Knowledge graph-based interpretable teaching cognition interaction method and system, and medium

Abstract

The invention discloses an interpretable teaching cognition interaction method and system based on a knowledge graph, and a medium, and relates to the field of education management information systems, and the method specifically comprises the steps: collecting answer data of a learner, calculating a knowledge mastering state vector, mapping the knowledge mastering state vector to a knowledge graph node, and carrying out the feature decomposition; obtaining a cognitive feature matrix and a knowledge node cognitive difficulty coefficient; constructing a cognitive path graph based on the cognitive difficulty coefficient, and calculating the migration probability of adjacent nodes to generate a cognitive migration path; and constructing a multi-layer knowledge network on the path, calculating node weights and knowledge circulation coefficients, generating a teaching knowledge point sequence, and encoding the sequence to generate stepped teaching content. According to the method, personalized content recommendation and scientific allocation of teaching resources can be provided for an education platform, and intelligent optimization and management transparency of the teaching process are realized.

Description

Technical Field

[0001] This invention relates to the field of educational management information systems, specifically to interpretable teaching cognitive interaction methods, systems, and media based on knowledge graphs. Background Technology

[0002] With the rapid development of online education and smart learning platforms, how to achieve differentiated teaching and efficient resource allocation among a large user base has become a crucial issue. Existing teaching recommendation methods largely rely on static question banks or simple accuracy statistics, lacking dynamic analysis and interpretation mechanisms for learners' cognitive states. This results in insufficient transparency in personalized recommendations, inefficient allocation of teaching content, and hinders the rational management of educational resources and the improvement of learning outcomes. Summary of the Invention

[0003] The purpose of this invention is to provide an interpretable teaching cognitive interaction method, system, and medium based on knowledge graphs, which enables dynamic modeling of learners' cognitive states, improves the rationality of teaching content allocation, and makes the education process more transparent and efficient.

[0004] The knowledge graph-based interpretable teaching cognitive interaction method provided in this embodiment of the invention includes the following steps:

[0005] Collect learners' response data and calculate the knowledge mastery state vector based on the response data;

[0006] The knowledge mastery state vector is mapped to knowledge graph nodes and feature decomposition is performed to obtain the cognitive feature matrix. The cognitive difficulty coefficient of the knowledge node is calculated based on the cognitive feature matrix.

[0007] A cognitive path graph is constructed based on the cognitive difficulty coefficient. The node migration probability between adjacent knowledge nodes is calculated in the cognitive path graph to generate a cognitive migration path.

[0008] A multi-layered knowledge network is constructed along the cognitive transfer path. The importance weight and knowledge flow coefficient of the nodes in the multi-layered knowledge network are calculated to generate a sequence of teaching knowledge points.

[0009] The sequence of teaching knowledge points is encoded to generate tiered teaching content.

[0010] Furthermore, learners' response data is collected, and a knowledge mastery state vector is calculated based on this data, including:

[0011] Collect learners' response data and perform time-series segmentation to generate a time-series response sequence;

[0012] The temporal response sequence is decomposed using wavelet decomposition to extract multi-scale fluctuation features, and the cognitive state value is calculated based on the multi-scale fluctuation features.

[0013] The forgetting compensation curve is adjusted using the cognitive state value. When the cognitive state value increases, the compensation coefficient is decreased, and when the cognitive state value decreases, the compensation coefficient is increased. The compensation coefficient is then multiplied by the answer data to obtain the corrected answer value.

[0014] A fluctuation synchronicity analysis is performed on the corrected answer values ​​to calculate the cognitive correlation between knowledge points. A knowledge correlation matrix is ​​constructed based on the cognitive correlation, and knowledge groups with synchronous change characteristics are identified from the knowledge correlation matrix.

[0015] The cognitive relevance of the knowledge group and the cognitive state value are combined to generate a knowledge mastery state vector.

[0016] Furthermore, the knowledge mastery state vector is mapped to knowledge graph nodes and feature decomposition is performed to obtain the cognitive feature matrix, including:

[0017] Based on the hierarchical relationship and degree of association between knowledge points in the knowledge graph, the mapping weight between nodes is calculated, and the knowledge mastery state vector is mapped to the knowledge graph node according to the mapping weight to obtain the node state value.

[0018] The node state values ​​are subjected to spectral decomposition to obtain a sequence of feature values ​​and a set of feature vectors;

[0019] Based on the distribution pattern of the feature value sequence and the hierarchical relationship between nodes, the weight coefficient of the cognitive dimension is calculated, and the feature vector group is combined with the weight coefficient to obtain the cognitive feature matrix.

[0020] Furthermore, the cognitive difficulty coefficient of knowledge nodes calculated based on the cognitive feature matrix includes:

[0021] The cognitive feature matrix is ​​normalized, and the degree of dispersion of each dimension in the normalized cognitive feature matrix is ​​calculated. The degree of dispersion is then converted into dimension weights.

[0022] The normalized cognitive feature matrix is ​​weighted using the aforementioned dimensional weights to obtain the weighted cognitive features of knowledge nodes.

[0023] Calculate the principal component values ​​of the weighted cognitive features, and convert the principal component values ​​into cognitive complexity;

[0024] Based on the ratio of the positive to the negative components of the principal component value, the cognitive complexity is compensated and corrected to generate the cognitive difficulty coefficient of the knowledge node. The positive component represents the degree to which the knowledge node facilitates understanding, while the negative component represents the degree to which the knowledge node hinders understanding.

[0025] Furthermore, a cognitive path graph is constructed based on the cognitive difficulty coefficient. The node transfer probability between adjacent knowledge nodes is calculated within this cognitive path graph, generating cognitive transfer paths including:

[0026] The cognitive difficulty coefficient is mapped to an interval to obtain the node weight value, and a cognitive node weighting matrix is ​​constructed based on the node weight value;

[0027] The cognitive distance between nodes is calculated using a cognitive node weighting matrix. The cognitive distance is combined with the node weight value to generate the path connection strength. A directed weighted cognitive path graph is then constructed based on the path connection strength.

[0028] Extract node topological features from the directed weighted cognitive path graph, fuse node topological features with path connection strength, and calculate cognitive jump probability;

[0029] A state transition matrix is ​​constructed using cognitive jump probabilities. Markov chain analysis is performed on the state transition matrix to obtain steady-state transition paths. A node transition probability matrix is ​​then generated based on the steady-state transition paths.

[0030] Random walk sampling is performed based on the node migration probability matrix. The sampled paths are scored and ranked, and the path with the highest score is selected as the cognitive transfer path.

[0031] Furthermore, a multi-layered knowledge network is constructed along the cognitive transfer path, and the importance weights and knowledge flow coefficients of the nodes in the multi-layered knowledge network are calculated to generate a sequence of teaching knowledge points, including:

[0032] In the cognitive transfer path, knowledge points are mapped to different cognitive levels according to their cognitive difficulty coefficients. The original cognitive transfer relationship is maintained within the same cognitive level, and vertical connections are established between different cognitive levels to build a multi-layered knowledge network.

[0033] The degree of association is obtained by calculating the number of connections between nodes and adjacent nodes in the multi-layer knowledge network. The mediation ability is calculated based on the number of times a node is transmitted in the cognitive transfer path. The importance weight of a node is determined according to the degree of association and the mediation ability.

[0034] Based on the nodes and multi-layered knowledge network in the cognitive transfer path, the knowledge flow coefficient representing the nodes is calculated;

[0035] Based on the importance weights and knowledge flow coefficients, nodes are sorted and organized in the multi-layered knowledge network to generate a sequence of teaching knowledge points.

[0036] Furthermore, the sequence of teaching knowledge points is hierarchically coded, and tiered teaching content is generated based on this hierarchical coding, including:

[0037] For each knowledge point in the sequence of teaching knowledge points, a hierarchical code is generated based on its cognitive level and teaching order. The hierarchical code includes a level number and a sequence number.

[0038] Based on the hierarchical coding analysis, the difference in hierarchical numbers between adjacent knowledge points is analyzed. When the hierarchical number of adjacent knowledge points changes, the knowledge point at the changed position is identified as the cognitive leap point, and the cognitive span is obtained by calculating the difference in hierarchical numbers at the cognitive leap point.

[0039] Cognitive importance is calculated by statistically analyzing the number of times knowledge points are transmitted and the number of connections in the cognitive transfer path, and cognitive gap value is calculated based on cognitive importance and cognitive span.

[0040] Based on the cognitive gap value, transitional teaching content is configured at the cognitive leap point to generate tiered teaching content with hierarchical coding.

[0041] This invention provides an interpretable instructional cognitive interaction system based on knowledge graphs, the system comprising:

[0042] The data acquisition module collects learners' response data and calculates a knowledge mastery state vector based on the response data.

[0043] The feature extraction module maps the knowledge mastery state vector to knowledge graph nodes and performs feature decomposition to obtain the cognitive feature matrix. Based on the cognitive feature matrix, the cognitive difficulty coefficient of the knowledge node is calculated.

[0044] The cognitive path construction module constructs a cognitive path graph based on the cognitive difficulty coefficient, calculates the node migration probability between adjacent knowledge nodes in the cognitive path graph, and generates a cognitive migration path.

[0045] The knowledge network construction module constructs a multi-layer knowledge network along the cognitive transfer path, calculates the importance weight and knowledge flow coefficient of the nodes in the multi-layer knowledge network, and generates a sequence of teaching knowledge points.

[0046] The teaching content generation module encodes the sequence of teaching knowledge points to generate tiered teaching content.

[0047] One technical solution provided in this embodiment of the invention is an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps in any of the aforementioned methods.

[0048] One technical solution provided in this embodiment of the invention is a computer-readable storage medium storing computer program instructions, which, when executed by a processor, implement the steps in any of the aforementioned methods.

[0049] This invention enables dynamic modeling of learners' cognitive states and transforms this model into a basis for allocating teaching resources and optimizing teaching processes, thereby enhancing the scientific rigor and transparency of educational platforms in personalized recommendations and content scheduling. By introducing knowledge graphs for cognitive difficulty modeling and transfer path calculation, the platform can achieve precise delivery of teaching content in a large-scale user environment, avoiding the shortcomings of traditional methods such as coarse resource allocation and lack of interpretability. Simultaneously, the teaching knowledge point sequences generated through multi-layered knowledge networks support hierarchical management and intelligent supervision of the educational process, improving the utilization rate of teaching resources and learning efficiency. Ultimately, this achieves personalized, intelligent, and efficient educational services, enhancing learner experience and platform management capabilities.

[0050] The above description of the invention is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 A flowchart illustrating the interpretable teaching cognitive interaction method based on knowledge graphs provided in this embodiment of the invention;

[0053] Figure 2 A flowchart illustrating the construction of a cognitive transfer path provided in an embodiment of the present invention;

[0054] Figure 3 This is a schematic diagram of the structure of an interpretable teaching cognitive interaction system based on knowledge graphs, provided in an embodiment of the present invention. Detailed Implementation

[0055] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. In the following description relating to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements.

[0056] The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with one or more embodiments of this specification. It should be noted that the steps of the corresponding methods are not necessarily performed in the order shown and described in this specification in other embodiments. In some other embodiments, the methods may include more or fewer steps than described in this specification. Furthermore, a single step described in this specification may be broken down into multiple steps in other embodiments; and multiple steps described in this specification may be combined into a single step in other embodiments.

[0057] like Figure 1 As shown, Figure 1 A flowchart of a knowledge graph-based interpretable teaching cognitive interaction method provided in an embodiment of the present invention, the method comprising the following steps:

[0058] Collect learners' response data and calculate the knowledge mastery state vector based on the response data;

[0059] The knowledge mastery state vector is mapped to knowledge graph nodes and feature decomposition is performed to obtain the cognitive feature matrix. The cognitive difficulty coefficient of the knowledge node is calculated based on the cognitive feature matrix.

[0060] A cognitive path graph is constructed based on the cognitive difficulty coefficient. The node migration probability between adjacent knowledge nodes is calculated in the cognitive path graph to generate a cognitive migration path.

[0061] A multi-layered knowledge network is constructed along the cognitive transfer path. The importance weight and knowledge flow coefficient of the nodes in the multi-layered knowledge network are calculated to generate a sequence of teaching knowledge points.

[0062] The sequence of teaching knowledge points is encoded to generate tiered teaching content.

[0063] Furthermore, learners' response data is collected, and a knowledge mastery state vector is calculated based on this data, including:

[0064] Collect learners' response data and perform time-series segmentation to generate a time-series response sequence;

[0065] The temporal response sequence is decomposed using wavelet decomposition to extract multi-scale fluctuation features, and the cognitive state value is calculated based on the multi-scale fluctuation features.

[0066] The forgetting compensation curve is adjusted using the cognitive state value. When the cognitive state value increases, the compensation coefficient is decreased, and when the cognitive state value decreases, the compensation coefficient is increased. The compensation coefficient is then multiplied by the answer data to obtain the corrected answer value.

[0067] A fluctuation synchronicity analysis is performed on the corrected answer values ​​to calculate the cognitive correlation between knowledge points. A knowledge correlation matrix is ​​constructed based on the cognitive correlation, and knowledge groups with synchronous change characteristics are identified from the knowledge correlation matrix.

[0068] The cognitive relevance of the knowledge group and the cognitive state value are combined to generate a knowledge mastery state vector.

[0069] In this implementation, learners' answer data during the learning process is first collected and then segmented according to chronological order. The answer data includes information such as the learner's correctness of answers to questions related to the knowledge points, the time taken to answer, and the answering process. Chronological segmentation refers to dividing continuous answer data into multiple temporal segments according to preset time windows (such as weekly, daily, or per learning unit), forming a temporal answer sequence. For example, for a learner's answers over three consecutive weeks in a mathematics course, the data can be segmented by week, resulting in three temporal answer sequences.

[0070] The acquired time-series response sequence was processed using wavelet decomposition to extract multi-scale fluctuation features. Wavelet decomposition is the process of breaking down the response sequence into different frequency components, which can capture the changing patterns of the response data at different time scales. Discrete wavelet transform was used to decompose the time-series response sequence into approximation coefficients and detail coefficients. The approximation coefficients reflect the overall trend of the response sequence, while the detail coefficients reflect the fluctuation characteristics of the response sequence at different scales. Based on the extracted multi-scale fluctuation features, a cognitive state value was calculated. The cognitive state value is a quantitative indicator that measures the learner's mastery of knowledge points, and its calculation takes into account factors such as fluctuations in response accuracy, changes in response time, and the stability of response patterns.

[0071] The calculated cognitive state value is used to adjust the forgetting compensation curve. Based on the Ebbinghaus forgetting curve principle, the forgetting compensation curve simulates the learner's forgetting process. When the cognitive state value shows an upward trend, it indicates that the learner's understanding of the knowledge point is improving, so the compensation coefficient is reduced to decrease compensation for forgetting factors. Conversely, when the cognitive state value shows a downward trend, it indicates that the learner's mastery of the knowledge point is weakening, so the compensation coefficient is increased to increase compensation for forgetting factors. The adjustment range of the compensation coefficient is typically between 0.8 and 1.2, with the specific value determined based on the magnitude of change in the cognitive state value. Multiplying the compensation coefficient by the original answer data yields the corrected answer value, a process that more accurately reflects the learner's actual cognitive state.

[0072] A fluctuation synchronicity analysis was performed on the corrected response values ​​to calculate the cognitive correlation between different knowledge points. Fluctuation synchronicity analysis identifies potential cognitive correlations between knowledge points by examining the coordinated change patterns in response performance across different knowledge points. For example, when learners' performance on "fraction addition and subtraction" and "fraction multiplication" shows synchronous changes, it may indicate a close connection between these two knowledge points in the learner's cognitive structure. A sliding window correlation analysis method was used to calculate the correlation coefficients of response performance across different knowledge points, generating a knowledge correlation matrix. The knowledge correlation matrix is ​​a square matrix where each element represents the cognitive correlation between corresponding knowledge points, ranging from -1 to 1, with larger values ​​indicating stronger correlations.

[0073] This study identifies knowledge groups exhibiting synchronous changes from the knowledge association matrix. A knowledge group refers to a set of closely linked knowledge points in a learner's cognitive structure, which show synchronous changes during the learning process. Cluster analysis is used to group knowledge points with a cognitive association higher than a preset threshold (e.g., 0.7) into the same knowledge group. For a given learner, knowledge groups such as "area calculation of geometric figures," "perimeter calculation," and "proportional relationships" might be identified, indicating that these knowledge points are closely related in the learner's cognitive structure.

[0074] The cognitive relevance of identified knowledge groups is combined with cognitive state values ​​to generate a knowledge mastery state vector. This vector is a multi-dimensional feature representation describing a learner's mastery of the knowledge system, including information on the degree of mastery of knowledge points, the strength of connections between knowledge points, and the stability of cognitive states. This vector can be represented as an n-dimensional feature vector, where n is the number of relevant knowledge features. Each component of the vector corresponds to a specific knowledge feature, such as the degree of mastery of a particular knowledge point or the strength of the connection between a pair of knowledge points.

[0075] Through the above technical solution, this implementation method can achieve dynamic and accurate assessment of learners' cognitive states, capture subtle characteristics of changes in learners' cognition, and reveal the intrinsic connections between knowledge points, thereby providing data support for personalized teaching. This not only improves the accuracy and scientific rigor of teaching assessment but also provides teachers with interpretable teaching decision-making basis, helping them accurately identify learners' knowledge weaknesses and cognitive obstacles, formulate targeted teaching strategies, and ultimately achieve a significant improvement in teaching effectiveness and optimization of the learning experience.

[0076] Furthermore, the knowledge mastery state vector is mapped to knowledge graph nodes and feature decomposition is performed to obtain the cognitive feature matrix, including:

[0077] Based on the hierarchical relationship and degree of association between knowledge points in the knowledge graph, the mapping weight between nodes is calculated, and the knowledge mastery state vector is mapped to the knowledge graph node according to the mapping weight to obtain the node state value.

[0078] The node state values ​​are subjected to spectral decomposition to obtain a sequence of feature values ​​and a set of feature vectors;

[0079] Based on the distribution pattern of the feature value sequence and the hierarchical relationship between nodes, the weight coefficient of the cognitive dimension is calculated, and the feature vector group is combined with the weight coefficient to obtain the cognitive feature matrix.

[0080] In educational cognitive assessment, a knowledge graph is a network model describing the structure of a knowledge system, consisting of nodes representing knowledge points and edges representing the relationships between knowledge points. Knowledge graphs typically contain multi-level knowledge point structures. For example, in a middle school mathematics knowledge graph, "equation" might be a first-level knowledge point, with second-level knowledge points such as "linear equation in one variable" and "system of linear equations in two variables." These second-level knowledge points may further relate to knowledge points such as "solving equations" and "word problems." Based on this structured knowledge representation, learners' knowledge mastery state vectors can be mapped to corresponding nodes in the knowledge graph, enabling deeper extraction of cognitive features.

[0081] In this implementation, the mapping weights between nodes are first calculated based on the hierarchical relationships and correlations between knowledge points in the knowledge graph. The mapping weights reflect the strength and influence of the correlation between different knowledge points and are the foundation for mapping knowledge mastery state vectors to knowledge graph nodes. The calculation of mapping weights considers factors such as the hierarchical relationships between knowledge points, the prerequisite relationships between knowledge points, and the importance of knowledge points in the subject system. For example, for a pair of knowledge points A and B in the knowledge graph, if A is a prerequisite knowledge point of B, the mapping weight from A to B may be higher than the mapping weight from B to A; if knowledge point C is a core node connecting multiple knowledge points, the mapping weight of C may be relatively high. Mapping weights are usually represented in matrix form, where each element represents the mapping weight between corresponding knowledge points, typically ranging from 0 to 1, with larger values ​​indicating stronger correlations.

[0082] Based on the calculated mapping weights, the knowledge mastery state vector is mapped onto knowledge graph nodes, yielding node state values. The knowledge mastery state vector contains information about the learner's mastery of each knowledge point. Through calculation with the mapping weight matrix, the state value of each node in the knowledge graph can be obtained. Node state values ​​not only reflect the learner's mastery of the directly corresponding knowledge point but also consider the mastery of related knowledge points and their impact. For example, if a learner has a high level of mastery of "fraction multiplication," and there is a strong correlation between "fraction multiplication" and "fraction division," the state value of the "fraction division" node may also increase accordingly. The calculation process of node state values ​​can be understood as distributing and propagating the learner's knowledge mastery across the knowledge graph, forming a state distribution covering the entire knowledge graph.

[0083] Spectral decomposition is performed on the node state values ​​to obtain the eigenvalue sequence and eigenvector set. Spectral decomposition is a matrix factorization technique that decomposes the node state value matrix into a combination of eigenvalues ​​and eigenvectors. The node state value matrix can be viewed as a signal distribution on a knowledge graph, and spectral decomposition can identify the main patterns and variation characteristics in this distribution. In the spectral decomposition process, the Laplacian matrix of the node state values ​​is first constructed, and then the eigenvalues ​​and eigenvectors of this matrix are solved. The eigenvalue sequence is arranged in descending order, and each eigenvalue corresponds to an eigenvector. The larger the eigenvalue, the greater the contribution of the corresponding eigenvector to the overall distribution. The eigenvector can be understood as the basic pattern or dimension of the learner's cognitive state; different eigenvectors may correspond to different types of cognitive abilities or learning styles.

[0084] The weight coefficients of cognitive dimensions are calculated based on the distribution pattern of the feature value sequence and the hierarchical relationship between nodes. The distribution pattern of the feature value sequence reflects the importance of different cognitive dimensions in the learner's overall cognitive state, while the hierarchical relationship between nodes provides constraints on the knowledge structure. By comprehensively considering these two factors, appropriate weight coefficients can be assigned to each cognitive dimension (i.e., feature vector). Normalization is typically used in the calculation of weight coefficients so that the sum of all weight coefficients is 1. For example, if the first three feature values ​​in a learner's feature value sequence account for 80% of the total, and the corresponding feature vectors are mainly related to "logical reasoning," "calculation ability," and "spatial imagination," then these three cognitive dimensions will receive higher weight coefficients.

[0085] The cognitive feature matrix is ​​obtained by combining the feature vector groups with weight coefficients. The cognitive feature matrix is ​​a structured representation of a learner's cognitive state, where rows correspond to different knowledge points or cognitive dimensions, columns correspond to different feature vectors, and matrix element values ​​represent the weighted contribution of the feature vectors to their respective dimensions. Analysis of the cognitive feature matrix can identify the main characteristics, strengths, and weaknesses of a learner's cognitive state. For example, matrix analysis might show that a learner performs strongly in the "abstract thinking" dimension but is relatively weak in the "application ability" dimension, providing a clear direction for subsequent personalized instruction.

[0086] Through the above technical solution, this implementation method can integrate discrete knowledge point mastery states into a unified knowledge graph framework, revealing the inherent connections between knowledge points. Using spectral decomposition technology, it can extract key cognitive features and patterns from complex node state distributions. Based on feature value distribution and knowledge hierarchy relationships, it can determine the relative importance of different cognitive dimensions, achieving a quantitative expression of cognitive features. This method not only improves the accuracy and dimensionality of teaching assessment but also provides a scientific basis for personalized teaching decisions. Ultimately, it helps teachers more accurately identify learners' cognitive characteristics, formulate more targeted teaching strategies, and improve overall teaching effectiveness and learning efficiency.

[0087] Furthermore, the cognitive difficulty coefficient of knowledge nodes calculated based on the cognitive feature matrix includes:

[0088] The cognitive feature matrix is ​​normalized, and the degree of dispersion of each dimension in the normalized cognitive feature matrix is ​​calculated. The degree of dispersion is then converted into dimension weights.

[0089] The normalized cognitive feature matrix is ​​weighted using the aforementioned dimensional weights to obtain the weighted cognitive features of knowledge nodes.

[0090] Calculate the principal component values ​​of the weighted cognitive features, and convert the principal component values ​​into cognitive complexity;

[0091] Based on the ratio of the positive to the negative components of the principal component value, the cognitive complexity is compensated and corrected to generate the cognitive difficulty coefficient of the knowledge node. The positive component represents the degree to which the knowledge node facilitates understanding, while the negative component represents the degree to which the knowledge node hinders understanding.

[0092] In the process of educational cognitive assessment, the cognitive feature matrix carries learners' cognitive state information about each node in the knowledge graph. Each row in the matrix represents a knowledge node, and each column represents a cognitive dimension. The matrix element values ​​represent the feature strength of the knowledge node in the corresponding cognitive dimension. To eliminate differences in units and numerical ranges between different cognitive dimensions, the cognitive feature matrix needs to be normalized. Normalization uses a linear mapping method to map the value of each cognitive dimension to a range of 0 to 1, preserving the relative relationships within the dimension. For example, for the j-th column of the cognitive feature matrix (representing the j-th cognitive dimension), each element in that column can be subtracted from its minimum value, and then divided by the difference between the maximum and minimum values ​​in that column to obtain the normalized value. After this processing, the data from different cognitive dimensions are comparable, facilitating subsequent analysis.

[0093] The dispersion of each dimension of the normalized cognitive feature matrix is ​​calculated. Dispersion reflects the differences in learners' performance on that cognitive dimension and is an important basis for determining dimension weights. Dispersion can be measured by calculating the standard deviation or variance of the data for each cognitive dimension. The larger the standard deviation, the greater the differences in learners' performance on that dimension, and the more valuable that dimension is for distinguishing the cognitive difficulty of different knowledge nodes. When converting dispersion into dimension weights, a relative proportional relationship is usually used, so that dimensions with greater dispersion receive higher weights, and the sum of all dimension weights is 1. For example, assuming there are three cognitive dimensions with standard deviations of 0.2, 0.3, and 0.5, their weights might be 0.2, 0.3, and 0.5, respectively.

[0094] The weighted cognitive features of knowledge nodes are obtained by weighting the normalized cognitive feature matrix using the calculated dimensional weights. The weighting process involves multiplying the feature value of each cognitive dimension by its corresponding dimensional weight, and then summing the weighted values ​​for all dimensions. This weighting process highlights dimensions more valuable for judging cognitive difficulty while weakening dimensions with less influence, resulting in a more accurate final difficulty assessment. For example, if a knowledge node has normalized feature values ​​of 0.8, 0.6, and 0.4 in the dimensions of "abstract thinking," "memory ability," and "application ability," respectively, and the weights of these three dimensions are 0.5, 0.3, and 0.2, then the weighted cognitive feature of this knowledge node is 0.8 × 0.5 + 0.6 × 0.3 + 0.4 × 0.2 = 0.68.

[0095] The principal component values ​​of weighted cognitive features are calculated and then converted into cognitive complexity. Principal component analysis (PCA) is a dimensionality reduction technique that extracts the most representative components from multidimensional features. In this method, PCA is performed on the weighted cognitive features to obtain the principal component values ​​of knowledge nodes. These principal component values ​​reflect the comprehensive performance of knowledge nodes across the main cognitive dimensions and are the basis for assessing cognitive complexity. The calculation of principal component values ​​involves first constructing the covariance matrix of the weighted cognitive features, then solving for the eigenvectors and eigenvalues ​​of this matrix. The eigenvector with the largest eigenvalue is the first principal component. The weighted cognitive features are then projected onto the first principal component to obtain the principal component values. Principal component values ​​are typically converted into cognitive complexity through linear or nonlinear mappings. The mapping relationship can be determined based on experience in educational measurement and data analysis results. For example, principal component values ​​can be linearly mapped to a complexity range of 1-10, so that the larger the principal component value, the higher the corresponding cognitive complexity.

[0096] Based on the ratio of the positive to the negative components of the principal component value, cognitive complexity is compensated and corrected to generate the cognitive difficulty coefficient of the knowledge node. The principal component value can be decomposed into positive and negative components. The positive component represents the degree to which the knowledge node facilitates understanding, i.e., the positive impact of the knowledge node on the learner's comprehension of related knowledge; the negative component represents the degree to which the knowledge node hinders understanding, i.e., the obstacles the knowledge node may pose to the learner's comprehension. The calculation of the positive and negative components is based on the decomposition of the principal component and semantic analysis of cognitive features, typically requiring the combination of educational experts' knowledge and statistical analysis of large amounts of learning data. The ratio of the positive to the negative components directly affects the actual difficulty of the knowledge node; a larger ratio indicates a more significant facilitating effect and lower difficulty; a smaller ratio indicates a more significant hindering effect and higher difficulty. During the compensation and correction process, methods such as the reciprocal or logarithmic transformation of the ratio are typically used to convert the ratio relationship into a compensation coefficient, which is then multiplied by the cognitive complexity to obtain the final cognitive difficulty coefficient. For example, assuming the cognitive complexity of a certain knowledge node is 7 and the ratio of positive to negative components is 0.8, the possible compensation coefficient is 1 / 0.8 = 1.25, and the final cognitive difficulty coefficient is 7 × 1.25 = 8.75.

[0097] The above technical solution enables assessment based on learners' actual cognitive data, avoiding biases from subjective judgment. Weighting by dimensional dispersion highlights more discriminative cognitive dimensions, improving assessment accuracy. Introducing the analysis of positive and negative components considers the dual facilitating and hindering effects of knowledge nodes, making the assessment results more consistent with actual learning situations. This difficulty assessment method based on cognitive feature matrices provides precise knowledge difficulty references for personalized teaching, helping teachers rationally arrange teaching content and pace, and develop adaptive learning paths for learners, ultimately optimizing teaching effectiveness and improving learning efficiency.

[0098] Furthermore, a cognitive path graph is constructed based on the cognitive difficulty coefficient. The node transfer probability between adjacent knowledge nodes is calculated within this cognitive path graph, generating cognitive transfer paths including:

[0099] The cognitive difficulty coefficient is mapped to an interval to obtain the node weight value, and a cognitive node weighting matrix is ​​constructed based on the node weight value;

[0100] The cognitive distance between nodes is calculated using a cognitive node weighting matrix. The cognitive distance is combined with the node weight value to generate the path connection strength. A directed weighted cognitive path graph is then constructed based on the path connection strength.

[0101] Extract node topological features from the directed weighted cognitive path graph, fuse node topological features with path connection strength, and calculate cognitive jump probability;

[0102] A state transition matrix is ​​constructed using cognitive jump probabilities. Markov chain analysis is performed on the state transition matrix to obtain steady-state transition paths. A node transition probability matrix is ​​then generated based on the steady-state transition paths.

[0103] Random walk sampling is performed based on the node migration probability matrix. The sampled paths are scored and ranked, and the path with the highest score is selected as the cognitive transfer path.

[0104] like Figure 2 The diagram illustrates the cognitive transfer path construction process of this embodiment.

[0105] To convert the cognitive difficulty coefficient into node weight values ​​suitable for network analysis, interval mapping is required. Interval mapping uses piecewise linear or nonlinear functions to map the difficulty coefficient to an appropriate weight interval. For example, an inverse proportional relationship can be used, so that knowledge nodes with higher difficulty coefficients receive smaller weight values, reflecting the negative correlation between learning difficulty and learning preference. Specifically, the node weight value can be set to 1 minus the ratio of the difficulty coefficient to the maximum difficulty, so that nodes with lower difficulty receive higher weights and nodes with higher difficulty receive lower weights. If the cognitive difficulty coefficient of a "linear equation in one variable" in a knowledge graph is 3 and the maximum difficulty is 10, then its node weight value is 1 - (3 / 10) = 0.7. Based on the calculated node weight values, a cognitive node weighting matrix is ​​constructed. This matrix is ​​a square matrix, where the rows and columns correspond to the nodes in the knowledge graph, and the matrix elements represent the weight combination value of the corresponding node pairs, which can be calculated by multiplying or weighting the node weights.

[0106] Cognitive distance between nodes is calculated using a weighted matrix of cognitive nodes. Cognitive distance is a measure of the similarity or correlation between two knowledge nodes in the cognitive space; the smaller the distance, the stronger the cognitive correlation between the nodes. Cognitive distance can be calculated based on the element values ​​of the weighted matrix, combined with the topological relationships between nodes in the knowledge graph. Commonly used calculation methods include Euclidean distance, Manhattan distance, or cosine similarity; the choice of method depends on the data characteristics and application requirements.

[0107] The calculated cognitive distance is combined with the node weight values ​​to generate path connection strength. Path connection strength reflects the tightness of the connection between knowledge nodes and is the foundation for constructing a directed weighted cognitive path graph. Generally, path connection strength is inversely proportional to cognitive distance and directly proportional to node weight values. For example, path connection strength can be expressed as the sum of node weight values ​​divided by the cognitive distance, so that node pairs with high weights and close distances receive greater connection strength. A directed weighted cognitive path graph is constructed based on the path connection strength. In this graph, nodes represent knowledge points, edges represent the connection relationships between nodes, edge weights represent path connection strength, and edge directions represent cognitive flow or learning order.

[0108] Node topological features are extracted from a directed weighted cognitive path graph. Node topological features are a set of features describing the position and connection patterns of nodes in the graph structure, including degree, centrality, and clustering coefficient. Node degree is the number of edges directly connected to a node, which can be divided into in-degree and out-degree; centrality measures the importance of a node in the network, and commonly used centrality indicators include degree centrality, proximity centrality, and betweenness centrality; clustering coefficient reflects the tightness of connections between a node's neighbors. The extracted topological features can be represented in vector form, with each dimension of the vector corresponding to a feature. Node topological features are fused with path connection strength to calculate the cognitive jump probability. Feature fusion is the process of combining multiple features into a unified representation, which can be achieved using methods such as weighted summation, feature concatenation, or nonlinear transformation. The fused features are used to calculate the cognitive jump probability, i.e., the likelihood of a learner jumping from one knowledge node to another. The cognitive jump probability is directly proportional to the path connection strength and inversely proportional to the difficulty coefficient of the target node, reflecting the natural flow and preferences in the learning process.

[0109] A state transition matrix is ​​constructed using cognitive jump probabilities. The state transition matrix is ​​a square matrix where rows and columns correspond to nodes in the knowledge graph. It satisfies the constraint that the sum of all jump probabilities from any node is 1. Markov chain analysis is then performed on the state transition matrix to obtain steady-state transition paths. Markov chain analysis is a method for studying the long-term behavior of stochastic processes. By iteratively iterating the state transition matrix, the steady-state distribution of the system can be obtained. Specifically, it can start from any initial state vector and repeatedly right-multiply the state transition matrix until the result converges to the steady-state distribution. The steady-state distribution reflects the probability distribution of learners visiting various knowledge nodes during long-term learning, and the steady-state transition path is the most probable learning path constructed based on this distribution. A node transition probability matrix is ​​generated based on the steady-state transition path, describing the probability of a learner migrating from one knowledge node to another.

[0110] Random walk sampling is performed based on the node migration probability matrix. The sampled paths are then scored and ranked, and the path with the highest score is selected as the cognitive transfer path. Random walk sampling is a path generation method based on a probabilistic model. Starting from the initial node, the next movement direction is determined according to the node migration probability matrix until a termination condition is met. Through multiple random walks, a large number of possible learning path samples can be obtained. These sampled paths are scored and ranked, and the scoring criteria can include factors such as path length, average weight of nodes on the path, and path difficulty gradient. For example, paths with moderate total length, gradually increasing difficulty, and a large sum of node weights can be prioritized. The path with the highest score is selected as the final cognitive transfer path, serving as the optimal learning sequence recommended to the learner.

[0111] Through the aforementioned technical solution, this implementation can dynamically construct personalized learning paths based on learners' cognitive characteristics and the difficulty features of knowledge nodes, thereby optimizing learning efficiency. This method considers the cognitive connections and difficulty gradients between knowledge nodes, ensuring the coherence and progressiveness of the learning path; it introduces Markov chain analysis, giving the generated path theoretical stability and predictability; and it employs random walk sampling and multi-dimensional scoring mechanisms to enhance the flexibility and adaptability of path selection. This knowledge graph-based cognitive transfer path generation method provides core algorithmic support for intelligent education systems, recommending the most suitable knowledge learning sequence based on learners' cognitive states and learning goals, significantly improving learning effectiveness and experience.

[0112] Furthermore, a multi-layered knowledge network is constructed along the cognitive transfer path, and the importance weights and knowledge flow coefficients of the nodes in the multi-layered knowledge network are calculated to generate a sequence of teaching knowledge points, including:

[0113] In the cognitive transfer path, knowledge points are mapped to different cognitive levels according to their cognitive difficulty coefficients. The original cognitive transfer relationship is maintained within the same cognitive level, and vertical connections are established between different cognitive levels to build a multi-layered knowledge network.

[0114] The degree of association is obtained by calculating the number of connections between nodes and adjacent nodes in the multi-layer knowledge network. The mediation ability is calculated based on the number of times a node is transmitted in the cognitive transfer path. The importance weight of a node is determined according to the degree of association and the mediation ability.

[0115] Based on the nodes and multi-layered knowledge network in the cognitive transfer path, the knowledge flow coefficient representing the nodes is calculated;

[0116] Based on the importance weights and knowledge flow coefficients, nodes are sorted and organized in the multi-layered knowledge network to generate a sequence of teaching knowledge points.

[0117] Specifically, based on the generated cognitive transfer paths, knowledge points are mapped to different cognitive levels according to their cognitive difficulty coefficients, constructing a multi-layered knowledge network. Cognitive level division is typically based on a range of cognitive difficulty coefficients, with 3 to 5 levels. For example, for knowledge points with a cognitive difficulty coefficient range of 0-10, they can be divided into three cognitive levels: basic (0-3), intermediate (3-6), and advanced (6-10). Within the same cognitive level, the original cognitive transfer relationships between knowledge points are maintained, i.e., the established connections in the cognitive transfer path are preserved. Simultaneously, vertical connections are established between different cognitive levels, forming a hierarchical knowledge structure. The establishment of vertical connections is based on the prerequisite relationships and cognitive dependencies between knowledge points, typically pointing from lower-level knowledge points to higher-level knowledge points, representing the progression direction of the learning path. The weight of vertical connections can be determined based on the difficulty differences and correlation strength between knowledge points at different levels. For example, if the "linear equation in one variable" in the basic level and the "system of linear equations in two variables" in the intermediate level have a prerequisite relationship, a vertical connection is established between them, and the connection weight can be set as the reciprocal of the difference in the difficulty coefficients of the two knowledge points. By combining horizontal and vertical connections, a multi-layered knowledge network is formed, which reflects both the cognitive transfer relationship between knowledge points and the hierarchical nature of the knowledge structure.

[0118] The degree of association is determined by calculating the number of connections between nodes and their neighbors in a multi-layered knowledge network. The degree of association measures the tightness of connections between knowledge points in the network, calculated by counting the number of edges directly connected to the node, including horizontal connections within the same level and vertical connections across levels. A higher degree of association indicates more frequent interaction between the knowledge point and other knowledge points, and a more important position within the knowledge system. Mediation ability is calculated based on the number of times a node is transmitted in the cognitive transfer path. Mediation ability reflects the bridging role of a knowledge point in the knowledge transfer process, calculated by counting the number of times the node appears in all possible shortest paths. Nodes with high mediation ability are usually key nodes connecting different knowledge groups or different cognitive levels, playing a crucial role in knowledge transformation and integration during the learning process. The importance weight of nodes is determined based on the degree of association and mediation ability. Importance weight is an indicator that comprehensively evaluates the importance of a knowledge point in a multi-layered knowledge network, calculated as a weighted average of the degree of association and mediation ability. The weight allocation can be adjusted according to teaching objectives and learner characteristics. Generally, a weight of 0.4 for the degree of association and 0.6 for the mediation ability can be set, allowing nodes playing a key role in knowledge transfer to receive higher importance weights.

[0119] Based on the nodes in the cognitive transfer path and the multi-layered knowledge network, the knowledge flow coefficient, representing the nodes, is calculated. The knowledge flow coefficient is an indicator that measures the efficiency of information flow and cognitive transformation of knowledge points during the learning process, reflecting the contribution of knowledge points to the overall learning process. The calculation of the knowledge flow coefficient considers the node's position in the cognitive transfer path, its connection pattern in the multi-layered knowledge network, and the frequency of interaction with other nodes.

[0120] The specific calculation methods include: counting the frequency of nodes appearing in the cognitive transfer path, with higher frequency indicating stronger commonality of the node across different learning paths; analyzing the connectivity diversity of nodes in multi-layered knowledge networks, i.e., the degree to which they connect to different types or levels of nodes, with higher diversity indicating broader knowledge transfer; and evaluating the balance of knowledge inflow and outflow of nodes, i.e., the ratio of a node's in-degree to its out-degree, with better balance indicating higher efficiency in knowledge exchange. These three indicators are normalized and then weighted to obtain the final knowledge flow coefficient, ranging from 0 to 1, with higher values ​​indicating better flow.

[0121] Based on importance weights and knowledge flow coefficients, nodes in a multi-layered knowledge network are sorted and organized to generate a sequence of teaching knowledge points. The sorting process considers the combined influence of importance weights and knowledge flow coefficients, and the final ranking value of a node can be determined through linear combination or multiplicative relationships. For example, the final ranking value can be set as the importance weight multiplied by the knowledge flow coefficient, so that knowledge points that are both important and have good flowability receive a higher ranking. When organizing the sequence of teaching knowledge points, the hierarchical structure of the multi-layered knowledge network and the continuity constraints of cognitive transfer paths also need to be considered. A hierarchical traversal strategy is usually adopted, prioritizing knowledge points with high ranking values ​​while satisfying prerequisite relationships, forming a teaching sequence that conforms to cognitive patterns and highlights key points. For specific teaching scenarios, the length and complexity of the sequence can be adjusted according to teaching objectives and time constraints. For example, for introductory teaching, high-ranking knowledge points from the basic level and some intermediate levels can be selected; for advanced teaching, the proportion of intermediate and advanced level knowledge points can be increased.

[0122] Through the above technical solution, this implementation method realizes the transformation and optimization from cognitive transfer paths to sequences of teaching knowledge points, providing technical support for the organization of personalized teaching content. This method integrates multi-layer network analysis, mediation capability assessment, and knowledge flow computation techniques to achieve accurate evaluation of the importance of knowledge points and the teaching sequence. Compared with traditional linear teaching sequences, the multi-layer knowledge network constructed by this method more comprehensively reflects the hierarchical structure and relationships of knowledge. The generated teaching sequence considers both the inherent importance of knowledge points and the smoothness of the learning process, significantly improving teaching effectiveness and learning experience, and providing key technical support for personalized teaching and adaptive learning systems.

[0123] Furthermore, the sequence of teaching knowledge points is hierarchically coded, and tiered teaching content is generated based on this hierarchical coding, including:

[0124] For each knowledge point in the sequence of teaching knowledge points, a hierarchical code is generated based on its cognitive level and teaching order. The hierarchical code includes a level number and a sequence number.

[0125] Based on the hierarchical coding analysis, the difference in hierarchical numbers between adjacent knowledge points is analyzed. When the hierarchical number of adjacent knowledge points changes, the knowledge point at the changed position is identified as the cognitive leap point, and the cognitive span is obtained by calculating the difference in hierarchical numbers at the cognitive leap point.

[0126] Cognitive importance is calculated by statistically analyzing the number of times knowledge points are transmitted and the number of connections in the cognitive transfer path, and cognitive gap value is calculated based on cognitive importance and cognitive span.

[0127] Based on the cognitive gap value, transitional teaching content is configured at the cognitive leap point to generate tiered teaching content with hierarchical coding.

[0128] In this embodiment, hierarchical codes are generated for knowledge points in the teaching knowledge point sequence based on their cognitive level and teaching order. Hierarchical coding is a structured knowledge point identification method, comprising two main components: a hierarchy number and a sequence number. The hierarchy number indicates the cognitive level to which the knowledge point belongs, typically using numbers such as 1, 2, 3, etc., to represent different cognitive levels such as basic, intermediate, and advanced levels. The sequence number indicates the teaching order of knowledge points within the same cognitive level, numbered sequentially from first to last. The hierarchical coding adopts the format "hierarchical number.sequence number," for example, code "1.3" represents the third knowledge point in the basic level, and code "2.1" represents the first knowledge point in the intermediate level. Through hierarchical coding, the hierarchical affiliation and teaching order of knowledge points can be clearly expressed, providing a structured basis for subsequent teaching content organization and transition design.

[0129] By analyzing the differences in level codes between adjacent knowledge points using hierarchical coding, cognitive leap points are identified and cognitive spans are calculated. Differences in level codes between adjacent knowledge points reflect changes in the cognitive levels of the teaching content. When the level codes of adjacent knowledge points change, it indicates that the teaching content has leaped from one cognitive level to another, and this leap may affect learners' cognitive coherence. The knowledge points where the level codes change are identified as cognitive leap points, which are turning points in the teaching content that require special attention. The cognitive span is obtained by calculating the difference in level codes at the cognitive leap points, representing the magnitude of the change in cognitive level. For example, if the hierarchical codes of adjacent knowledge points are "1.5" and "2.1", then the level code changes from 1 to 2, with a difference of 1, indicating a leap from the basic level to the intermediate level, and the cognitive span is 1. If the hierarchical codes of adjacent knowledge points are "1.5" and "3.1", then the level code changes from 1 to 3, with a difference of 2, indicating a direct leap from the basic level to the advanced level, and the cognitive span is 2. The larger the cognitive span, the more significant the change in cognitive level, and the greater the cognitive challenge that learners may face.

[0130] Cognitive importance is calculated by statistically analyzing the number of times knowledge points are transmitted and the number of connections in the cognitive transfer path. Cognitive gap is then calculated based on cognitive importance and cognitive span. Cognitive importance is an indicator of the importance of a knowledge point in the overall learning process. The calculation method includes two aspects: first, counting the number of times a knowledge point is transmitted in the cognitive transfer path; a higher number of transmissions indicates stronger commonality of the knowledge point across different learning paths; second, calculating the number of connections of a knowledge point in the knowledge network; a higher number of connections indicates more frequent interaction between the knowledge point and other knowledge points. After normalizing these two indicators, a weighted average is used to obtain the final cognitive importance, with a value ranging from 0 to 1, where a higher value indicates higher importance. The cognitive gap is an indicator that comprehensively considers cognitive importance and cognitive span, used to quantify the difficulty of cognitive transformation between adjacent knowledge points. The cognitive gap is calculated by multiplying the cognitive span by a cognitive importance weighting coefficient, which is determined based on the specific application scenario and teaching objectives. For example, if the cognitive importance of a knowledge point at the cognitive transition point is 0.8, the cognitive span is 2, and the weighting coefficient is set to 0.5, then the cognitive gap value is 2 × (1 + 0.8 × 0.5) = 3. The larger the cognitive gap value, the higher the difficulty of cognitive transformation, and the more transitional teaching content is required.

[0131] Transitional teaching content is configured at cognitive leap points based on the cognitive gap value, generating tiered teaching content with hierarchical coding. Transitional teaching content is intermediate content designed to mitigate cognitive jumps and enhance learning coherence, including forms such as concept connections, case transitions, and analogies. The principle for configuring transitional teaching content is: the larger the cognitive gap value, the richer the transitional content; the higher the cognitive importance, the greater the depth and breadth of the transitional content. Specific configuration methods can be tiered according to the size of the cognitive gap value: when the cognitive gap value is small (e.g., less than 2), simple concept reviews and related prompts can be configured; when the cognitive gap value is medium (e.g., 2-4), case analysis and analogies can be added; when the cognitive gap value is large (e.g., greater than 4), complete transitional units can be designed, including intermediate concepts, exercises, and application scenarios. Transitional teaching content should also be assigned appropriate hierarchical coding, usually using decimal form to indicate its transitional nature. For example, transitional content between codes "1.5" and "2.1" can be coded as "1.5.1", "1.5.2", etc. By rationally allocating transitional teaching content, the resulting tiered teaching content maintains the hierarchical structure of knowledge points while enhancing the coherence and learnability between different levels.

[0132] Through the above technical solution, this implementation method achieves a structured transformation from a sequence of teaching knowledge points to tiered teaching content, providing technical support for personalized teaching content design. This method achieves structured identification of knowledge points through hierarchical coding, quantifies the cognitive transition difficulty of teaching content through cognitive leap point identification and cognitive gap value calculation, and solves the problem of jumping between different cognitive levels through intelligent configuration of transitional teaching content. Compared with traditional linear teaching content organization, the tiered teaching content generated by this method has a clear hierarchical structure and smooth cognitive transitions, effectively reducing learners' cognitive burden, improving learning efficiency and knowledge transfer ability, and providing a technical framework for content organization in intelligent education systems.

[0133] like Figure 3 As shown, Figure 3 This is a schematic diagram of the structure of a knowledge graph-based interpretable teaching cognitive interaction system provided in an embodiment of the present invention. The system includes:

[0134] The data acquisition module collects learners' response data and calculates a knowledge mastery state vector based on the response data.

[0135] The feature extraction module maps the knowledge mastery state vector to knowledge graph nodes and performs feature decomposition to obtain the cognitive feature matrix. Based on the cognitive feature matrix, the cognitive difficulty coefficient of the knowledge node is calculated.

[0136] The cognitive path construction module constructs a cognitive path graph based on the cognitive difficulty coefficient, calculates the node migration probability between adjacent knowledge nodes in the cognitive path graph, and generates a cognitive migration path.

[0137] The knowledge network construction module constructs a multi-layer knowledge network along the cognitive transfer path, calculates the importance weight and knowledge flow coefficient of the nodes in the multi-layer knowledge network, and generates a sequence of teaching knowledge points.

[0138] The teaching content generation module encodes the sequence of teaching knowledge points to generate tiered teaching content.

[0139] One technical solution provided in this embodiment of the invention is an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps in any of the aforementioned methods.

[0140] One technical solution provided in this embodiment of the invention is a computer-readable storage medium storing computer program instructions, which, when executed by a processor, implement the steps in any of the aforementioned methods.

[0141] The specific embodiments described above are preferred embodiments of the present invention and are not intended to limit the specific scope of the present invention. The scope of the present invention includes, but is not limited to, these specific embodiments. All equivalent changes made in accordance with the shape and structure of the present invention are within the protection scope of the present invention.

Claims

1. An explainable teaching cognitive interaction method based on a knowledge graph, characterized in that, The method comprises: collecting answer data of a learner, and calculating a knowledge mastery state vector according to the answer data; mapping the knowledge mastery state vector to knowledge graph nodes and performing feature decomposition to obtain a cognitive feature matrix, and calculating a cognitive difficulty coefficient of a knowledge node based on the cognitive feature matrix; constructing a cognitive path graph according to the cognitive difficulty coefficient, calculating a node migration probability between adjacent knowledge nodes in the cognitive path graph, and generating a cognitive migration path; constructing a multi-layer knowledge network on the cognitive migration path, calculating an importance weight and a knowledge circulation coefficient of a node in the multi-layer knowledge network, and generating a teaching knowledge point sequence; encoding the teaching knowledge point sequence to generate a step-by-step teaching content.

2. The method of claim 1, wherein, The method comprises: collecting answer data of a learner, and calculating a knowledge mastery state vector according to the answer data; collecting answer data of a learner and performing time sequence segmentation to generate a time sequence answer sequence; performing wavelet decomposition on the time sequence answer sequence to extract multi-scale fluctuation features, and calculating a cognitive state value based on the multi-scale fluctuation features; adjusting a forgetting compensation curve using the cognitive state value, decreasing the compensation coefficient when the cognitive state value rises, increasing the compensation coefficient when the cognitive state value falls, multiplying the compensation coefficient by the answer data to obtain a corrected answer value; performing fluctuation synchronicity analysis on the corrected answer value, calculating a cognitive correlation degree between knowledge points, constructing a knowledge correlation matrix according to the cognitive correlation degree, and identifying a knowledge group with a synchronous change feature from the knowledge correlation matrix; 3. The method of claim 1, wherein, performing feature combination on the cognitive correlation degree of the knowledge group and the cognitive state value to generate a knowledge mastery state vector. The method comprises: calculating a mapping weight between nodes based on the hierarchical relationship and correlation degree between knowledge points in the knowledge graph, mapping the knowledge mastery state vector to the knowledge graph nodes according to the mapping weight to obtain a node state value; performing spectral decomposition on the node state value to obtain a feature value sequence and a feature vector group; 4. The method of claim 1, wherein, calculating a weight coefficient of a cognitive dimension according to the distribution law of the feature value sequence and the hierarchical relationship between nodes, combining the feature vector group and the weight coefficient group to obtain a cognitive feature matrix. The method comprises: performing normalization processing on the cognitive feature matrix to calculate the discrete degree of each dimension in the normalized cognitive feature matrix, and converting the discrete degree into a dimension weight; performing weighting on the normalized cognitive feature matrix using the dimension weight to obtain a weighted cognitive feature of the knowledge node; calculating a principal component value of the weighted cognitive feature, and converting the principal component value into a cognitive complexity; 5. The method of claim 1, wherein, performing compensation correction on the cognitive complexity according to the ratio relationship between the positive component and the negative component of the principal component value to generate a cognitive difficulty coefficient of the knowledge node, wherein the positive component represents the understanding promotion degree of the knowledge node, and the negative component represents the understanding hindering degree of the knowledge node. The method comprises: calculating a node migration probability between adjacent knowledge nodes in the cognitive path graph to generate a cognitive migration path. The cognitive difficulty coefficient is interval mapped to obtain a node weight value, and a cognitive node weighted matrix is constructed based on the node weight value; The cognitive distance between nodes is calculated using the cognitive node weighted matrix, the cognitive distance and the node weight value are combined to generate a path connection strength, and a directed weighted cognitive path graph is constructed according to the path connection strength; The node topology feature is extracted from the directed weighted cognitive path graph, the node topology feature and the path connection strength are fused, and the cognitive jump probability is calculated; A state transition matrix is constructed using the cognitive jump probability, Markov chain analysis is performed on the state transition matrix, a steady-state migration path is obtained, and a node migration probability matrix is generated according to the steady-state migration path; Random walk sampling is performed based on the node migration probability matrix, the sampling path is scored and sorted, and the path with the highest score is selected as the cognitive migration path.

6. The method of claim 1, wherein, A multi-layer knowledge network is constructed on the cognitive migration path, the importance weight and the knowledge circulation coefficient of the nodes in the multi-layer knowledge network are calculated, and a teaching knowledge point sequence is generated including: On the cognitive migration path, the knowledge points are mapped to different cognitive levels according to the cognitive difficulty coefficient of the knowledge points, the original cognitive migration relationship is maintained within the same cognitive level, and vertical connections are established between different cognitive levels to construct a multi-layer knowledge network; The connection number of the nodes and adjacent nodes in the multi-layer knowledge network is calculated to obtain the correlation degree, the intermediary ability is calculated based on the number of transmissions of the nodes in the cognitive migration path, and the importance weight of the nodes is determined according to the correlation degree and the intermediary ability; The knowledge circulation coefficient representing the nodes is calculated according to the nodes in the cognitive migration path and the multi-layer knowledge network; Based on the importance weight and the knowledge circulation coefficient, the nodes in the multi-layer knowledge network are sorted and organized to generate a teaching knowledge point sequence.

7. The method of claim 1, wherein, The teaching knowledge point sequence is hierarchically coded, and a step-by-step teaching content is generated according to the hierarchical coding including: The knowledge points in the teaching knowledge point sequence are hierarchically coded based on their cognitive levels and teaching sequences, and the hierarchical coding contains a level number and a serial number; According to the hierarchical coding, the level number difference between adjacent knowledge points is analyzed, when the level number of adjacent knowledge points changes, the knowledge point at the change position is determined as a cognitive crossing point, and the cognitive span is obtained by calculating the level number difference value at the cognitive crossing point; The cognitive importance is calculated by counting the number of transmissions and the number of connections of the knowledge points in the cognitive migration path, and the cognitive drop value is calculated according to the cognitive importance and the cognitive span; According to the cognitive drop value, transition teaching content is configured at the cognitive crossing point to generate a step-by-step teaching content with hierarchical coding.

8. An explainable teaching cognitive interaction system based on a knowledge graph for implementing the method of any one of claims 1-7, characterized in that, The system includes: A collection module collects the answer data of the learners, and calculates a knowledge mastery state vector based on the answer data; A feature extraction module maps the knowledge mastery state vector to a knowledge graph node and performs feature decomposition to obtain a cognitive feature matrix, and calculates a cognitive difficulty coefficient of a knowledge node based on the cognitive feature matrix; A cognitive path construction module constructs a cognitive path graph according to the cognitive difficulty coefficient, calculates the node migration probability between adjacent knowledge nodes in the cognitive path graph, and generates a cognitive migration path. The knowledge network construction module constructs a multi-layer knowledge network on a cognitive transfer path, calculates an importance weight and a knowledge flow coefficient of a node in the multi-layer knowledge network, and generates a teaching knowledge point sequence; The teaching content generation module encodes the teaching knowledge point sequence to generate a step-by-step teaching content.

9. An electronic device, comprising: The computer program is stored in the memory and executable on the processor, and the processor executes the computer program to implement the steps in the method of any one of claims 1 to 7. The computer program instructions are stored on the computer readable storage medium, and the processor executes the computer program instructions to implement the steps in the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, ​

Images