. The LSTM network performs correlation analysis on the vectors in the sequence, capturing long-distance temporal dependencies and outputting temporal state features that incorporate contextual information. Next, in the output layer or post-processing module, the model calculates the difference between the temporal state features of adjacent nodes to obtain the state change gradient. Finally, all gradient values are traversed, and the locations of nodes with gradient values greater than a preset gradient threshold are marked, forming the target change sequence segment.
[0058] In this embodiment, by converting the mastery sequence into a feature vector and using a temporal feature extraction layer for in-depth correlation analysis and gradient calculation, the dynamic evolution of the learning state can be accurately quantified at the micro level. This effectively filters out minute random fluctuations, thereby accurately identifying key sequence segments that undergo drastic changes or abnormal transitions. This provides a reliable data foundation for the subsequent accurate identification of learning obstacles and significantly improves the sensitivity and accuracy of teaching diagnosis.
[0059] In some embodiments, generating updated teaching content based on key sequence points includes: obtaining knowledge point attributes of key content nodes corresponding to key sequence points; obtaining key features corresponding to key sequence points based on mastery degree sequences; querying a preset teaching strategy mapping table based on knowledge point attributes and key features to obtain a target teaching strategy; querying a teaching content knowledge base corresponding to teaching knowledge points that match the teaching content; the teaching content knowledge base stores at least one type of teaching content corresponding to each teaching knowledge point; and obtaining updated teaching content from the teaching content knowledge base based on the target teaching strategy and knowledge point attributes.
[0060] Among them, key content nodes refer to specific teaching content nodes corresponding to the identified key sequence points, that is, knowledge point units in the learning path of the learner that have problems and require key intervention or remediation. Knowledge point attributes refer to the label information used to describe the inherent cognitive characteristics and types of knowledge points, including at least the dimensions of concept understanding (emphasizing the memorization and comprehension of definitions and principles), logical deduction (emphasizing causal analysis and proof processes), and computational skills (emphasizing formula application and calculation operations). The pre-constructed teaching strategy mapping table refers to a pre-constructed lookup table or set of logical rules that defines which teaching methods should be adopted under different combinations of knowledge point attributes and student status characteristics. For example, for knowledge points of the "concept understanding" category and the characteristic showing "vague memory", the mapping table may point to the strategy of "concept restatement + graphic assistance". The target teaching strategy refers to the teaching intervention plan specifically customized for the current problem, determined by querying the mapping table, which can specify the form of teaching content (such as videos, exercises, and graphics) and the focus (such as strengthening the basics and expanding applications). A teaching content knowledge base refers to a database that stores a large amount of teaching resources. It is indexed by knowledge points and is used to store teaching content materials of various forms and difficulties corresponding to each teaching knowledge point.
[0061] For example, firstly, the server parses key sequence points, identifies corresponding key content nodes, and reads the pre-stored metadata tags (such as "calculation skills") of these nodes as knowledge point attributes. Simultaneously, it extracts features from the mastery level sequence to obtain a feature vector sequence, from which the vector corresponding to the key sequence point is obtained as the key feature. Then, the server inputs the knowledge point attributes and key features into a preset teaching strategy mapping table. This table uses a key-value pair structure, precisely matching attribute tags with feature classifications (such as "calculation skills + high-frequency errors") to output specific target teaching strategies (such as "step-by-step guidance and error correction"). Finally, the server queries the teaching content knowledge base using the ID of the key content node as an index, filtering out resources that match the target teaching strategy (such as exercises with step-by-step prompts) and knowledge point attributes, and generating updated teaching content. For example, in mathematics teaching, the key content node identified is "application of the Pythagorean theorem," with the knowledge point attribute "calculation skills." The obtained key features show that the student's mastery level vector at this node exhibits a "broken logical chain" characteristic. Querying the mapping table yields the target teaching strategy as "step-by-step demonstration." Search the knowledge base of teaching content for video resources related to "Pythagorean theorem" that include "step-by-step breakdown demonstrations", and push a teaching video containing step-by-step calculation animations to students as updated teaching content.
[0062] In this embodiment, by combining the attribute characteristics of knowledge points with the personalized learning status characteristics of students for dual-dimensional analysis, and by using a preset mapping table and knowledge base for linkage retrieval, teaching strategies that not only conform to the characteristics of knowledge point types but also fit the actual weak points of students can be accurately matched, thereby generating highly customized updated teaching content, effectively reducing the interference of redundant information caused by indiscriminate push, and significantly improving the pertinence and effectiveness of teaching intervention.
[0063] In some embodiments, updating teaching content includes content explanation resources, example demonstration resources, and exercise resources; generating updated teaching content based on key sequence points includes: determining the target teaching stage based on the key mastery level corresponding to the key sequence points, wherein the target teaching stage is any one of the content explanation stage, example demonstration stage, or exercise practice stage; when the target teaching stage is the content explanation stage, generating explanation content containing the knowledge point breakdown logic; when the target teaching stage is the example demonstration stage, retrieving target example questions from a preset question bank that match the knowledge point attributes corresponding to the key sequence points, and generating demonstration content of the solution steps for the target example questions; when the target teaching stage is the exercise practice stage, determining the target difficulty level based on the key mastery level, and generating exercise questions that meet the target difficulty level.
[0064] Here, "key mastery level" refers to a numerical value or level that quantitatively represents the learner's mastery of a specific knowledge point, corresponding to a key sequence point. "Targeted teaching stage" refers to the learning state level most suitable for the learner, determined by the system based on the key mastery level. This embodiment divides it into three progressive stages: "content explanation stage" (basic cognition), "example demonstration stage" (deepening understanding), and "exercise practice stage" (consolidation and application).
[0065] For example, firstly, the server obtains the key mastery level value corresponding to the key sequence point and compares it with a preset mastery level range threshold. If the value is in the low range (e.g., 0-0.4), the target teaching stage is determined to be the "content explanation stage." The server calls a preset knowledge graph to decompose the key knowledge points into sub-modules such as "definition-principle-characteristics," generating explanation content containing the knowledge point breakdown logic. If the value is in the middle range (e.g., 0.4-0.7), the stage is determined to be the "example demonstration stage." The server searches the preset question bank using knowledge point attributes as keywords, extracts typical examples, and uses a template engine to generate problem-solving step demonstration content with key step annotations. If the value is in the high range (e.g., 0.7-1.0), the stage is determined to be the "exercise practice stage." The server maps the specific high or low value to the corresponding target difficulty level and selects practice questions that match that level from the question bank.
[0066] In this embodiment, the teaching intervention is precisely divided into three progressive stages—explanation, demonstration, and practice—based on the students' mastery of key concepts. Specific content generation logic (deconstruction of explanation, step-by-step demonstration, and difficulty-level practice) is configured for each stage. This enables a refined and tiered supply of teaching content, effectively avoiding cognitive overload caused by directly pushing high-difficulty exercises to students with low mastery levels, or low learning efficiency caused by repeatedly pushing basic explanations to students with high mastery levels. This ensures that the difficulty of the teaching intervention is matched to the students' actual level, thereby significantly improving learning efficiency and the utilization rate of teaching resources.
[0067] In some embodiments, sending updated teaching content to a target terminal corresponding to the learner to display the updated teaching content on the target terminal includes: calling a preset digital human rendering engine to obtain target digital human driving data that matches the type of the updated teaching content; the driving data includes facial expression data, body movement data, and voice broadcast data; fusing and rendering the updated teaching content with the target digital human driving data to generate a digital human teaching video stream; and sending the digital human teaching video stream to the target terminal so that the target terminal can perform interactive questioning actions with the learner through the digital human image while playing the updated teaching content.
[0068] The digital human rendering engine refers to a software module or system integrating computer graphics, motion capture, and artificial intelligence technologies. It is used to build, drive, and render virtual human figures, generating high-fidelity facial expressions, body movements, and voice output in real time based on input command data. Target digital human driving data refers to a set of parameterized data used to control the performance of the virtual digital human figure. This may include: facial expression data (such as phoneme positioning for lip-sync, eyebrow and eye movement parameters for expressing emotions), body movement data (such as skeletal animation parameters for gestures, standing posture, nodding, and shaking), and voice broadcasting data (such as audio streams generated by text-to-speech technology). Fusion rendering refers to the process of synthesizing visual elements of teaching content (such as text, images, and exercise interfaces) with visual elements of the digital human figure at the screen level. Through techniques such as layer overlay, background transparency, or 3D scene embedding, the digital human figure can naturally blend into the teaching content display environment. The digital human teaching video stream refers to a continuous video data stream generated after fusion rendering, containing both digital human figure demonstrations and teaching content displays.
[0069] For example, the server invokes a pre-built digital human rendering engine to parse and update the type of teaching content (such as "exercises"), loads the corresponding teaching script from the database, and synthesizes speech data using text-to-speech technology. Then, based on the emotional tags of the speech text, the server generates matching facial expression data (such as a serious contemplative expression) and body movement data (such as a finger pointing at the screen), which together constitute the target digital human's driving data. Next, the digital human rendering engine uses the graphics processor for real-time fusion rendering, overlaying the digital human layer onto the teaching content interface and encoding it to generate a digital human teaching video stream. Finally, the video stream is sent to the target terminal via a real-time transmission protocol.
[0070] In this embodiment, by introducing a digital human rendering engine and generating multimodal driving data including face, body and voice, static or traditional teaching content is transformed into vivid digital human teaching video streams. This technical solution can significantly enhance the immersion and interactivity of the teaching process, simulate a real "teacher-student learning companion" scenario, effectively alleviate the boredom and loneliness of students during the learning process, and improve students' focus and learning interest through more intuitive and vivid interactive questioning actions.
[0071] In one exemplary embodiment, the server determines the scope of content to be taught in this lesson based on the subject syllabus. The syllabus, combined with real-time student classroom behavior records, determines the content requiring personalized instruction, such as the solution to a quadratic equation. The explanation process involves a digital teacher interacting with students through interactive questioning.
[0072] like Figure 4 As shown, real-time interactive data of students during the content explanation process is collected (corresponding to obtaining multimodal interactive data of teaching subjects regarding teaching content), and the students' mastery is judged (corresponding to obtaining a mastery level sequence corresponding to the teaching content). Using a personalized recommendation system, exclusive teaching content suitable for students' current mastery level is recommended (updated teaching content will be sent to the target terminal), and the digital human teacher interacts with students through interactive questioning to explain content, examples, or exercises.
[0073] In this embodiment, a student's mastery level of the current teaching content can be determined based on a mastery level sequence. A recommendation model (corresponding to the target network model) then determines the recommended difficulty level based on this mastery level. Matching teaching content or questions are then retrieved from a teaching content library or question bank based on the recommended difficulty level. For example, the teaching content library or question bank may have difficulty labels, such as being divided into easy, medium, and difficult levels. Alternatively, the mastery level can be determined based on the student's answer accuracy rate (corresponding to multimodal interaction data). Then, content, examples, and exercises of different difficulties (corresponding to updated teaching content) are recommended based on the mastery level. For example, a high accuracy rate results in higher difficulty content or questions, while a low accuracy rate results in lower difficulty content or questions.
[0074] In this embodiment, by analyzing students' classroom behavior in real time and based on their actual understanding, the online course provides personalized content explanations, personalized examples, and personalized exercises to truly achieve individualized teaching; and for the same lesson, each student learns completely different content, realizing customized teaching.
[0075] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.
[0076] Based on the same inventive concept, this application also provides a personalized teaching device for implementing the personalized teaching method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more personalized teaching device embodiments provided below can be found in the limitations of the personalized teaching method described above, and will not be repeated here.
[0077] In one exemplary embodiment, such as Figure 3 A personalized teaching device 300 is provided, comprising: an acquisition module 301, a sequence module 302, a processing module 303, and a communication module 304, wherein:
[0078] The acquisition module 301 is used to acquire multimodal interaction data of the teaching subjects in relation to the teaching content; the teaching content includes at least one teaching content node.
[0079] The sequence module 302 is used to obtain the node mastery level corresponding to each teaching content node based on multimodal interactive data, and to obtain the mastery level sequence corresponding to the teaching content based on the mastery level of each node.
[0080] The processing module 303 is used to input the teaching content and mastery level sequence into the target network model, analyze the state change information of the mastery level sequence through the target network model, identify key sequence points in the mastery level sequence based on the state change information, and generate updated teaching content based on the key sequence points.
[0081] The communication module 304 is used to send updated teaching content to the target terminal corresponding to the teaching object so that the updated teaching content can be displayed on the target terminal.
[0082] In some embodiments, the multimodal interaction data includes node interaction data corresponding to each teaching content node. In analyzing the state change information of the mastery degree sequence through the target network model and identifying key sequence points in the mastery degree sequence based on the state change information, the processing module 303 is specifically used for: analyzing the state change information corresponding to the mastery degree sequence through the temporal feature extraction layer in the target network model; identifying the target change sequence segment in the state change information; for each teaching content node corresponding to the target change sequence segment, fusing the node interaction data corresponding to the teaching content node through the multimodal fusion layer in the target network model to obtain the modal conflict reference value corresponding to the teaching content node; determining the interactive conflict node in the teaching content based on the target modal conflict reference value; the target modal conflict reference value being greater than a preset modal conflict reference value; acquiring a knowledge graph corresponding to the teaching content; determining the preceding content node associated with the interactive conflict node based on the knowledge graph; and determining the key sequence point in the mastery degree sequence based on the preceding content node.
[0083] In some embodiments, in analyzing the state change information corresponding to the mastery degree sequence through the temporal feature extraction layer in the target network model, and identifying the target change sequence segment in the state change information, the processing module 303 is specifically used to: convert the mastery degree sequence into a feature vector sequence, input the feature vector sequence into the temporal feature extraction layer, perform correlation analysis on each vector in the feature vector sequence through the temporal feature extraction layer to obtain temporal state features, the temporal state features representing the state change information of each teaching content node in the time dimension; calculate the state change gradient between adjacent teaching content nodes based on the temporal state features; determine the target change sequence segment based on the target state change gradient; and ensure that the target state change gradient is greater than a preset gradient threshold.
[0084] In some embodiments, in generating and updating teaching content based on key sequence points, the processing module 303 is specifically used to: obtain the knowledge point attributes of key content nodes corresponding to key sequence points; the knowledge point attributes include at least conceptual understanding, logical deduction, and computational skills; obtain key features corresponding to key sequence points based on the mastery level sequence; query a preset teaching strategy mapping table based on the knowledge point attributes and key features to obtain the target teaching strategy; query the teaching content knowledge base corresponding to the teaching knowledge points that match the teaching content; the teaching content knowledge base stores at least one type of teaching content corresponding to each teaching knowledge point; and obtain updated teaching content from the teaching content knowledge base based on the target teaching strategy and knowledge point attributes.
[0085] In some embodiments, in generating and updating teaching content based on key sequence points, the processing module 303 is specifically used to: determine the target teaching stage based on the key mastery level corresponding to the key sequence points, wherein the target teaching stage is any one of the content explanation stage, example demonstration stage, or exercise practice stage; when the target teaching stage is the content explanation stage, generate explanation content containing the knowledge point decomposition logic; when the target teaching stage is the example demonstration stage, retrieve target example questions from a preset question bank that match the knowledge point attributes corresponding to the key sequence points, and generate demonstration content of the solution steps of the target example questions; when the target teaching stage is the exercise practice stage, determine the target difficulty level based on the key mastery level, and generate exercise questions that meet the target difficulty level.
[0086] In some embodiments, in sending updated teaching content to the target terminal corresponding to the teaching object so as to display the updated teaching content on the target terminal, the communication module 303 is specifically used to: call a preset digital human rendering engine to obtain target digital human driving data that matches the type of the updated teaching content; the driving data includes facial expression data, body movement data, and voice broadcast data; fuse and render the updated teaching content with the target digital human driving data to generate a digital human teaching video stream; and send the digital human teaching video stream to the target terminal so that the target terminal can perform interactive questioning actions with the teaching object through the digital human image during the playback of the updated teaching content.
[0087] Each module in the aforementioned personalized teaching device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each module.
[0088] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 5As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The database stores students' historical interaction data and other relevant data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network. When the computer program is executed by the processor, it implements a personalized teaching method.
[0089] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0090] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.
[0091] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.
[0092] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.
[0093] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.
[0094] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application 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 memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic 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. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.
[0095] The technical features of the above embodiments can be combined in any way. For the sake of brevity, 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.
[0096] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. A personalized teaching method, characterized in that, The method includes: Acquire multimodal interaction data of the learners in response to the teaching content; the teaching content includes at least one teaching content node; Based on the multimodal interaction data, the node mastery level corresponding to each of the teaching content nodes is obtained, and based on the node mastery level, a mastery level sequence corresponding to the teaching content is obtained; The teaching content and the mastery level sequence are input into the target network model. The target network model analyzes the state change information of the mastery level sequence, identifies key sequence points in the mastery level sequence based on the state change information, and generates updated teaching content based on the key sequence points. The updated teaching content is sent to the target terminal corresponding to the teaching object so that the updated teaching content can be displayed on the target terminal.
2. The method according to claim 1, characterized in that, The multimodal interaction data includes node interaction data corresponding to each of the teaching content nodes. The step of analyzing the state change information of the mastery level sequence through the target network model, and identifying key sequence points in the mastery level sequence based on the state change information, includes: The temporal feature extraction layer in the target network model is used to analyze the state change information corresponding to the mastery level sequence; the target change sequence segment in the state change information is identified. For each teaching content node corresponding to the target change sequence segment, the node interaction data corresponding to the teaching content node is fused through the multimodal fusion layer in the target network model to obtain the modal conflict reference value corresponding to the teaching content node; Interactive conflict nodes in the teaching content are determined based on a target modal conflict reference value; the target modal conflict reference value is greater than a preset modal conflict reference value. Obtain the knowledge graph corresponding to the teaching content; Based on the knowledge graph, determine the preceding content nodes associated with the interaction conflict nodes; The key sequence points in the mastery level sequence are determined based on the preceding content nodes.
3. The method according to claim 2, characterized in that, The process involves analyzing the state change information corresponding to the mastery sequence through the temporal feature extraction layer in the target network model; identifying the target change sequence segment in the state change information, including: The mastery sequence is converted into a feature vector sequence, and the feature vector sequence is input into the temporal feature extraction layer. The temporal feature extraction layer performs correlation analysis on each vector in the feature vector sequence to obtain temporal state features. The temporal state features represent the state change information of each teaching content node in the time dimension. Based on the temporal state characteristics, calculate the state change gradient between adjacent teaching content nodes; The target change sequence segment is determined based on the gradient of the target state change; the gradient of the target state change is greater than a preset gradient threshold.
4. The method according to claim 1, characterized in that, The generation and updating of teaching content based on the key sequence points includes: Obtain the knowledge point attributes of the key content nodes corresponding to the key sequence points; the knowledge point attributes include at least conceptual understanding, logical deduction, and computational skills. Based on the mastery level sequence, key features corresponding to the key sequence points are obtained; Based on the knowledge point attributes and the key features, the target teaching strategy is obtained by querying the preset teaching strategy mapping table. The system queries the teaching content knowledge base corresponding to the teaching knowledge points that match the teaching content; the teaching content knowledge base stores at least one type of teaching content corresponding to each teaching knowledge point. The updated teaching content is obtained from the teaching content knowledge base based on the target teaching strategy and the knowledge point attributes.
5. The method according to claim 1, characterized in that, The updated teaching content includes content explanation resources, example demonstration resources, and exercise resources; the generation of updated teaching content based on the key sequence points includes: Based on the level of mastery of the key sequence points, the target teaching stage is determined. The target teaching stage can be any one of the content explanation stage, example demonstration stage, or exercise practice stage. When the target teaching stage is the content explanation stage, explanation content containing the logic of knowledge point breakdown is generated; When the target teaching stage is the example demonstration stage, target example questions that match the knowledge point attributes corresponding to the key sequence points are retrieved from the preset question bank, and the solution steps of the target example questions are generated. When the target teaching stage is the exercise practice stage, the target difficulty level is determined based on the level of mastery of the key points, and exercise questions that meet the target difficulty level are generated.
6. The method according to claim 1, characterized in that, Sending the updated teaching content to the target terminal corresponding to the learner, so as to display the updated teaching content on the target terminal, includes: The preset digital human rendering engine is invoked to obtain target digital human driving data that matches the type of the updated teaching content; the target digital human driving data includes facial expression data, body movement data, and voice broadcast data. The updated teaching content is fused and rendered with the target digital human driving data to generate a digital human teaching video stream; The digital human teaching video stream is sent to the target terminal, so that the target terminal can perform interactive questioning actions with the teaching object through the digital human image while playing the updated teaching content.
7. The method according to any one of claims 1 to 6, characterized in that, The multimodal interaction data includes at least one of screen touch data, voice response data, and facial image data.
8. A personalized teaching device, characterized in that, The device includes: The acquisition module is used to acquire multimodal interaction data of the teaching subjects in response to the teaching content; the teaching content includes at least one teaching content node; The sequence module is used to obtain the node mastery level corresponding to each of the teaching content nodes based on the multimodal interaction data, and to obtain the mastery level sequence corresponding to the teaching content based on the mastery level of each node. The processing module is used to input the teaching content and the mastery level sequence into the target network model, analyze the state change information of the mastery level sequence through the target network model, identify key sequence points in the mastery level sequence based on the state change information, and generate updated teaching content based on the key sequence points. A communication module is used to send the updated teaching content to the target terminal corresponding to the teaching object, so as to display the updated teaching content on the target terminal.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.
10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.
