Intelligent teaching assistance method and system based on large language model

Abstract

The invention belongs to the technical field of artificial intelligence education, and relates to an intelligent teaching assistance method and system based on a large language model. The method comprises the steps that text data, voice data and eye movement track data of students are acquired, feature extraction is carried out, and a three-dimensional cognitive state tensor model is constructed; constructing a cognitive state tensor decomposition constraint optimization function, performing tensor decomposition, and outputting a core tensor and a factor matrix; generating a teaching strategy, and defining a cross-modal kernel function to realize feature association; tensor decomposition parameters are dynamically adjusted; and outputting the updated teaching strategy. According to the method, the cognitive change trajectory of a learner can be captured from multiple dimensions, and the accuracy and comprehensiveness of cognitive state characterization are remarkably improved through a cross-modal kernel function and a gradient coupling mechanism; constructing a cognitive state tensor decomposition constraint optimization function to describe an evolution law of a cognitive state, and capturing a continuous change characteristic of the cognitive state along with time through a dynamic constraint condition of a time factor matrix.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of artificial intelligence education, and specifically relates to an intelligent teaching auxiliary method and system based on a large language model. BACKGROUND

[0002] At present, with the rapid development of artificial intelligence technology, intelligent teaching systems, as an important carrier of educational informatization, are profoundly changing the traditional teaching mode. Intelligent teaching systems can greatly improve teaching efficiency and learning outcomes by automatically analyzing learner status and generating teaching strategies to provide decision support for educators. In the context of large-scale online education and distance learning, intelligent teaching systems can effectively address the problem of insufficient teacher resources, realize the educational philosophy of teaching students in accordance with their aptitude, and meet the diverse educational needs of a lifelong learning society.

[0003] Various technical routes have been formed in the field of existing intelligent teaching systems. Teaching systems based on knowledge graphs construct hierarchical association networks between subject knowledge points to realize learning path planning and knowledge point recommendation. Systems based on learning behavior analysis use sequence models to process learning trajectory data to predict knowledge point mastery and learning difficulties. Systems based on natural language processing analyze learners' text answers to assess understanding depth and concept confusion. Systems based on multimedia analysis use audio and video processing techniques to capture learners' emotional state and attention distribution. These technical solutions have been applied in various educational scenarios and have achieved certain teaching effects.

[0004] Although existing intelligent teaching technology has made significant progress, there are still some deficiencies. First, existing systems rely on single-modal data for analysis. Since cognitive processes inherently involve multi-channel information processing, single-modal analysis cannot fully capture learners' internal cognitive changes, resulting in incomplete state representation. Second, existing cognitive modeling methods mostly use static representation frameworks, which fail to effectively capture time-varying factors such as memory decay and attention fluctuations, resulting in insufficient prediction accuracy for long-term learning processes. Third, traditional teaching strategy generation lacks a systematic theoretical framework, making it difficult to balance cognitive goals and teaching costs while ignoring key teaching constraints, resulting in poor strategy feasibility. In addition, existing systems generally lack a closed-loop feedback mechanism, making it difficult to dynamically adjust parameters based on learning outcomes and address individual differences. SUMMARY

[0005] To solve the above technical problems, the present application provides an intelligent teaching auxiliary method and system based on a large language model.

[0006] In a first aspect, the present application provides an intelligent teaching auxiliary method based on a large language model, comprising: The text data, voice data and eye movement trajectory data of students are acquired, feature extraction is performed, and a three-dimensional cognitive state tensor model is constructed; A cognitive state tensor decomposition constraint optimization function is constructed, the cognitive state tensor is subjected to tensor decomposition with a constraint condition, and a core tensor and a factor matrix are outputted; A teaching strategy is generated according to the core tensor and the factor matrix; Feature correlation is realized by defining a cross-modal kernel function; The tensor decomposition parameters are dynamically adjusted according to the strategy execution result and the feature correlation degree, and the updated core tensor and factor matrix are obtained; The updated teaching strategy is outputted according to the updated core tensor and factor matrix.

[0007] In a second aspect, the present application provides an intelligent teaching auxiliary system based on a large language model, comprising a multi-modal data processing unit, a cognitive state modeling unit, a teaching strategy generation unit, a multi-modal fusion unit, a feedback control unit and an update output unit; The multi-modal data processing unit is used for acquiring text data, voice data and eye movement trajectory data of students, performing feature extraction, and constructing a three-dimensional cognitive state tensor model; The cognitive state modeling unit is used for constructing a cognitive state tensor decomposition constraint optimization function, performing tensor decomposition on the cognitive state tensor with a constraint condition, and outputting a core tensor and a factor matrix; The teaching strategy generation unit is used for generating a teaching strategy according to the core tensor and the factor matrix; The multi-modal fusion unit is used for realizing feature correlation by defining a cross-modal kernel function; The feedback control unit is used for dynamically adjusting tensor decomposition parameters according to the strategy execution result and the feature correlation degree, and obtaining an updated core tensor and factor matrix; The update output unit is used for outputting an updated teaching strategy according to the updated core tensor and factor matrix.

[0008] On the basis of the above technical solution, the present application can also be improved as follows.

[0009] Further, the text data, voice data and eye movement trajectory data of students are acquired, feature extraction is performed, and a three-dimensional cognitive state tensor model is constructed, comprising: The text feature vector is calculated according to the text data; let be the input text sequence, be the classification identifier of the model, be the semantic vector dimension, represent a real number, be a real vector space of dimension, then the text feature vector is represented as: ; According to the voice data, a voice feature matrix is calculated; let be the voice spectrum graph, be the Mel cepstrum coefficient extraction method, be the cepstrum coefficient order, be a real number matrix space, then the voice feature matrix is represented as: ; According to the eye movement trajectory data, an eye movement feature function is calculated; let be the eye movement trajectory data, be the time horizontal gaze coordinates, be the time vertical gaze coordinates, be the time pupil diameter, be the gaze density function, be the starting time of the teaching segment, be the end time of the teaching segment, be the feature dimension, be a real number vector space of dimension, then the eye movement feature function is represented as: ; Let the cognitive state tensor be , , be the total number of knowledge points, be the number of time slices, be the number of modalities, be the feature dimension of each modality, be the trainable weight parameter, be the feature vector of the modality at time , then the three-dimensional cognitive state tensor model is represented as: .

[0010] Further, a cognitive state tensor decomposition constraint optimization function is constructed, including: let be the cognitive state tensor, be the core tensor, representing the cross-modal knowledge association, be the knowledge point factor matrix, be the time factor matrix, be the modality factor matrix, be the tensor mod product, be the Frobenius norm, and The regularization coefficient is . If the identity matrix is ​​denoted by , then the cognitive state tensor decomposition constrained optimization function is expressed as: ; set up Adjusting parameters for knowledge point sparsity. The core tensor rank constraint value, To extract the diagonal elements of the matrix, Represents the core tensor If the rank is , then the constraint condition is expressed as: .

[0011] Furthermore, the cognitive state tensor is decomposed with constraints to output the core tensor and factor matrix, including applying orthogonal constraints to the factor matrix; let... Time factor matrix The row vectors The first of the time factor matrix row vectors It is an exponentially decaying term. It is a periodic term. For the standard Euclidean inner product, For teaching cycle parameters, To represent the memory decay rate, applying orthogonal constraints to the factor matrix is ​​expressed as: .

[0012] Furthermore, teaching strategies are generated based on the core tensor and factor matrix, including: Based on the core tensor and the knowledge point factor matrix, a mixed-integer nonlinear programming model for teaching strategies is constructed. set up For teaching strategy vectors, Indicates the selection of the first One knowledge point, This is the knowledge point factor matrix for the current teaching cycle. This is the knowledge point factor matrix of the previous teaching cycle. and These are the weighting coefficients. For the first The teaching cost of each knowledge point For the pedagogical constraint matrix, This is the threshold vector for achieving teaching methodology standards. This represents the maximum number of knowledge points covered in a single lesson. For a set of conflicting knowledge points, For the core tensor, For tensor Modular product, is the Frobenius norm, the mixed integer nonlinear programming model of teaching strategy is represented as: ; ; ; ; The mixed integer nonlinear programming model of teaching strategy is solved by using a branch and bound algorithm, and the objective function value corresponding to the optimal teaching strategy is output.

[0013] Further, the feature correlation is realized by defining a cross-modal kernel function, including: a modal factor matrix and an original feature, defining a cross-modal kernel function; Let and be a multi-modal input sample, be the number of modalities, be the feature mapping function of the th modal, be the weight matrix of the th modal, be the modal correlation matrix, , be the kernel scale parameter, be the th modal feature vector of the sample , be the cross-modal kernel function value, then the cross-modal kernel function is represented as: ; Let be the cross-modal coupling coefficient, be the gradient of the th modal mapping to the th modal input, be the gradient vector of the kernel function to the th modal input feature, be the hidden space mapping function of the th modal, and the gradient coupling condition is satisfied: ; Output the modal correlation matrix.

[0014] Further, for the feature mapping function, a nonlinear transformation is realized by a multi-layer perception, and the LeakyReLU activation function is selected as the hidden layer activation function, and the weight matrix is updated through a differentiable convex optimization layer; let be the number of batch samples, represent the weighted norm, be the The kernel scale parameter, updated in the next iteration, is used to control the sensitivity of the distance metric in the kernel function. For the first in the batch A multimodal input sample, For the first in the batch Each of the multimodal input samples contains three modalities: text, speech, and eye-tracking features. This is a multimodal joint mapping function that maps samples from the original feature space. Mapped to a unified latent space representation For the iteration round counter, for 3D modal correlation matrix For modal number; kernel scale parameter Dynamically adjusted based on feature distribution, expressed as: .

[0015] Furthermore, set To predict the cognitive state tensor, These are the predicted values ​​from tensor decomposition reconstruction. Let L2 norm be the vector. The cognitive state prediction error is given by the cognitive state tensor. , ; The cognitive state prediction error is expressed as: ; set up For the first The weight of teaching strategies in each teaching cycle For the first The weight of teaching strategies in each teaching cycle , and For PID control parameters, For the first Prediction error for each teaching cycle Add variables to the teaching cycle. For the first The prediction error value for each teaching cycle, adjusted using a PID controller to adjust the weights of the teaching strategy, is expressed as follows: ; set up For the first The maximum number of knowledge points allowed per teaching cycle. To adjust the step size for capacity, The rate of change of the prediction error for cognitive state. For symbolic functions, If the function is a floor function, then the dynamic adjustment of the teaching capacity constraint is expressed as: .

[0016] The beneficial effects of this invention are: (1) This invention constructs a dynamic cognitive state representation model through multimodal data fusion and tensor decomposition, realizing the collaborative analysis of text, speech and eye-tracking data in the teaching process, effectively overcoming the problem of incomplete information in a single modality, and capturing the cognitive change trajectory of learners from multiple dimensions. Through cross-modal kernel function and gradient coupling mechanism, the system can identify the implicit association patterns between various modalities, significantly improving the accuracy and comprehensiveness of cognitive state representation. (2) The present invention constructs a cognitive state tensor decomposition constraint optimization function to describe the evolution law of cognitive state. By separating the core tensor and factor matrix, it represents the potential features of the three dimensions of knowledge points, time and modality, so that the teaching system can deeply understand the memory decay law and attention fluctuation pattern in the learning process. The cognitive state tensor decomposition constraint optimization function not only has good interpretability, but also captures the continuous change characteristics of cognitive state over time through the dynamic constraint conditions of the time factor matrix, providing a precise quantitative basis for the generation of subsequent teaching strategies. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of an intelligent teaching assistance method based on a large language model provided in Embodiment 1 of the present invention; Figure 2 This is a block diagram illustrating the principle of an intelligent teaching support system based on a large language model, as provided in Embodiment 2 of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0019] Example 1 As an example, see the attached document. Figure 1 As shown, to solve the above-mentioned technical problems, this embodiment provides an intelligent teaching assistance method based on a large language model, including: Acquire students' text data, voice data, and eye-tracking trajectory data, extract features, and construct a three-dimensional cognitive state tensor model; Construct a constraint optimization function for the cognitive state tensor decomposition, perform constraint-based tensor decomposition on the cognitive state tensor, and output the core tensor and factor matrix. generate teaching strategies according to the core tensor and the factor matrix; realize feature association by defining a cross-modal kernel function; dynamically adjust tensor decomposition parameters according to the strategy execution result and the feature association degree, obtain an updated core tensor and a factor matrix; output an updated teaching strategy according to the updated core tensor and the factor matrix.

[0020] The application realizes the collaborative analysis of text, speech and eye movement data in the teaching process by constructing a dynamic cognitive state representation model through multi-modal data fusion and tensor decomposition, effectively overcomes the problem of incomplete information of a single mode, can capture the cognitive change trajectory of learners from multiple dimensions, and through the cross-modal kernel function and the gradient coupling mechanism, the system can identify the implicit association mode between modes, and significantly improves the accuracy and comprehensiveness of cognitive state representation.

[0021] The application constructs a cognitive state tensor decomposition constraint optimization function to describe the evolution law of the cognitive state, separates the latent features of the three dimensions of knowledge points, time and modes through the core tensor and the factor matrix, so that the teaching system can deeply understand the memory decay law and attention fluctuation mode in the learning process, the cognitive state tensor decomposition constraint optimization function not only has good interpretability, but also captures the continuous change characteristics of the cognitive state with time through the dynamic constraint condition of the time factor matrix, and provides accurate quantitative basis for subsequent teaching strategy generation.

[0022] Optionally, text data, speech data and eye movement trajectory data of students are acquired, and feature extraction is performed to construct a three-dimensional cognitive state tensor model, including: a text feature vector is calculated according to the text data; let be the input text sequence, be the classification identifier of the model, be the semantic vector dimension, represent a real number, be a real number vector space of dimension, then the text feature vector is represented as: ; a speech feature matrix is calculated according to the speech data; let be a speech spectrogram, be a mel cepstrum coefficient extraction method, be a cepstrum coefficient order, be a real number matrix space of dimension, then the speech feature matrix is represented as: ; According to the eye movement trajectory data, an eye movement feature function is calculated; set as the eye movement trajectory data, as the horizontal gaze coordinates, as the vertical gaze coordinates, as the pupil diameters, as the gaze density function, as the start time of the teaching segment, as the end time of the teaching segment, as the feature dimension, as a real vector space of dimension d, the eye movement feature function is represented as: ; ; Let the cognitive state tensor be , , as the total number of knowledge points, as the number of time slices, as the number of modalities, as the feature dimension of each modality, as the trainable weight parameters, as the feature vector of the i-th modality at time t, the three-dimensional cognitive state tensor model is represented as: . .

[0023] Through the above processing flow, the alignment and fusion of heterogeneous teaching data in a unified mathematical representation space are realized, laying a data foundation for subsequent cognitive state analysis and strategy generation of the system.

[0024] The standardized feature tensor is decomposed into the modal product form of the core tensor and the factor matrix using the Tucker decomposition framework. The decomposition objective function is defined as the weighted sum of the reconstruction error and the regularization term.

[0025] Optionally, a cognitive state tensor decomposition constraint optimization function is constructed, including: let as the cognitive state tensor, as the core tensor representing cross-modal knowledge association, as the knowledge point factor matrix, as the time factor matrix, as the modality factor matrix, as the tensor modular product, as the Frobenius norm, and as the regularization coefficients, ​​​is an identity matrix, the cognitive state tensor decomposition constrained optimization function is represented as: ; The physical interpretability of the decomposition result is guaranteed by imposing three constraints: first, the knowledge point factor matrix is constrained by an upper bound on the Frobenius norm to prevent overfitting and maintain feature stability; second, the time factor matrix is required to be column-wise unitized to ensure the scale consistency of the temporal features; finally, the rank of the core tensor is limited to a pre-set threshold to capture the main cross-modal association patterns.

[0026] Let be the knowledge point sparsity adjustment parameter, be the core tensor rank constraint value, be the diagonal elements of the extraction matrix, denote the rank of the core tensor , the constraint condition is represented as: .

[0027] Optionally, the cognitive state tensor is subjected to constrained tensor decomposition, and the output core tensor and factor matrices include orthogonal constraints imposed on the factor matrices. Let be the row vector of the time factor matrix , be the row vector of the time factor matrix, be the exponential decay term, be the periodic term, be the standard Euclidean inner product, be the teaching cycle parameter, be the memory decay rate, the orthogonal constraint imposed on the factor matrices is represented as: .

[0028] For the characteristics of the time dimension, dynamic constraints are introduced to strengthen the temporal continuity, and the inner product of adjacent time slice factor vectors satisfies the exponential decay periodic law.

[0029] The alternating least squares method is used for optimization and solution, and the core tensor and each factor matrix are iteratively updated. After each iteration, the factor matrices are normalized and projected: the is projected into the Frobenius sphere, the column vectors of are unitized, and the low-rank approximation of the core tensor is achieved through singular value decomposition. After the optimization process converges, the decomposed , , and to downstream modules for policy generation.

[0030] Through the above process, the interpretable modeling of multi-modal cognitive state is realized, the core tensor encodes the cross-modal interaction mode, and the factor matrix represents the knowledge point correlation strength, time evolution law and mode contribution weight respectively, providing quantitative basis for system adaptive policy generation.

[0031] Based on the analysis results of cognitive state, an optimized teaching scheme is generated, and the specific implementation is as follows: According to the core tensor and the knowledge point factor matrix , a multi-objective optimization function is constructed to balance the cognitive development needs and teaching implementation cost, and the objective function is defined as a weighted double criterion optimization form: .

[0032] Optionally, according to the core tensor and the factor matrix, a teaching strategy is generated, including: According to the core tensor and the knowledge point factor matrix, a mixed integer nonlinear programming model of teaching strategy is constructed; Let be the teaching strategy vector, represent the selection of the th knowledge point, be the knowledge point factor matrix of the current teaching period, be the knowledge point factor matrix of the previous teaching period, and be the weight coefficient, be the teaching cost of the th knowledge point, be the teaching method constraint matrix, be the teaching method threshold value vector, be the maximum number of knowledge points for single teaching, be the conflict knowledge point pair set, be the core tensor, be the tensor mod product, be the Frobenius norm, then the mixed integer nonlinear programming model of the teaching strategy is expressed as: ; The feasibility and effectiveness of the teaching scheme are ensured by imposing three constraints. First, the teaching method adaptability constraint is introduced: ; This constraint forces the teaching strategy to comply with the preset teaching method principle, such as the requirement of knowledge point correlation strength under constructivism guidance.

[0033] Secondly, set the knowledge point capacity constraint: ; Maximum number of knowledge points in single teaching It represents the maximum number of knowledge points allowed to be covered in a single teaching cycle, and its value is dynamically adjusted by the feedback control module. This constraint prevents cognitive overload and ensures teaching quality. Preferably, the initial value is set to 10.

[0034] Finally, define the conflict avoidance constraint: ; This constraint avoids mutual interference of knowledge points being included in the teaching plan at the same time, such as the timing conflict between prerequisite knowledge points and advanced knowledge points.

[0035] Solve the mixed integer nonlinear programming model of teaching strategy by using branch and bound algorithm, and output the objective function value corresponding to the optimal teaching strategy. First, take the continuous relaxation problem as the lower bound estimation, and gradually approach the optimal solution by iteratively partitioning the feasible region and calculating the boundary value; introduce the cut plane technique to accelerate convergence during the solving process, and add effective inequalities to eliminate non-integer solutions; finally output the binary strategy vector , , wherein , represents selecting the th knowledge point into the current teaching cycle, is the optimal solution of the mixed integer nonlinear programming model of teaching strategy, that is, the optimal teaching strategy vector finally output by the algorithm; is the th component or element in the optimal teaching strategy vector .

[0036] By generating the teaching strategy according to the core tensor and the factor matrix as described above, the quantitative generation and optimization of the teaching strategy are realized, and under the premise of ensuring the compliance of the teaching method, the dynamic balance between the cognitive development goal and the resource investment cost is achieved, providing executable scheme guidance for subsequent teaching implementation.

[0037] The present application realizes multi-objective optimization of teaching strategy based on mixed integer programming method, generates the optimal teaching plan under the constraint condition by balancing the two key indicators of cognitive state change and teaching cost. The strategy generation mechanism considers multiple constraints such as teaching method adaptability, knowledge point capacity and conflict avoidance, ensuring the practicality and executability of the teaching strategy, effectively solving the problem of lack of theoretical basis in strategy generation in traditional teaching system.

[0038] Optionally, feature association is realized by defining a cross-modal kernel function, including: Modal factor matrix and original features, define a cross-modal kernel function; Let and For multimodal input samples, For the number of modes, For the first Modal feature mapping function, For modality The weight matrix, The modal incidence matrix, , For kernel scale parameters, For the sample The Modal eigenvectors If the value is a cross-modal kernel function, then the cross-modal kernel function is expressed as: ; A cross-modal kernel function is constructed to achieve feature alignment and interaction. The kernel function is defined as an exponential distance metric. This kernel function measures the similarity of cross-modal samples through weighted Mahalanobis distance. The exponential transformation maps linearly inseparable features to a high-dimensional separable space.

[0039] set up For cross-modal coupling coefficients, For the first Modal mapping pairs Gradient of modal input, For kernel function on the 1st The gradient vector of the modal input features. For the first The latent space mapping function of the modality satisfies the gradient coupling condition as follows: ; Gradient coupling constraints are introduced to enhance intermodal correlation. This constraint, implemented via backpropagation, forces gradient update paths from different modes to exhibit synergistic effects in the latent space. Preferably, the coupling coefficient matrix... It satisfies the row normalization condition.

[0040] Output the modal correlation matrix.

[0041] Optionally, for the feature mapping function, a nonlinear transformation is implemented using a multilayer perceptron, with the LeakyReLU activation function chosen for the hidden layer, and the weight matrix... Update the layer using differentiable convex optimization; let... This represents the batch sample size. Denotes the weighted norm. For the first The kernel scale parameter, updated in the next iteration, is used to control the sensitivity of the distance metric in the kernel function. For the first in the batch A multimodal input sample, for the i-th multi-modal input sample in the batch , each of which contains text, speech and eye movement modalities, is a multi-modal joint mapping function that maps the samples in the original feature space to a unified latent space representation, is an iteration round counter, is a dimensional modality correlation matrix, is the number of modalities; the kernel scale parameter is dynamically adjusted according to the feature distribution, denoted as: .

[0042] The mapping function and the weight matrix are jointly trained using an alternating optimization strategy. For the mapping function , a multi-layer perceptron is used to realize the nonlinear transformation, and the LeakyReLU activation function is preferred to alleviate the gradient vanishing problem. The weight matrix is updated through a differentiable convex optimization layer to ensure its semi-definiteness and maintain the effectiveness of the distance metric.

[0043] The output modality correlation matrix provides a quantitative basis for cross-modal dependency for dynamic system parameter adjustment. Through the above process, the deep semantic alignment and collaborative representation of multi-modal features such as text, speech and eye movement are realized, providing a unified fusion feature space for subsequent cognitive state modeling.

[0044] By monitoring the difference between the predicted value and the actual observation of the cognitive state in real time, an error signal driven parameter adjustment mechanism is constructed. The cognitive state prediction error is defined as the normalized difference of the multi-modal tensor.

[0045] Optionally, the tensor decomposition parameters are dynamically adjusted according to the strategy execution result and the feature correlation degree, including: Let be the predicted cognitive state tensor, be the reconstructed predicted value of the tensor decomposition, be the L2 norm of the vector, be the cognitive state prediction error, and the cognitive state tensor be , ; The cognitive state prediction error is denoted as: ; This error term quantifies the system prediction ability and serves as the core feedback signal of the control loop. A proportional-integral-derivative control (PID control) algorithm is used to adjust the weight parameters of the strategy generation module.

[0046] Let be the teaching strategy weight of the i-th teaching cycle, be the teaching strategy weight of the i-th teaching cycle, , be the teaching strategy weight of the i-th teaching cycle, , and be the PID control parameters, be the prediction error of the i-th teaching cycle, be the accumulated variable of the teaching cycle, be the prediction error value of the i-th teaching cycle, the teaching strategy weight is adjusted by the PID controller, which is expressed as: ; ; Let be the maximum number of knowledge points allowed in the i-th teaching cycle, be the capacity adjustment step size, be the cognitive state prediction error change rate, be the sign function, be the floor function, then the dynamic adjustment of the teaching capacity constraint is expressed as: .

[0047] A double-level feedback mechanism is established: the bottom layer of fast response layer performs parameter fine-tuning once every teaching cycle, and the top layer of slow adaptation layer re-trains the cognitive state model every K cycles. When re-training, the core tensor is frozen, and only the factor matrices , and are updated to maintain the stability of the cross-modal association pattern. Preferably, the re-training cycle number K=5 to balance the model adaptability and computational overhead.

[0048] By establishing a double-level feedback mechanism, a complete "perception-decision-execution" closed-loop control loop is formed, realizing the collaborative optimization of teaching strategy parameters and system constraint thresholds, and ensuring the robustness and adaptability of the system in a dynamic teaching environment.

[0049] The present application introduces a closed-loop feedback control mechanism to dynamically adjust system parameters, compares the difference between the predicted cognitive state and the actual observation in real time, and updates the teaching strategy weight and knowledge point capacity constraint using the PID control algorithm. This adaptive adjustment mechanism enables the system to optimize the teaching plan in real time according to the learning effect, significantly improving the adaptability of the teaching system to individual differences, and achieving precise intervention and continuous optimization in the intelligent teaching process.

[0050] Example 2 ​​Based on the same principle as the method shown in embodiment 1 of the present application, an intelligent teaching auxiliary system based on a large language model is also provided in the embodiments of the present application, which comprises a multi-modal data processing unit, a cognitive state modeling unit, a teaching strategy generation unit, a multi-modal fusion unit, a feedback control unit and an update output unit; The multi-modal data processing unit is used to acquire text data, speech data and eye movement trajectory data of a student, and perform feature extraction to construct a three-dimensional cognitive state tensor model; The cognitive state modeling unit is used to construct a cognitive state tensor decomposition constraint optimization function, perform tensor decomposition on the cognitive state tensor with a constraint condition, and output a core tensor and a factor matrix; The teaching strategy generation unit is used to generate a teaching strategy according to the core tensor and the factor matrix; The multi-modal fusion unit is used to realize feature association by defining a cross-modal kernel function; The feedback control unit is used to dynamically adjust tensor decomposition parameters according to a strategy execution result and a feature association degree, to obtain an updated core tensor and factor matrix; The update output unit is used to output an updated teaching strategy according to the updated core tensor and factor matrix.

[0051] Optionally, the text data, speech data and eye movement trajectory data of the student are acquired, feature extraction is performed, and a three-dimensional cognitive state tensor model is constructed, comprising: The text feature vector is calculated according to the text data; let be the input text sequence, be the classification identifier of the model, be the semantic vector dimension, represent a real number, be a real number vector space of dimension d, then the text feature vector is represented as: ; The speech feature matrix is calculated according to the speech data; let be the speech spectrogram, be the mel cepstrum coefficient extraction method, be the cepstrum coefficient order, be a real number matrix space of dimension D, then the speech feature matrix is represented as: ; The eye movement feature function is calculated according to the eye movement trajectory data; let be the eye movement trajectory data, be the horizontal gaze coordinates at time , be the time vertical gaze coordinates, for time pupil diameter, for gaze density function, for start time of teaching segment, for end time of teaching segment, for feature dimension, for real vector space, then the eye movement feature function is represented as: ; Let the cognitive state tensor be , , total number of knowledge points, number of time slices, number of modalities, feature dimension of each modality, trainable weight parameters, feature vector of the modality at time , then the three-dimensional cognitive state tensor model is represented as: .

[0052] Optionally, a cognitive state tensor decomposition constraint optimization function is constructed, including: let be the cognitive state tensor, be the core tensor, representing cross-modality knowledge association, be the knowledge point factor matrix, be the time factor matrix, be the modality factor matrix, be the tensor mod product, be the Frobenius norm, and be regularization coefficients, be the identity matrix, then the cognitive state tensor decomposition constraint optimization function is represented as: ; Let be the knowledge point sparsity adjustment parameter, be the core tensor rank constraint value, be the diagonal elements of the extraction matrix, represent the rank of , then the constraint condition is represented as: .

[0053] Optionally, the tensor decomposition with constraints on the cognitive state tensor is performed to output a core tensor and a factor matrix, and further includes imposing an orthogonal constraint on the factor matrix; let be the first row vector of the time factor matrix , be the first row vector of the time factor matrix , be an exponential decay term, be a periodic term, be a standard Euclidean inner product, be a teaching cycle parameter, be a memory decay rate, imposing the orthogonal constraint on the factor matrix is represented as: .

[0054] Optionally, the teaching strategy is generated according to the core tensor and the factor matrix, including: constructing a mixed integer nonlinear programming model of the teaching strategy according to the core tensor and the knowledge point factor matrix; let be a teaching strategy vector, represent selecting the th knowledge point, be a knowledge point factor matrix of a current teaching cycle, be a knowledge point factor matrix of a previous teaching cycle, and be weight coefficients, be a teaching cost of the th knowledge point, be a teaching method constraint matrix, be a teaching method threshold value vector, be a maximum number of knowledge points for single teaching, be a set of conflict knowledge point pairs, be the core tensor, be a tensor modulus product, be a Frobenius norm, the mixed integer nonlinear programming model of the teaching strategy is represented as: . . . . The mixed integer nonlinear programming model of the teaching strategy is solved by using a branch and bound algorithm, and an objective function value corresponding to an optimal teaching strategy is output.

[0055] Optionally, feature association is achieved by defining a cross-modal kernel function, including: modal factor matrix and original features, define a cross-modal kernel function; Let and be a multi-modal input sample, be the number of modalities, be the feature mapping function of the th modality, be the weight matrix of the th modality, be the modal correlation matrix, , be the kernel scale parameter, be the th modality feature vector of the sample , be the cross-modal kernel function value, then the cross-modal kernel function is represented as: ; Let be the cross-modal coupling coefficient, be the gradient of the th modality mapping to the th modality input, be the gradient vector of the kernel function to the th modality input feature, be the hidden space mapping function of the th modality, and the gradient coupling condition is satisfied: ; Output the modal correlation matrix.

[0056] Optionally, for the feature mapping function, a nonlinear transformation is realized by a multilayer perception, the hidden layer activation function is selected as a LeakyReLU activation function, and the weight matrix is updated through a differentiable convex optimization layer; let be the number of batch samples, denote the weighted norm, be the kernel scale parameter after the th iteration update, used to control the sensitivity of distance measurement in the kernel function, be the th multi-modal input sample in the batch, be the th multi-modal input sample in the batch, each multi-modal input sample containing text, speech and eye movement three modal features, be a multi-modal joint mapping function that maps the sample in the original feature space to a unified hidden space representation, be an iteration round counter, be a dimensional modal correlation matrix, is the number of modalities; and is dynamically adjusted according to the feature distribution, and is expressed as: .

[0057] Optionally, the tensor decomposition parameters are dynamically adjusted according to the strategy execution result and the feature correlation degree, including: Let be the predicted cognitive state tensor, be the predicted value reconstructed by tensor decomposition, be the L2 norm of the vector, be the cognitive state prediction error, and the cognitive state tensor be , ; The cognitive state prediction error is expressed as: ; Let be the teaching strategy weight of the i-th teaching cycle, be the teaching strategy weight of the i-th teaching cycle, , be the PID control parameters, be the prediction error of the i-th teaching cycle, be the teaching cycle cumulative variable, be the prediction error value of the i-th teaching cycle, and the teaching strategy weight is adjusted by the PID controller, and is expressed as: ; Let be the prediction error of the i-th teaching cycle, and the teaching strategy weight is adjusted by the PID controller, and is expressed as: ; Let be the maximum number of knowledge points allowed in the i-th teaching cycle, be the capacity adjustment step size, be the cognitive state prediction error change rate, be the sign function, be the floor function, and the dynamic adjustment of the teaching capacity constraint is expressed as: . .

[0058] The above is only a preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.​​

Claims

1. A large language model-based intelligent teaching assistance method, characterized in that, The method comprises the following steps: Obtain text data, speech data and eye movement trajectory data of students, and perform feature extraction to construct a three-dimensional cognitive state tensor model; Construct a cognitive state tensor decomposition constraint optimization function, perform tensor decomposition on the cognitive state tensor with constraint conditions, and output core tensor and factor matrix; Generate teaching strategies according to the core tensor and the factor matrix; Realize feature association by defining a cross-modal kernel function; Dynamically adjust the tensor decomposition parameters according to the strategy execution result and the feature association degree to obtain the updated core tensor and factor matrix; Output the updated teaching strategies according to the updated core tensor and factor matrix. 2.The intelligent teaching assistance method based on a large language model of claim 1, wherein, Obtain text data, speech data and eye movement trajectory data of students, and perform feature extraction to construct a three-dimensional cognitive state tensor model, comprising: According to the text data, a text feature vector is calculated; set for the input text sequence, for the classification identifier of the model, for the semantic vector dimension, representing a real number, for a vector space of real numbers, the text feature vector is represented as: ； According to the voice data, a voice feature matrix is calculated; and is a voice spectrum graph, is a mel-frequency cepstral coefficient extraction method, is a mel-frequency cepstral coefficient order, is a mel-frequency cepstral coefficient, is a voice feature matrix, and the voice feature matrix is represented as: ； According to the eye movement trajectory data, an eye movement feature function is calculated; set for the eye movement trajectory data, for the time horizontal gaze coordinates, for the time vertical gaze coordinates, for the time pupil diameter, for the gaze density function, for the starting time of the teaching segment, for the ending time of the teaching segment, for the feature dimension, for a real vector space of dimension, then the eye movement feature function is expressed as: ； Let the cognitive state tensor be , , is the total number of knowledge points, is the number of time slices, is the number of modalities, is the dimension of each modality feature, is the trainable weight parameter, is the feature vector of the th modality at the th time slice, then the three-dimensional cognitive state tensor model is represented as: 。 3.The intelligent teaching assistance method based on a large language model of claim 1, wherein, The cognitive state tensor decomposition constraint optimization function is constructed, including: is a cognitive state tensor, is a core tensor, representing cross-modal knowledge association, is a knowledge point factor matrix, is a time factor matrix, is a modal factor matrix, is a tensor modular product, is a Frobenius norm, and is a regularization coefficient, is a unit matrix, and the cognitive state tensor decomposition constraint optimization function is represented as: ； Let be the knowledge point sparsity adjustment parameter, be the core tensor rank constraint value, be the extracted matrix diagonal element, denote the rank of the core tensor , then the constraint condition is expressed as: 。 4.The intelligent teaching assistance method based on a large language model of claim 1, wherein, The cognitive state tensor is subjected to constrained tensor decomposition, and a core tensor and a factor matrix are output, and the factor matrix is subjected to orthogonal constraint. is a time factor matrix is a first row vector of the time factor matrix, is a second row vector of the time factor matrix, is an exponential decay term, is a periodic term, is a standard Euclidean inner product, is a teaching period parameter, is a memory decay rate, and the orthogonal constraint imposed on the factor matrix is represented as: 。 5.The intelligent teaching assistance method based on a large language model of claim 1, wherein, Generate teaching strategies according to the core tensor and the factor matrix, comprising: According to the core tensor and the knowledge point factor matrix, a mixed integer nonlinear programming model of the teaching strategy is constructed; Let be the teaching strategy vector, represent the selection of the th knowledge point, be the knowledge point factor matrix of the current teaching period, be the knowledge point factor matrix of the previous teaching period, and be the weight coefficients, be the teaching cost of the th knowledge point, be the teaching method constraint matrix, be the teaching method threshold vector, be the maximum number of knowledge points for a single teaching, be the set of conflicting knowledge point pairs, be the core tensor, be the tensor mod product, be the Frobenius norm, then the mixed integer nonlinear programming model of the teaching strategy is represented as: ； ； ； ； Solve the mixed integer nonlinear programming model of the teaching strategy by using a branch and bound algorithm, and output the objective function value corresponding to the optimal teaching strategy. 6.The intelligent teaching assistance method based on a large language model of claim 1, wherein, Realize feature association by defining a cross-modal kernel function, comprising: Define a cross-modal kernel function based on the modal factor matrix and the original features; Let and be a multi-modal input sample, be the number of modalities, be the feature mapping function for the -th modality, be the weight matrix for the -th modality, be the modality correlation matrix, , be the kernel scale parameter, be the -th modality feature vector for the sample , and be the cross-modality kernel function value, then the cross-modality kernel function is represented as: ； Let be the cross-modal coupling coefficient, be the gradient of the modal mapping with respect to the modal input, be the gradient vector of the kernel function with respect to the modal input feature, be the hidden space mapping function of the modal, satisfying the gradient coupling condition ； Output the modal association matrix.

7. The intelligent teaching assistance method based on a large language model according to claim 6, characterized in that, For the feature mapping function, a nonlinear transformation is realized by a multi-layer perceptron, and the LeakyReLU activation function is selected for the hidden layer activation function, and the weight matrix is updated by differentiable convex optimization layer; set is the batch size, represents the weighted norm, is the kernel scale parameter of the iteration, used to control the sensitivity of the distance metric in the kernel function, is the multi-modal input sample in the batch, is the multi-modal input sample in the batch, each multi-modal input sample contains text, speech and eye movement three modal features, is a multi-modal joint mapping function that maps the original feature space sample to a unified hidden space representation, is the iteration round counter, is the dimensional modal correlation matrix, is the number of modalities; the kernel scale parameter is dynamically adjusted according to the feature distribution, which is represented as: 。 8.The intelligent teaching assistance method based on a large language model of claim 1, wherein, Dynamically adjust the tensor decomposition parameters according to the strategy execution result and the feature association degree, comprising: Let be the predicted cognitive state tensor, be the predicted value of the tensor decomposition reconstruction, be the L2 norm of a vector, be the cognitive state prediction error, the cognitive state tensor being , ; The cognitive state prediction error is represented as: ； Let be the teaching strategy weight for the th teaching cycle, be the teaching strategy weight for the th teaching cycle, , and be the PID control parameters, be the prediction error for the th teaching cycle, be the teaching cycle cumulative variable, be the prediction error value for the th teaching cycle, the teaching strategy weight adjusted by the PID controller is expressed as: ； Let be the maximum number of knowledge points allowed for the current teaching cycle, be the capacity adjustment step size, be the cognitive state prediction error change rate, be the sign function, be the floor function, then the dynamic adjustment of the teaching capacity constraint is represented as: 。 9.A large language model-based intelligent teaching assistance system, characterized in that, The system comprises a multi-modal data processing unit, a cognitive state modeling unit, a teaching strategy generation unit, a multi-modal fusion unit, a feedback control unit and an update output unit. The multi-modal data processing unit is used to obtain text data, speech data and eye movement trajectory data of students, and perform feature extraction to construct a three-dimensional cognitive state tensor model. The cognitive state modeling unit is used to construct a cognitive state tensor decomposition constraint optimization function, perform tensor decomposition on the cognitive state tensor with constraint conditions, and output core tensor and factor matrix. The teaching strategy generation unit is used to generate teaching strategies according to the core tensor and the factor matrix. The multi-modal fusion unit is used to realize feature association by defining a cross-modal kernel function. The feedback control unit is used to dynamically adjust the tensor decomposition parameters according to the strategy execution result and the feature association degree to obtain the updated core tensor and factor matrix. The update output unit is used to output the updated teaching strategies according to the updated core tensor and factor matrix.

Images