Teaching content integration method and system based on knowledge graph and storage medium

Abstract

The application relates to the technical field of data processing, and discloses a teaching content integration method and system based on a knowledge graph and a storage medium. The method comprises the following steps: annotating cognitive difficulty of multiple-source teaching resources, and constructing a concept ontology library; fusing semantic similarity and cognitive span through a preposition dependency matrix algorithm to generate a preposition dependency graph; performing six-layer cognitive level coding to obtain a multilayer semantic matrix; dynamically planning and recombining a learner's cognitive ability and the semantic matrix to generate an adaptive content sequence; constructing a timestamp fusion space, evolving the graph through a relationship decay model to obtain a dynamic personalized learning path. The application solves the problems of missing cognitive load quantification, inaccurate concept dependency relationship mining, static knowledge representation limitation and dynamic updating lag in the existing teaching content integration method based on the knowledge graph.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a teaching content integration method and system based on a knowledge graph and a storage medium. BACKGROUND

[0002] The existing teaching content integration method mainly adopts a rule-based knowledge organization method and a static content recommendation strategy. The traditional method organizes teaching content through a predefined course outline and fixed knowledge point relationship, arranges teaching resources using simple keyword matching or classification tags, and recommends content based on the historical performance or basic information of learners. These methods can classify and manage teaching resources to a certain extent, and provide preliminary technical support for teaching activities.

[0003] However, there is a lack of effective cognitive load quantification mechanism, which cannot accurately assess the cognitive pressure that learners need to bear when learning specific concepts, resulting in unreasonable teaching content arrangement; secondly, the concept dependency relationship is not accurately mined, which is only based on simple co-occurrence statistics or preset rules, and cannot identify the real pre-dependency relationship and cognitive span between concepts; thirdly, a single level of knowledge representation is used, which cannot reflect the semantic difference and complexity of teaching concepts at different cognitive levels; finally, there is a lack of dynamic updating mechanism, and the knowledge graph remains static once it is constructed, which cannot timely reflect the changes of teaching content and learning effect feedback.

[0004] Based on the above analysis, the existing technology lacks a comprehensive technical solution that can simultaneously solve the problems of cognitive load quantification, accurate dependency relationship mining, multi-level semantic representation, and dynamic evolution. Because the cognitive load cannot be accurately quantified, it is impossible to perform personalized content reorganization based on the cognitive ability of learners; because the concept dependency relationship is not accurately mined, it further affects the construction quality of multi-cognitive level semantic representation; because there is a lack of multi-level semantic representation, the dynamic planning reorganization based on cognitive load lacks an effective data basis; finally, because of the lack of dynamic updating mechanism, the entire teaching content integration system cannot adapt to the timeliness changes of knowledge and the evolution of individualized needs of learners. SUMMARY

[0005] The present application provides a teaching content integration method and system based on a knowledge graph and a storage medium, which solves the problems of lack of cognitive load quantification, inaccurate concept dependency relationship mining, static knowledge representation limitations, and dynamic updating lag in the existing teaching content integration method based on a knowledge graph, and improves the individualization degree and adaptive ability of teaching content integration.

[0006] In a first aspect, the present application provides a teaching content integration method based on a knowledge graph, which comprises:

[0007] The multi-source teaching resources are annotated with cognitive difficulty coefficients to obtain a teaching concept ontology library containing cognitive attribute vectors;

[0008] The teaching concept ontology library is subjected to semantic similarity and cognitive span fusion calculation processing by a preposition dependency matrix algorithm to obtain a concept preposition dependency graph;

[0009] The teaching concepts are subjected to six-layer cognitive level decomposition and coding processing according to the concept preposition dependency graph to obtain a multi-cognitive level semantic representation matrix;

[0010] The learner cognitive ability vector and the multi-cognitive level semantic representation matrix are subjected to dynamic programming reorganization processing under cognitive load constraints to obtain a cognitive load adaptive content sequence;

[0011] A timestamp-labeled incremental update fusion space is constructed based on the cognitive load adaptive content sequence, and a knowledge graph evolution process is performed through a relationship strength decay model to obtain a dynamic personalized learning path.

[0012] In a second aspect, the present application provides a knowledge graph-based teaching content integration system, which comprises:

[0013] An annotation module is configured to annotate multi-source teaching resources with cognitive difficulty coefficients to obtain a teaching concept ontology library containing cognitive attribute vectors;

[0014] A calculation module is configured to perform semantic similarity and cognitive span fusion calculation processing on the teaching concept ontology library by a preposition dependency matrix algorithm to obtain a concept preposition dependency graph;

[0015] An encoding module is configured to perform six-layer cognitive level decomposition and coding processing on teaching concepts according to the concept preposition dependency graph to obtain a multi-cognitive level semantic representation matrix;

[0016] A reorganization module is configured to perform dynamic programming reorganization processing on a learner cognitive ability vector and the multi-cognitive level semantic representation matrix under cognitive load constraints to obtain a cognitive load adaptive content sequence;

[0017] An update module is configured to construct a timestamp-labeled incremental update fusion space based on the cognitive load adaptive content sequence, and perform a knowledge graph evolution process through a relationship strength decay model to obtain a dynamic personalized learning path.

[0018] In a third aspect, a knowledge graph-based teaching content integration device is provided, comprising a memory and at least one processor, the memory storing instructions; the at least one processor invoking the instructions in the memory to enable the knowledge graph-based teaching content integration device to perform the knowledge graph-based teaching content integration method described above.

[0019] In a fourth aspect, a computer-readable storage medium is provided, the computer-readable storage medium storing instructions that, when executed on a computer, enable the computer to perform the knowledge graph-based teaching content integration method described above.

[0020] In the technical solutions provided in the present application, the pre-dependent matrix algorithm is used to realize accurate fusion calculation of semantic similarity and cognitive span. The algorithm can accurately identify the real dependency relationship between teaching concepts, overcoming the limitations of the prior art based on simple co-occurrence statistics, and providing a reliable relationship basis for subsequent cognitive level decomposition. The six-layer cognitive level decomposition and coding processing is based on Bloom's cognitive level theory, and the teaching concepts are semantically represented at the memory, understanding, application, analysis, evaluation, and creation levels, breaking through the limitations of traditional single-level knowledge representation, so that the teaching content can be finely organized according to different cognitive complexity. The dynamic programming reorganization processing under the constraint of cognitive load realizes content sequence optimization based on individual cognitive characteristics by matching the learner's cognitive ability vector with the multi-cognitive level semantic representation matrix, solving the key problem that the prior art cannot quantify cognitive load. The time stamp marked incremental update fusion space combines the relationship strength decay model to establish a dynamic evolution mechanism of the knowledge graph, so that the teaching content integration can adapt to the timeliness changes of knowledge and the dynamic adjustment of learning needs.

[0021] The three-dimensional fusion calculation feature of the pre-dependent matrix algorithm significantly improves the mining accuracy of concept dependency relationships, enabling the identification of implicit dependency relationships that cannot be discovered by traditional methods, laying a solid foundation for the construction of personalized learning paths. The construction feature of the multi-cognitive level semantic representation matrix enables the same teaching concept to exhibit differentiated semantic features at different cognitive levels, supporting more refined cognitive ability matching and load assessment. This multi-level representation capability is not achievable by traditional flat knowledge organization methods. The optimization feature of the dynamic programming reorganization algorithm under the constraint of cognitive load ensures that the generated learning sequence meets both the logical dependency relationships between concepts and the cognitive bearing capacity of learners, realizing truly personalized teaching content arrangement. The timeliness adjustment feature of the relationship strength decay model enables the knowledge graph to reflect the dynamic changes of teaching knowledge, ensuring the timeliness and accuracy of teaching content integration. This dynamic evolution capability fills the technical gap in timeliness of existing static knowledge graphs, providing key technical support for the continuous optimization of intelligent education systems. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of one embodiment of the knowledge graph-based teaching content integration method in this application.

[0024] Figure 2 This is a schematic diagram of one embodiment of the knowledge graph-based teaching content integration system in this application.

[0025] Figure 3 This is a schematic block diagram of the teaching content integration device based on knowledge graphs in an embodiment of the present invention. Detailed Implementation

[0026] This application provides a method, system, and storage medium for integrating teaching content based on a knowledge graph. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0027] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the knowledge graph-based teaching content integration method in this application includes:

[0028] Step S101: Perform cognitive difficulty coefficient annotation on the multi-source teaching resources to obtain a teaching concept ontology library containing cognitive attribute vectors;

[0029] Step S102: Perform semantic similarity and cognitive span fusion calculation on the teaching concept ontology using the pre-dependency matrix algorithm to obtain the concept pre-dependency graph;

[0030] Step S103, according to the concept preposition dependency graph, the teaching concept is processed by six-layer cognitive hierarchy decomposition coding, and a multi-cognitive level semantic representation matrix is obtained.

[0031] Step S104, the learner cognitive ability vector and the multi-cognitive level semantic representation matrix are processed by dynamic programming reorganization under cognitive load constraint, and a cognitive load adaptive content sequence is obtained.

[0032] Step S105, based on the cognitive load adaptive content sequence, a timestamp marked incremental update fusion space is constructed, a knowledge graph evolution process is performed through a relationship strength attenuation model, and a dynamic personalized learning path is obtained.

[0033] It can be understood that the execution subject of the present application can be a knowledge graph-based teaching content integration system, and can also be a terminal or a server, and the specific place is not limited. The server is taken as an example for description.

[0034] Specifically, the teaching text is segmented by dividing the continuous text into independent text segments according to the knowledge point boundary, and each segment corresponds to a main teaching concept. Then, the original text segment is extracted based on the TF-IDF weight calculation, and the TF-IDF algorithm identifies important words in the text by calculating the product of word frequency and inverse document frequency, thereby obtaining a core teaching concept identifier list. Then, the concept identifier is input into the cognitive difficulty evaluation model for quantitative calculation, and the model outputs difficulty coefficient and learning time coefficient according to the concept abstraction degree, prerequisite knowledge requirement and learning complexity. Finally, a three-dimensional cognitive attribute vector containing difficulty coefficient, learning time coefficient and number of preposition concepts is constructed, and unified identification is performed through a concept identifier mapping table.

[0035] The semantic similarity and cognitive span fusion calculation is performed by the preposition dependency matrix algorithm. The preposition dependency matrix is an n x n matrix, where n represents the total number of teaching concepts, and the matrix element value represents the preposition dependency strength between concepts. First, the teaching concepts are input into the Word2Vec model for vectorization processing. The Word2Vec model maps words to high-dimensional dense vectors through neural network training, capturing the semantic relationship between concepts. Then, based on the concept semantic vector set, the preposition dependency matrix is constructed, and the initial preposition dependency strength matrix is obtained through the fusion calculation of the semantic similarity, temporal dependency and cognitive span three dimensions. The semantic similarity is calculated by the cosine similarity of the included angle cosine value of two concept vectors, the temporal dependency is calculated based on the appearance order of concepts in the teaching content, and the cognitive span is determined according to the cognitive difficulty difference between concepts. The three-dimensional cognitive attribute vector is input into the weight optimization algorithm for parameter adjustment to obtain the weight coefficient combination of the three dimensions. Finally, the initial preposition dependency strength matrix is weighted and fused according to the weight coefficient combination, and the dependency relationship is selected by the threshold judgment algorithm. When the dependency strength value exceeds the preset threshold, it is determined that there is a preposition dependency relationship between concepts.

[0036] The six-layer cognitive hierarchy decomposition and coding processing is performed according to the concept preposition dependency graph. Based on Bloom's cognitive hierarchy theory, teaching concepts are divided into six cognitive levels: memory level, understanding level, application level, analysis level, evaluation level and creation level, each representing different cognitive complexity and thinking requirements. The cognitive level judgment algorithm identifies the main cognitive level of each concept by analyzing the lexical features, grammatical features and contextual features of the concept, including the complexity and professionalism of the concept vocabulary, the syntactic structure of the concept expression, and the contextual information of the concept in the teaching material. Then, the teaching concepts are processed by hierarchical semantic coding, generating semantic vectors for each concept at six cognitive levels, forming a concept hierarchical semantic vector set. The graph neural network models the semantic relationship between concepts through neighbor information aggregation processing, and the attention mechanism allocates different weights according to the importance between concepts, and the updated hierarchical semantic vectors are organized into a multi-cognitive level semantic representation matrix according to the hierarchical structure.

[0037] The learner cognitive ability vector is reorganized and processed by dynamic programming under the cognitive load constraint of the multi-cognitive level semantic representation matrix. The learner cognitive ability vector is constructed based on historical learning data and cognitive test results, and contains six dimensions of memory capacity, understanding speed, application proficiency, analysis depth, evaluation accuracy and innovation tendency. The matching degree calculation process is carried out by corresponding matching of the six-dimensional ability characteristics of the learner with the six-layer cognitive semantic vector in the multi-cognitive level semantic representation matrix, forming the ability-demand matching matrix. The Euclidean distance algorithm measures the difference between the learner's ability and the concept demand, and calculates the concept cognitive difference measure value. The cognitive load mapping function converts the difference measure value into a standardized cognitive load intensity value, and the sigmoid function performs nonlinear transformation processing to obtain the normalized cognitive load intensity coefficient. The time complexity factor adjusts the cognitive load in the time dimension, and the weighted product calculation is carried out with the concept learning time. The dynamic programming algorithm optimizes the teaching content combination under the constraint of cognitive load, and the learning efficiency maximization objective function evaluates the sequence quality through the balance calculation of learning benefit and cognitive load.

[0038] The time stamp marked incremental update fusion space is constructed based on the cognitive load adaptive content sequence. The time stamp records the creation time, update time and access time of each teaching concept, forming a time dimension data tracking mechanism. The incremental update fusion space adopts a four-tuple structure, including concept set, relationship set, time set and update operation set, and the space index algorithm processes data organization. The relationship strength decay model calculates the timeliness of concept relationship by exponential decay function, and the relationship strength between concepts gradually weakens over time, reflecting the timeliness characteristics of knowledge. The knowledge graph update process updates the graph structure through the weighted fusion calculation of new and old relationship strength and the conflict detection mechanism, and the conflict detection mechanism identifies and solves the contradiction between new knowledge and existing knowledge. The A-star search algorithm generates a dynamic personalized learning path by combining learner feedback data, and the heuristic function estimates the path cost according to the concept difficulty and learner ability.

[0039] In a specific embodiment, the process of step S101 can specifically include the following steps:

[0040] The teaching text is segmented and processed, and the original text segment set is obtained according to the knowledge point boundary;

[0041] The original text segment set is processed by keyword extraction based on TF-IDF weight calculation, and a list of core teaching concept identifiers is obtained;

[0042] The list of core teaching concept identifiers is input into the cognitive difficulty evaluation model for quantitative calculation, and the concept attribute parameters including difficulty coefficient and learning time coefficient are obtained;

[0043] According to the cognitive attribute vector construction processing of the teaching concept according to the concept attribute parameter, a three-dimensional cognitive attribute vector is obtained.

[0044] Based on the concept identifier mapping table, the three-dimensional cognitive attribute vector is uniformly identified, and a teaching concept ontology library containing the cognitive attribute vector is obtained.

[0045] Specifically, the first step of cognitive difficulty coefficient annotation processing of multi-source teaching resources is teaching text segmentation processing. This processing is a unified cutting operation for three types of heterogeneous data sources: teaching material text, video subtitles and exercise library. The teaching text segmentation processing adopts a segmentation strategy based on knowledge point boundaries. By identifying features such as paragraph titles, chapter separators and keyword density changes, the position of the knowledge point boundary is determined. The continuous teaching text is cut into independent text segments according to the principle of semantic integrity. Each original text segment corresponds to a complete knowledge point content, including the definition, explanation, example and exercise of the knowledge point, forming an original text segment set as the basis data unit for subsequent processing.

[0046] The key step of identifying core teaching concepts is the keyword extraction processing of the original text segment set based on TF-IDF weight calculation. The TF-IDF algorithm quantifies the importance of a word in a specific document by calculating the product of term frequency and inverse document frequency. The term frequency reflects the frequency of a word in a single text segment, and the inverse document frequency reflects the rarity of a word in the entire text segment set. The algorithm first performs word segmentation and stop word removal on each text segment, then calculates the TF value of each word, which is the number of occurrences of the word in the current segment divided by the total number of words in the segment, then calculates the IDF value, which is the logarithm value of the total number of text segments divided by the number of segments containing the word, and finally multiplies the TF value and the IDF value to obtain the TF-IDF weight. By setting a weight threshold, the words with the highest TF-IDF value are selected as the keywords of the text segment. After concept recognition and standardization processing, the keyword forms a list of core teaching concept identifiers.

[0047] The input of the core teaching concept identifier list into the cognitive difficulty assessment model for quantitative calculation processing is the core link for constructing the concept attribute parameters. The cognitive difficulty assessment model is an evaluation algorithm based on multi-feature fusion. The model receives the concept identifier as input, calculates the cognitive difficulty of the concept by analyzing multiple dimensional features such as vocabulary complexity, abstract degree, prerequisite knowledge requirement, and learning level. The vocabulary complexity is measured by the number of phonetic symbols, character length, and professional term density of the concept vocabulary, the abstract degree is evaluated by the concreteness score and visualization degree of the concept, the prerequisite knowledge requirement is determined by analyzing the dependency relationship between the concept and other concepts, and the learning level is divided according to the six levels in Bloom's cognitive taxonomy. The model inputs these feature values into a neural network for weighted calculation, and outputs concept attribute parameters including difficulty coefficient and learning duration coefficient. The difficulty coefficient reflects the cognitive complexity of the concept, and the learning duration coefficient represents the average time required to master the concept.

[0048] According to the concept attribute parameters, the cognitive attribute vector construction processing of the teaching concept is the process of integrating multi-dimensional attribute information into a standardized vector representation. The cognitive attribute vector construction processing combines the difficulty coefficient, the learning duration coefficient, and the number of prerequisite concepts obtained through dependency relationship analysis into a three-dimensional vector structure. The vector construction process first normalizes the three parameters, mapping the parameter values of different dimensions to the standard interval of zero to one, and then organizes the three normalized values into a vector form according to a fixed dimension order. The first dimension of the three-dimensional cognitive attribute vector represents the normalized difficulty coefficient, the second dimension represents the normalized learning duration coefficient, and the third dimension represents the normalized number of prerequisite concepts. The length of the vector reflects the overall cognitive complexity of the concept.

[0049] Based on the concept identifier mapping table, the unified identification processing of the three-dimensional cognitive attribute vector is the final step of establishing the concept ontology library. The concept identifier mapping table is a data structure containing concept name, unique identifier, synonym list, and upper and lower relationship information, which is used to solve the naming inconsistency problem of the same concept in different data sources. The unified identification processing identifies different expressions pointing to the same concept through string matching, semantic similarity calculation, expert rules, and other methods, and maps them to a unified identifier. The processing process establishes an association between each three-dimensional cognitive attribute vector and the corresponding unified identifier, and records metadata information such as the source information, creation time, and update history of the concept. After the unified identification processing, the data is organized according to the ontology structure to form a teaching concept ontology library containing concept entities, attribute relationships, and hierarchical structure.

[0050] In a specific embodiment, the process of performing step S102 can specifically include the following steps:

[0051] The teaching concepts in the teaching concept ontology library are input into a Word2Vec model for vectorization processing to obtain a concept semantic vector set;

[0052] A pre-requisite dependency matrix is constructed based on the concept semantic vector set, wherein the rows and columns of the matrix correspond to the teaching concepts, and the matrix element values represent the pre-requisite dependency strength between the concepts. Through the fusion calculation and processing of three dimensions of semantic similarity, time sequence dependency, and cognitive span, an initial pre-requisite dependency strength matrix is obtained.

[0053] The three-dimensional cognitive attribute vectors are input into a weight optimization algorithm for parameter adjustment processing to obtain a weight coefficient combination of semantic similarity, time sequence dependency, and cognitive span.

[0054] The initial pre-requisite dependency strength matrix is weighted and fused according to the weight coefficient combination to obtain a weighted pre-requisite dependency matrix.

[0055] The weighted pre-requisite dependency matrix is input into a threshold judgment algorithm for dependency relationship screening processing. When the dependency strength value exceeds a preset threshold, it is determined that there is a pre-requisite dependency relationship between the concepts, and a concept pre-requisite dependency graph is obtained.

[0056] Specifically, the first step of the pre-requisite dependency matrix algorithm is to input the teaching concepts in the teaching concept ontology library into the Word2Vec model for vectorization processing. The Word2Vec model is a word embedding technology based on neural networks that learns the semantic representation of words by analyzing the context relationships of words in large-scale teaching corpus. The model uses a skip-gram model architecture to learn the distributed representation of words by predicting the context words around the target words. Each teaching concept is mapped to a fixed-dimensional dense vector. The vectorization processing process first takes the concept identifiers in the teaching concept ontology library as input words, and converts the discrete concept identifiers into continuous numerical vectors through the embedding layer of the model. Each concept corresponds to a high-dimensional vector containing semantic information. These vectors can reflect the semantic similarity between concepts, and concepts with similar semantics are closer in vector space, thereby forming a concept semantic vector set as the basis data for subsequent matrix construction.

[0057] The construction of the prerequisite dependency matrix based on the concept semantic vector set is the core step of establishing the dependency relationship between concepts. The prerequisite dependency matrix is a square matrix, and the rows and columns of the matrix correspond to all concepts in the teaching concept ontology library. The matrix element value represents the prerequisite dependency strength of the row concept to the column concept. The construction process determines the dependency relationship strength between concepts through the fusion calculation of three dimensions of semantic similarity, time sequence dependency, and cognitive span. Semantic similarity is obtained by calculating the cosine similarity of two concept semantic vectors. The cosine similarity is equal to the product of the dot product of two vectors divided by the product of the vector length, with a value range of negative one to positive one. Time sequence dependency is calculated based on the order of concept appearance in teaching materials. By analyzing the positional relationship of concepts in the course outline and textbook chapters, the temporal dependency is determined. Cognitive span is calculated according to the difference in cognitive difficulty between concepts. By comparing the difficulty coefficients in the three-dimensional cognitive attribute vectors of two concepts, the gap in cognitive complexity is measured. The calculation results of the three dimensions are fused according to a specific mathematical relationship to form the initial prerequisite dependency strength matrix.

[0058] Inputting the three-dimensional cognitive attribute vector into the weight optimization algorithm for parameter adjustment processing is the key link to determine the weight coefficients of the three dimensions. The weight optimization algorithm uses an optimization strategy based on gradient descent to learn the optimal weight combination by minimizing the error between the predicted dependency relationship and the actual teaching sequence. The algorithm takes the three-dimensional cognitive attribute vector of the teaching concept as the feature input, including the difficulty coefficient, learning duration coefficient, and number of prerequisite concepts. The optimization process quantifies the difference between the prediction results and the true labels by constructing a loss function, which considers the accuracy of concept dependency relationship and the rationality of learning path. The algorithm uses the backpropagation mechanism to update the weight parameters, gradually adjusts the weight coefficients of semantic similarity, time sequence dependency, and cognitive span through multiple rounds of iterative training, and finally converges to the weight combination that best fits the teaching rules.

[0059] According to the weight combination, the initial prerequisite dependency strength matrix is weighted and fused for calculation and processing, which is an important step in integrating multi-dimensional information. Weighted fusion calculation linearly weights and sums the values of semantic similarity, time sequence dependency, and cognitive span according to the coefficients obtained by the weight optimization algorithm. The fusion process calculates the weighted sum for each element in the matrix, i.e., multiplies the semantic similarity at that position by the semantic weight coefficient, multiplies the time sequence dependency by the time sequence weight coefficient, and multiplies the cognitive span by the cognitive weight coefficient. Then, the three product results are added to obtain the final dependency strength value at that position. Weighted fusion calculation ensures the reasonable integration of different dimensional information, solves the possible bias problem of single-dimensional judgment of dependency relationship, and forms a weighted prerequisite dependency matrix that considers multiple factors comprehensively.

[0060] The weighted preposition dependency matrix is input into a threshold judgment algorithm for dependency relationship screening processing, which is a decision step for determining the final concept dependency relationship. The threshold judgment algorithm distinguishes strong dependency relationships from weak dependency relationships by setting a preset threshold. When the dependency strength value of a matrix element exceeds the preset threshold, it is determined that there is a preposition dependency relationship between the corresponding concepts. The threshold setting is based on empirical data and statistical analysis results in teaching practice, and usually selects a numerical value that can filter out noise relationships but retain important dependency relationships. The judgment process checks each element value in the weighted preposition dependency matrix one by one, and marks the positions exceeding the threshold as having a dependency relationship, and marks the positions below the threshold as having no dependency relationship. The binary matrix after threshold screening is converted into a graph structure representation, where concepts are nodes and dependency relationships are directed edges, and finally forms a concept preposition dependency graph.

[0061] In a specific embodiment, the process of performing step S103 can specifically include the following steps:

[0062] Based on Bloom's cognitive level theory, the teaching concepts are classified and processed according to cognitive levels, and the concepts are divided into six cognitive levels: memory level, understanding level, application level, analysis level, evaluation level and creation level, to obtain concept cognitive level classification labels;

[0063] The teaching concepts in the concept preposition dependency graph are input into a cognitive level judgment algorithm for main cognitive level identification processing, and through comprehensive analysis of lexical features, grammatical features and contextual features, the main cognitive level identification of each concept is obtained;

[0064] According to the main cognitive level identification, the teaching concepts are processed for hierarchical semantic coding, and semantic vectors are generated for each concept at six cognitive levels to obtain a concept hierarchical semantic vector set;

[0065] The concept hierarchical semantic vector set is input into a graph neural network for neighbor information aggregation processing, and the semantic relationship between concepts is modeled through an attention mechanism to obtain updated hierarchical semantic vectors;

[0066] Based on the updated hierarchical semantic vectors, matrix organization processing is performed, and the six-layer cognitive semantic vectors of all concepts are arranged according to the hierarchical structure to obtain a multi-cognitive level semantic representation matrix.

[0067] Specifically, the cognitive level classification of teaching concepts based on Bloom's cognitive level theory is the theoretical basis step of six-level cognitive level decomposition coding. Bloom's cognitive level theory divides the learning process into six progressive cognitive levels, each representing different levels of thinking complexity and cognitive requirements. The memory layer requires learners to be able to recall and identify facts, terms, basic concepts, and other basic information. The understanding layer requires learners to be able to explain, summarize, and convert the meaning of information. The application layer requires learners to be able to apply learned knowledge to solve problems in new situations. The analysis layer requires learners to be able to decompose complex information and identify the relationship between parts. The evaluation layer requires learners to be able to judge and evaluate based on standards and criteria. The creation layer requires learners to be able to integrate information to generate new ideas or solutions. Classification processing classifies each concept into the corresponding cognitive level by analyzing the content characteristics, learning objectives, and cognitive requirements of teaching concepts, establishes the mapping relationship between concepts and cognitive levels, and forms the cognitive level classification label of concepts as the classification basis for subsequent processing.

[0068] The main cognitive level identification processing of the teaching concepts in the concept precedence dependency graph into the cognitive level determination algorithm is the core link of determining the main cognitive level of the concept. The cognitive level determination algorithm analyzes the cognitive level attributes of the concept through multi-feature fusion. The algorithm receives the teaching concepts in the concept precedence dependency graph as input data. Vocabulary feature analysis includes language features such as vocabulary complexity, professional term density, and abstraction level of concept names. The language complexity of the concept is quantified through indicators such as the number of phonetic elements, character length, and word frequency statistics. Syntax feature analysis considers features such as syntax structure, syntax complexity, and modifier usage of concept expression on the syntax level. The syntax complexity information of the concept is extracted through syntax analysis tree and dependency relationship analysis. Context feature analysis focuses on contextual information, associated concepts, and position of the concept in the teaching materials. The cognitive positioning of the concept is determined by analyzing the vocabulary distribution, semantic association, and teaching sequence around the concept. The algorithm calculates the weighted fusion of the three types of features, predicts the cognitive level of the concept through the machine learning classifier, and outputs the probability distribution of each concept on the six cognitive levels. The probability of the highest level is selected as the main cognitive level identifier of the concept.

[0069] The hierarchical semantic encoding of teaching concepts based on the main cognitive level identifier is a key step in constructing a multi-level semantic representation. The hierarchical semantic encoding process is based on the main cognitive level identifier of each concept, and generates corresponding semantic vector representations for the concept at all six cognitive levels. The encoding process first determines the semantic intensity of the concept at the main level based on the main cognitive level identifier of the concept, and then calculates the semantic intensity of the concept at other cognitive levels through the analysis of cognitive relevance between levels. The cognitive relevance between levels reflects the differences in the performance of the same concept at different cognitive levels. The concept representation at the memory level focuses on the encoding of factual information, the representation at the understanding level emphasizes the association between concepts, the representation at the application level highlights the actual application scenario, the representation at the analysis level focuses on the structural decomposition of the concept, the representation at the evaluation level embodies the judgment standard, and the representation at the creation level reflects the innovation potential. The encoding algorithm generates fixed-dimensional dense vectors for each concept at six cognitive levels. These vectors capture the semantic characteristics of the concept at different cognitive levels, forming a set of hierarchical semantic vectors of the concept as input data for the graph neural network.

[0070] The input of the set of hierarchical semantic vectors of the concept into the graph neural network for neighbor information aggregation processing is the core algorithm of semantic relationship modeling. The graph neural network is a deep learning model specifically designed to process graph-structured data. This network treats concepts as graph nodes and the pre-requisite dependency relationships between concepts as graph edges. It propagates and aggregates information on the graph through a message passing mechanism. The neighbor information aggregation process updates the representation vector of each concept node by collecting information from its neighbor nodes. The aggregation function performs weighted summation or averaging operations on the semantic vectors of neighbor nodes. The attention mechanism dynamically calculates the importance weights of different neighbor nodes during the aggregation process. By learning attention scores, it determines which neighbor concepts are more important for the semantic representation of the current concept. The calculation of attention scores is based on factors such as semantic similarity between concepts, pre-requisite dependency strength, and cognitive level differences. A higher score indicates a greater impact of the corresponding neighbor concept on the current concept. The graph neural network iteratively updates the semantic vectors of all concepts through multiple rounds of message passing. In each iteration, the new representation of a concept incorporates information from its neighbor concepts, and eventually converges to a stable updated hierarchical semantic vector.

[0071] The final step in constructing a multi-cognitive-level semantic representation matrix is ​​to perform matrix organization processing based on the updated hierarchical semantic vectors. Matrix organization systematically arranges the six-layer cognitive semantic vectors of all concepts according to a hierarchical structure, forming a three-dimensional matrix structure. The first dimension of the matrix corresponds to the number of teaching concepts, the second dimension corresponds to the six cognitive levels, and the third dimension corresponds to the dimension of the semantic vectors. The organization process first determines the position of a concept in the matrix according to the identifier order, and then arranges the six-layer semantic vectors of each concept in the order of memory layer, understanding layer, application layer, analysis layer, evaluation layer, and creation layer. The value at each position in the matrix represents the semantic feature strength of the corresponding concept at a specific cognitive level, and the overall structure of the matrix reflects the complete semantic representation relationship of all concepts across multiple cognitive levels.

[0072] In one specific embodiment, the process of executing step S104 may specifically include the following steps:

[0073] Based on learners' historical learning data and cognitive test results, a six-dimensional cognitive ability feature is constructed, including memory capacity, comprehension speed, application proficiency, analytical depth, evaluation accuracy, and innovation tendency, resulting in a learner's cognitive ability vector;

[0074] The matching degree of learners’ cognitive ability vectors and multi-cognitive level semantic representation matrices is calculated. Through the difference analysis between conceptual cognitive needs and learners’ abilities, a set of conceptual cognitive load coefficients is obtained.

[0075] A cognitive load constraint function is constructed based on the set of conceptual cognitive load coefficients. The constraint condition that the total cognitive load does not exceed the learner's tolerance threshold is set. The teaching content combination optimization is carried out through dynamic programming algorithm to obtain a set of candidate content sequences that meet the constraint conditions.

[0076] The candidate content sequence set is input into the learning efficiency maximization objective function for sequence evaluation. By balancing the learning benefits and cognitive load, the optimal content sequence ranking result is obtained.

[0077] Based on the optimal content sequence sorting results, the teaching concepts are reorganized temporally to ensure that the prerequisite dependencies are satisfied and the cognitive load is evenly distributed, thus obtaining a cognitive load adaptive content sequence.

[0078] Specifically, the cognitive ability state of the learner is comprehensively characterized by multi-source data fusion. The historical learning data includes the performance records of the learner in past learning activities, covering the correct answer rate, learning duration, knowledge point mastery, learning path selection, and other behavior data, which reflect the ability performance of the learner in the actual learning process. The cognitive test results come from specially designed cognitive ability evaluation tests, including memory tests, understanding tests, application ability tests, and other standardized evaluation tools. The construction process of the six-dimensional cognitive ability feature maps the raw data to the six cognitive dimensions. Memory capacity is calculated by analyzing the learner's performance in memory tasks, including information retention time, memory accuracy, memory capacity size, and other indicators. Understanding speed is calculated based on the time required for the learner to understand new concepts and the depth of understanding. Application proficiency is evaluated by the learner's performance in actual problem solving. Analysis depth examines the learner's ability to decompose complex problems. Evaluation accuracy measures the learner's judgment and evaluation accuracy. Innovation tendency reflects the learner's performance tendency in creative tasks. The construction process normalizes the raw data of each dimension, unifies the data of different dimensions to the interval of zero to one, and finally forms a six-dimensional learner cognitive ability vector.

[0079] Matching the learner cognitive ability vector with the multi-cognitive level semantic representation matrix is a key step in cognitive load quantification. The matching degree calculation process determines the learning difficulty by comparing the learner's cognitive ability with the cognitive demand of the teaching concept. This process compares the six dimensions of the learner's cognitive ability vector with the semantic vectors of the corresponding cognitive levels in the multi-cognitive level semantic representation matrix. The matching degree calculation uses a vector similarity algorithm to measure the matching degree by calculating the Euclidean distance between the learner's ability vector and the concept demand vector. The smaller the distance, the more matched the learner's ability and concept demand. The larger the distance, the greater the ability gap. The difference analysis between concept cognitive demand and learner ability compares the ability demand of each concept one by one, analyzes the demand intensity of each concept in the six cognitive levels, and analyzes the gap between the corresponding ability of the learner. The difference analysis result is converted into a cognitive load coefficient, which reflects the amount of cognitive resources the learner needs to invest when learning a specific concept. The larger the difference, the higher the cognitive load, and the smaller the difference, the lower the cognitive load. The calculation process organizes the cognitive load coefficients of all concepts into a set form, providing quantitative load evaluation data for subsequent optimization algorithms.

[0080] The cognitive load constraint function is constructed based on the concept cognitive load coefficient set, which is the constraint condition setting step of dynamic programming optimization. The cognitive load constraint function is a mathematical function that takes the teaching content sequence as the input variable and outputs the total cognitive load corresponding to the sequence. The construction of the constraint function is based on the accumulation principle of cognitive load, that is, the total cognitive load of the learning sequence is equal to the sum of the cognitive load coefficients of all concepts in the sequence. The constraint condition sets the total cognitive load not to exceed the threshold value of the learner, which is determined comprehensively according to the cognitive ability level of the learner, the learning time limit, the fatigue degree and other factors. Dynamic programming algorithm is an algorithm used to solve optimization problems, which divides the complex teaching content combination optimization problem into multiple sub-problems, and constructs the global optimal solution by solving the optimal solution of the sub-problems step by step. The algorithm maintains a state transition table in the search process, which records the optimal choices in different states, and calculates the cost of transferring from one state to another state through the state transition equation. The optimization process finds the sequence combination containing the most effective teaching content under the premise of meeting the cognitive load constraint, and generates multiple candidate content sequence sets that meet the constraint conditions.

[0081] The sequence evaluation processing of the candidate content sequence set into the learning efficiency maximization target function is the decision step of optimal sequence selection. The learning efficiency maximization target function is an evaluation function that measures the ratio between the learning effect of the teaching content sequence and the cognitive load. The design of the target function considers the balance calculation of learning benefit and cognitive load. Learning benefit is calculated by evaluating the importance of concepts, knowledge coverage, skill improvement potential and other factors in the sequence, and cognitive load is calculated by the cognitive load coefficient set described above. The balance calculation adopts a weighted scoring mechanism, taking learning benefit as a positive factor and cognitive load as a negative factor, and balancing the influence of the two by setting appropriate weight coefficients. The sequence evaluation processing calculates the target function value of each sequence in the candidate content sequence set respectively, and the higher the function value, the better the learning efficiency of the sequence. The evaluation process also considers the integrity, logicality, difficulty gradient and other quality factors of the sequence, and ensures that the selected sequence is both efficient and reasonable through multi-dimensional evaluation, and finally outputs the optimal content sequence sorting result sorted according to the target function value.

[0082] The time sequence reorganization of the teaching concepts according to the optimal content sequence ranking result is a key step for generating the final learning path. The time sequence reorganization is based on the optimal content sequence ranking result to determine the learning order of the teaching concepts, and the process needs to meet two constraint conditions of the preposition dependency relationship and the uniform distribution of cognitive load. The preposition dependency relationship ensures that the learner has mastered all the preposition concepts before learning a concept, avoiding learning difficulties due to lack of basic knowledge. The uniform distribution of cognitive load requires reasonable allocation of concepts with high cognitive load in the learning sequence to avoid cognitive overload caused by concentrated learning of too many high-load concepts in a short time. The reorganization process adopts a topological sorting algorithm combined with a load balancing strategy. First, the basic order of the concepts is determined according to the preposition dependency relationship, and then the position of the concepts is adjusted to balance the distribution of cognitive load under the premise of meeting the dependency relationship. The adjustment process optimizes the load distribution through strategies such as swapping adjacent concept positions, inserting buffer concepts, and splitting high-load concept groups, and finally generates a cognitive load adaptive content sequence that meets both logical dependency and reasonable load distribution.

[0083] In a specific embodiment, the process of performing step S105 can specifically include the following steps:

[0084] Perform layer-by-layer corresponding matching processing on the six-dimensional ability features in the learner cognitive ability vector and the six-layer cognitive semantic vector in the multi-cognitive level semantic representation matrix to obtain an ability-demand pairing matrix;

[0085] Based on the ability-demand pairing matrix, perform cognitive demand intensity calculation processing on each teaching concept, measure the difference between the learner's ability and the concept's demand through the Euclidean distance algorithm, and obtain a concept cognitive difference measurement value;

[0086] According to the concept cognitive difference measurement value, construct a cognitive load mapping function to convert the difference measurement value into a standardized cognitive load intensity value, and perform nonlinear transformation processing through a sigmoid function to obtain a normalized cognitive load intensity coefficient;

[0087] Perform weighted product calculation processing on the normalized cognitive load intensity coefficient and the concept learning time length, introduce a time complexity factor to adjust the cognitive load in the time dimension, and obtain a time-weighted cognitive load coefficient;

[0088] Based on the time-weighted cognitive load coefficient, perform set organization processing on all teaching concepts, and index and arrange them according to the concept identifier to obtain a concept cognitive load coefficient set.

[0089] Specifically, the six-dimensional ability features in the learner cognitive ability vector are matched with the six-layer cognitive semantic vectors in the multi-cognitive hierarchical semantic representation matrix layer by layer, which is the basic step of establishing the ability-demand relationship. The matching process compares the cognitive ability of the learner with the cognitive demand of the teaching concept one by one through dimensional correspondence. The memory capacity dimension of the learner cognitive ability vector is matched with the memory layer vector of the multi-cognitive hierarchical semantic representation matrix, the understanding speed dimension is matched with the understanding layer vector, the application proficiency dimension is matched with the application layer vector, the analysis depth dimension is matched with the analysis layer vector, the evaluation accuracy dimension is matched with the evaluation layer vector, and the innovation tendency dimension is matched with the creation layer vector. The matching process pairs the semantic vector of each teaching concept at the six cognitive levels with the ability value of the corresponding dimension of the learner, forming an ability-demand pairing matrix as the input data for difference calculation. The ability-demand pairing matrix is a two-dimensional data structure, with rows corresponding to teaching concepts and columns corresponding to the six cognitive levels. Each matrix element contains the ability value of the learner in the corresponding cognitive dimension and the demand intensity value of the concept in the corresponding cognitive level. The pairing matrix directly reflects the corresponding relationship between the learner's ability and the demand of each concept.

[0090] The cognitive demand intensity calculation process for each teaching concept based on the ability-demand pairing matrix is the core link of quantifying the learning difficulty. The cognitive demand intensity calculation process determines the difficulty level of the learner learning a specific concept by analyzing each row of data in the ability-demand pairing matrix. This process calculates the cognitive demand intensity of each teaching concept separately. The Euclidean distance algorithm is a mathematical method for measuring the distance between two points in a multi-dimensional space. This algorithm treats the six-dimensional ability features of the learner and the six-dimensional cognitive demand of the teaching concept as two points in a six-dimensional space. It measures the matching degree of ability and demand by calculating the straight-line distance between the two points. The distance calculation process first calculates the difference between the learner's ability value and the concept's demand value in each dimension, then squares the six differences, adds the six squared values, and finally takes the square root of the sum to get the Euclidean distance value. The larger the distance value, the greater the difference between the learner's ability and the concept's demand, and the higher the difficulty of learning the concept. The smaller the distance value, the better the matching degree, and the lower the learning difficulty. Each teaching concept is calculated by the Euclidean distance algorithm to obtain the corresponding concept cognitive difference measurement value, which quantifies the degree of cognitive challenge faced by the learner when learning a specific concept.

[0091] The construction of the cognitive load mapping function according to the concept cognitive difference measure value is a key step to convert the difference measure into a standardized load coefficient. The cognitive load mapping function is a mathematical transformation function that maps the original difference measure value to the standardized cognitive load intensity interval. The construction of the mapping function is based on the nonlinear relationship principle between cognitive load and learning difficulty in cognitive psychology, that is, the growth of cognitive load is not a simple linear relationship with the difficulty difference. The sigmoid function is a S-shaped curve function, which has good nonlinear transformation characteristics and can map any range of input values to the output interval of zero to one. The nonlinear transformation process transforms the concept cognitive difference measure value through the sigmoid function. In the transformation process, smaller difference measure values get lower load coefficients after function transformation, and larger difference measure values get higher load coefficients. The S-shaped characteristics of the transformation curve ensure the smooth transition and reasonable distribution of the load coefficient. The transformation process solves the extreme value problem and the dimension problem of the original difference measure value. The output normalized cognitive load intensity coefficient is uniformly distributed in the standard interval of zero to one, which is convenient for subsequent calculation and comparison operations.

[0092] The weighted product calculation of the normalized cognitive load intensity coefficient and the concept learning duration is an important step to introduce the time dimension. The weighted product calculation process fully evaluates the cognitive load level of learners by combining cognitive intensity and time factors. This process takes into account the cognitive pressure learners bear in a unit of time. Concept learning duration refers to the average time learners need to master a specific concept. This duration data comes from teaching practice statistics and cognitive experiment measurement results. The time complexity factor is a regulating parameter that reflects the degree of influence of time pressure on cognitive load. The factor value is set based on the relationship between time pressure and cognitive performance in cognitive psychology. The weighted product calculation multiplies the normalized cognitive load intensity coefficient with the concept learning duration, and then performs a second product operation with the time complexity factor. The calculation result reflects the comprehensive cognitive load level of learners learning the concept under specific time constraints. The time dimension adjustment ensures the comprehensiveness and accuracy of cognitive load evaluation, avoiding the limitation of only considering cognitive intensity and ignoring time factors. The time-weighted cognitive load coefficient obtained is more close to the actual learning situation.

[0093] The final step of constructing the load coefficient set is to organize and process all teaching concepts based on the time-weighted cognitive load coefficient. The organization and processing will arrange and store the scattered time-weighted cognitive load coefficients according to a unified data structure, ensuring the orderliness and accessibility of the data. The concept identifier is the unique identifier of each teaching concept, which has been determined in the construction of the teaching concept ontology library and has the characteristics of uniqueness and invariability. The index arrangement sorts the time-weighted cognitive load coefficients according to the lexicographical order or numerical order of the concept identifiers. The arrangement process establishes the mapping relationship between the concept identifiers and the corresponding load coefficients. The organization and processing also include data integrity checking and outlier processing to ensure that each concept has a corresponding load coefficient and the value is within a reasonable range. The processing result forms the concept cognitive load coefficient set, which stores the concept identifier and the corresponding time-weighted cognitive load coefficient in the form of key-value pairs. The set structure supports fast query and batch processing operations.

[0094] In a specific embodiment, the process of performing step S106 can specifically include the following steps:

[0095] Timestamping each teaching concept in the cognitive load adaptive content sequence, recording the creation time, update time, and access time of the concept, and obtaining a timestamped concept sequence;

[0096] Based on the timestamped concept sequence, a four-tuple fusion space structure is constructed, including four dimensions of concept set, relationship set, time set, and update operation set. The data is organized and processed through a spatial indexing algorithm to obtain an incremental update fusion space;

[0097] Input the concept relationship in the concept preposition dependency graph into the relationship strength attenuation model for timeliness calculation and processing. The time weight of the relationship strength is adjusted through an exponential attenuation function to obtain a time-decayed relationship strength matrix;

[0098] According to the time-decayed relationship strength matrix, the incremental update fusion space is updated for the knowledge graph, and through the weighted fusion calculation and conflict detection mechanism of the new and old relationship strength, the evolved knowledge graph structure is obtained;

[0099] Based on the evolved knowledge graph structure, the cognitive load adaptive content sequence is processed for path optimization, and through the A-star search algorithm combined with learner feedback data, the dynamic personalized learning path is obtained.

[0100] Specifically, the cognitive load adaptive content sequence is time-stamped for each teaching concept, which is the basic step of building a time-effective knowledge graph. Time-stamping tracks the life cycle status of the concept by attaching time information to each teaching concept. This process records three key time information to fully reflect the time-effective characteristics of the concept. The creation time records the time point when the concept is first added to the knowledge graph, which marks the initial establishment of the concept in the teaching system. The update time records the latest time point when the concept content or attribute is modified, reflecting the dynamic change process of the concept information. The access time records the time point when the concept is last accessed or used by the learner, reflecting the actual application frequency and importance of the concept. The marking process establishes a relationship between the three time stamps and the corresponding teaching concept, forming a concept data structure containing time dimension information. The time stamp is standardized in a unified time format to ensure consistency and comparability of time information from different sources. The result of the processing generates a time-stamped concept sequence, which adds time dimension information to the original cognitive load adaptive content sequence. Each concept contains not only cognitive attributes and dependency relationships, but also complete time tracking information.

[0101] The time-stamped concept sequence is used to build a four-tuple fusion space structure, which is the core step of establishing a multi-dimensional knowledge representation. The four-tuple fusion space structure is a multi-dimensional data organization model that fully describes the state and evolution of the knowledge graph through four different dimension sets. The concept set contains all teaching concepts and their attribute information, and each concept element carries cognitive attribute vectors and time-stamped information. The relationship set stores various relationships between concepts, including pre-dependence relationships, semantic similarity relationships, cognitive level relationships, and other types of concept connections. The time set records all time-related information in the knowledge graph, including concept timestamps, relationship time effectiveness, and time series of updates. The update operation set records all modification operations on the knowledge graph, including concept addition, relationship update, attribute modification, and other operation types and execution times. The space indexing algorithm is an efficient data organization method that supports fast data querying and updating operations by establishing a multi-dimensional indexing structure. Data organization processing arranges and stores data in four dimensions according to the indexing structure, establishing multiple indexing relationships such as concept identifiers, relationship types, time intervals, and operation types. Incremental update fusion space supports dynamic data insertion, deletion, and modification operations, maintaining data consistency and integrity.

[0102] The input of concept relationship in the concept precedence dependency graph into the relationship strength decay model for timeliness calculation and processing is the key step of introducing time factors. The relationship strength decay model is a time modeling method that dynamically adjusts the strength of the relationship between concepts based on the timeliness principle of knowledge. The decay model considers the timeliness characteristics of teaching knowledge, that is, the importance and relevance of some concept relationships change over time. The exponential decay function is a mathematical function that simulates the change of relationship strength over time through the decay law of exponential form. The decay function takes the difference between the current time and the relationship establishment time as the input variable, and outputs the corresponding decay coefficient for adjusting the original relationship strength. Time weight adjustment calculates the adjusted relationship strength by multiplying the original relationship strength with the decay coefficient, and the adjustment process considers the relationship establishment time, the latest update time, the access frequency and other time factors. The calculation and processing are performed on each relationship edge in the concept precedence dependency graph, and the corresponding decay degree is determined according to the time attribute of the relationship. The processing result generates a time-decayed relationship strength matrix, which reflects the relationship strength distribution between concepts considering timeliness factors.

[0103] According to the time-decayed relationship strength matrix, the knowledge graph update processing of the incremental update fusion space is an important step to realize dynamic evolution. The knowledge graph update processing maintains the timeliness and accuracy of the knowledge graph by integrating new relationship strength information, which involves the fusion and conflict resolution mechanism of new and old data. The weighted fusion calculation of new and old relationship strength uses the weighted average method to combine the new values in the time-decayed relationship strength matrix with the original values in the incremental update fusion space. The fusion calculation considers the timeliness, credibility, source authority and other factors of the data to determine the weight distribution, ensuring that the fusion result not only preserves historical information but also reflects the latest changes. The conflict detection mechanism identifies inconsistencies and contradictions between new and old data, and determines the type and severity of conflicts through rule reasoning and expert knowledge. The detection mechanism includes numerical conflict detection, logical conflict detection, semantic conflict detection and other types of consistency verification. Conflict resolution strategies use different processing methods according to the conflict type, including data coverage, weighted average, expert arbitration, user confirmation and other solutions. The update processing finally generates the evolved knowledge graph structure, which integrates the relationship information adjusted for timeliness and the consistent data resolved for conflict.

[0104] The path optimization processing of the content sequence adaptive to the cognitive load based on the evolved knowledge graph structure is the final link of generating the dynamic learning path. The path optimization processing adapts to the dynamic changes of the knowledge graph by re-planning the learning sequence, and considers various factors such as the evolved concept relationship, the updated cognitive load, the feedback information of the learner, etc. The A-star search algorithm is a heuristic path search algorithm, which guides the search process through an evaluation function, reduces the search space while ensuring that the optimal path is found. The search algorithm takes the learning goal as the end point and the current knowledge state as the starting point, and selects the search direction by evaluating the estimated cost of each intermediate node to the end point. The heuristic function combines the cognitive load of the concept, the pre-dependence relationship, the learning benefit and other factors to estimate the path cost, and the function design ensures that the estimated value does not exceed the actual cost to ensure the optimality of the algorithm. The learner feedback data includes learning effect evaluation, learning preference survey, cognitive ability test and other information, which reflects the actual learning status and individualized needs of the learner. The path re-planning adjusts the original content sequence adaptive to the cognitive load by combining the learner feedback data, optimizes the learning order of the concept, adjusts the cognitive load distribution, adds or deletes related concepts, etc. The re-planning process ensures that the new learning path meets the pre-dependence relationship of the concept and adapts to the cognitive characteristics and feedback requirements of the learner, and finally generates a dynamic and individualized learning path.

[0105] The above describes the teaching content integration method based on the knowledge graph in the embodiments of the present application, and the following describes the teaching content integration system based on the knowledge graph in the embodiments of the present application. Please refer to Figure 2 An embodiment of the teaching content integration system based on the knowledge graph in the embodiments of the present application includes:

[0106] The annotation module is configured to perform cognitive difficulty coefficient annotation processing on the multi-source teaching resources to obtain a teaching concept ontology library containing cognitive attribute vectors.

[0107] The calculation module is configured to perform semantic similarity and cognitive span fusion calculation processing on the teaching concept ontology library by using a pre-dependence matrix algorithm to obtain a concept pre-dependence graph.

[0108] The encoding module is configured to perform six-layer cognitive level decomposition and encoding processing on the teaching concepts according to the concept pre-dependence graph to obtain a multi-cognitive level semantic representation matrix.

[0109] The recombination module is configured to perform cognitive load constraint dynamic programming recombination processing on the learner cognitive ability vector and the multi-cognitive level semantic representation matrix to obtain a cognitive load adaptive content sequence.

[0110] An updating module is configured to construct a timestamp-labeled incremental update fusion space based on the cognitive load adaptive content sequence, perform knowledge graph evolution processing through a relation strength decay model, and obtain a dynamic personalized learning path.

[0111] The above Figure 2 The knowledge graph-based teaching content integration system in the embodiment of the present application is described in detail from the perspective of modular functional entities, and the knowledge graph-based teaching content integration device in the embodiment of the present application is described in detail from the perspective of hardware processing.

[0112] Referring to Figure 3 The embodiment of the present application also provides a knowledge graph-based teaching content integration device, which can be a server, and the internal structure thereof can be as shown in Figure 3 The knowledge graph-based teaching content integration device comprises a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. The processor of the computer is configured to provide computing and control capabilities. The memory of the knowledge graph-based teaching content integration device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium. The database of the knowledge graph-based teaching content integration device is configured to store corresponding data in the embodiment. The network interface of the knowledge graph-based teaching content integration device is configured to communicate with an external terminal through a network connection. The computer program is executed by the processor to implement the above method.

[0113] Those skilled in the art can understand Figure 3 The structure shown in the embodiment of the present application is only a block diagram of part of the structure related to the present application scheme, and does not constitute a limitation on the knowledge graph-based teaching content integration device to which the present application scheme is applied.

[0114] The present application also provides a computer readable storage medium, which can be a non-volatile computer readable storage medium or a volatile computer readable storage medium. The computer readable storage medium stores instructions, and when the instructions are executed on a computer, the computer executes the steps of the knowledge graph-based teaching content integration method.

[0115] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, system and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described herein.

[0116] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application or the entire or part of the technical solutions that essentially contribute to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a knowledge graph-based teaching content integration device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0117] The above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those skilled in the art that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A knowledge graph-based teaching content integration method, characterized in that, The method comprises: Cognitive difficulty coefficient annotation processing is performed on multi-source teaching resources to obtain a teaching concept ontology library containing cognitive attribute vectors; A concept pre-dependence graph is obtained by performing semantic similarity and cognitive span fusion calculation processing on the teaching concept ontology library through a pre-dependence matrix algorithm, including: inputting teaching concepts in the teaching concept ontology library into a Word2Vec model for vectorization processing to obtain a concept semantic vector set; a pre-dependence matrix is constructed based on the concept semantic vector set, wherein the rows and columns of the matrix correspond to the teaching concepts respectively, and the matrix element values represent the pre-dependence strength between the concepts; an initial pre-dependence strength matrix is obtained by fusion calculation processing in three dimensions of semantic similarity, time sequence dependence and cognitive span, wherein the time sequence dependence determines the time dependence by analyzing the positional relationship of the concepts in the course outline and the textbook chapter, and the cognitive span measures the difference in cognitive complexity by comparing the difficulty coefficients in the three-dimensional cognitive attribute vectors of two concepts; the three-dimensional cognitive attribute vectors are input into a weight optimization algorithm for parameter adjustment processing to obtain a weight coefficient combination of the semantic similarity, the time sequence dependence and the cognitive span, wherein the three-dimensional cognitive attribute vectors contain the difficulty coefficient, the learning time coefficient and the number of pre-concepts of the teaching concepts; the initial pre-dependence strength matrix is weighted and fused by the weight coefficient combination to obtain a weighted pre-dependence matrix; the weighted pre-dependence matrix is input into a threshold judgment algorithm for dependence relationship screening processing, and when the dependence strength value exceeds a preset threshold, it is determined that there is a pre-dependence relationship between the concepts, thereby obtaining the concept pre-dependence graph; According to the concept pre-dependence graph, teaching concepts are processed by six-layer cognitive level decomposition and coding to obtain a multi-cognitive level semantic representation matrix; A cognitive load adaptive content sequence is obtained by performing dynamic programming reorganization processing on the learner cognitive ability vector and the multi-cognitive level semantic representation matrix under the cognitive load constraint, wherein the learner cognitive ability vector is constructed based on the historical learning data and the cognitive test results of the learner to obtain six-dimensional cognitive ability features, including memory capacity, understanding speed, application proficiency, analysis depth, evaluation accuracy and innovation tendency; the learner cognitive ability vector and the multi-cognitive level semantic representation matrix are matched and calculated to obtain a concept cognitive load coefficient set through the difference analysis of concept cognitive demand and learner ability. The adaptive content sequence based on the cognitive load is used for time stamp marking, an incremental update fusion space is constructed, a knowledge graph evolution process is performed through a relationship strength attenuation model, and a dynamic personalized learning path is obtained, including: each teaching concept in the adaptive content sequence based on the cognitive load is subjected to time stamp marking processing, the creation time, update time and access time of the concept are recorded, and a time stamp marked concept sequence is obtained; a four tuple fusion space structure is constructed based on the time stamp marked concept sequence, including four dimensions of a concept set, a relationship set, a time set and an update operation set, data organization processing is performed through a space index algorithm, and an incremental update fusion space is obtained; the concept relationship in the concept preposition dependency graph is input into a relationship strength attenuation model for timeliness calculation processing, the relationship strength attenuation model dynamically adjusts the strength of the relationship between concepts based on the knowledge timeliness principle; the time weight of the relationship strength is adjusted through an exponential attenuation function, and a time-decayed relationship strength matrix is obtained; the incremental update fusion space is subjected to knowledge graph update processing according to the time-decayed relationship strength matrix, and an evolved knowledge graph structure is obtained through a weighted fusion calculation of new and old relationship strengths and a conflict detection mechanism; the adaptive content sequence based on the cognitive load is subjected to path optimization processing based on the evolved knowledge graph structure, the path is re-planned through an A-star search algorithm combined with learner feedback data, and a dynamic personalized learning path is obtained. 2.The knowledge graph-based teaching content integration method of claim 1, wherein, The cognitive difficulty coefficient annotation processing is performed on the multi-source teaching resources, and a teaching concept ontology library containing a cognitive attribute vector is obtained, including: Teaching text segmentation processing is performed on the textbook text, video subtitles and exercise library, and an original text segment set cut according to the knowledge point boundary is obtained; Key word extraction processing is performed on the original text segment set based on TF-IDF weight calculation, and a core teaching concept identifier list is obtained; The core teaching concept identifier list is input into a cognitive difficulty evaluation model for quantitative calculation processing, and a concept attribute parameter containing a difficulty coefficient and a learning time coefficient is obtained; Cognitive attribute vector construction processing is performed on the teaching concept according to the concept attribute parameter, and a three-dimensional cognitive attribute vector is obtained; Uniform identification processing is performed on the three-dimensional cognitive attribute vector based on a concept identifier mapping table, and a teaching concept ontology library containing a cognitive attribute vector is obtained. 3.The knowledge graph-based teaching content integration method of claim 2, wherein, The six-layer cognitive level decomposition coding processing is performed on the teaching concept according to the concept preposition dependency graph, and a multi-cognitive level semantic representation matrix is obtained, including: Cognitive level classification processing is performed on the teaching concept based on Bloom's cognitive level theory, the concept is divided into six cognitive levels of memory level, understanding level, application level, analysis level, evaluation level and creation level, and a concept cognitive level classification label is obtained; The teaching concept in the concept preposition dependency graph is input into a cognitive level judgment algorithm for main cognitive level identification processing, the comprehensive analysis of the vocabulary features, syntax features and context features is performed, and the main cognitive level identifier of each concept is obtained; According to the main cognitive level identifier, the teaching concepts are subjected to hierarchical semantic coding processing, and a semantic vector is generated for each concept at six cognitive levels to obtain a concept hierarchical semantic vector set, wherein the coding process determines the semantic strength of the concept at the main level according to the main cognitive level identifier of the concept, and then calculates the semantic strength of the concept at other cognitive levels through cognitive correlation analysis between levels, the cognitive correlation between levels reflects the difference in the performance of the same concept at different cognitive levels, the concept at the memory level represents the encoding of factual information, the concept at the understanding level emphasizes the association between concepts, the concept at the application level highlights the actual application scenario, the concept at the analysis level focuses on the structural decomposition of the concept, the concept at the evaluation level embodies the judgment standard, and the concept at the creation level reflects the innovation potential, and a fixed-dimensional dense vector is generated for each concept at six cognitive levels; The concept hierarchical semantic vector set is input into a graph neural network for neighbor information aggregation processing, and the semantic relationship between concepts is modeled through an attention mechanism to obtain updated hierarchical semantic vectors; Based on the updated hierarchical semantic vectors, matrix organization processing is performed, and the six-layer cognitive semantic vectors of all concepts are arranged according to the hierarchical structure to obtain a multi-cognitive level semantic representation matrix. 4.The knowledge graph-based teaching content integration method of claim 1, wherein, The learning cognitive ability vector and the multi-cognitive level semantic representation matrix are subjected to dynamic programming reorganization processing under cognitive load constraints to obtain a cognitive load adaptive content sequence, including: A cognitive load constraint function is constructed based on the concept cognitive load coefficient set, a constraint condition that the total cognitive load does not exceed the learner's threshold is set, and dynamic programming algorithm is used for teaching content combination optimization processing to obtain a candidate content sequence set that meets the constraint condition; The candidate content sequence set is input into a learning efficiency maximization objective function for sequence evaluation processing, and the optimal content sequence sorting result is obtained through the balance calculation of learning benefit and cognitive load; According to the optimal content sequence sorting result, the teaching concepts are subjected to time sequence reorganization processing to ensure that the preposition dependency relationship is satisfied and the cognitive load is uniformly distributed, and a cognitive load adaptive content sequence is obtained. 5.The knowledge graph-based teaching content integration method of claim 4, wherein, The learning cognitive ability vector and the multi-cognitive level semantic representation matrix are subjected to matching degree calculation processing, and the concept cognitive load coefficient set is obtained through the difference analysis of concept cognitive demand and learner ability, including: The six-dimensional ability features in the learning cognitive ability vector are matched with the six-layer cognitive semantic vectors in the multi-cognitive level semantic representation matrix layer by layer to obtain an ability-demand pairing matrix; Based on the ability-demand pairing matrix, the cognitive demand strength of each teaching concept is calculated, and the difference degree between the learner ability and the concept demand is measured by the Euclidean distance algorithm to obtain a concept cognitive difference metric value; According to the concept cognitive difference metric value, a cognitive load mapping function is constructed to convert the difference metric value into a normalized cognitive load strength value, and a sigmoid function is used for nonlinear transformation processing to obtain a normalized cognitive load strength coefficient; The normalized cognitive load intensity coefficient is multiplied by the concept learning duration to obtain a time-weighted cognitive load coefficient by introducing a time complexity factor to adjust the cognitive load in the time dimension. All teaching concepts are organized and processed based on the time-weighted cognitive load coefficient, and are indexed and arranged according to concept identifiers to obtain a concept cognitive load coefficient set. 6.A knowledge graph-based teaching content integration system, characterized by, The knowledge graph-based teaching content integration system for implementing the knowledge graph-based teaching content integration method according to any one of claims 1-5 comprises: A labeling module is configured to label and process cognitive difficulty coefficients of multi-source teaching resources to obtain a teaching concept ontology library containing cognitive attribute vectors. A calculation module is configured to perform semantic similarity and cognitive span fusion calculation processing on the teaching concept ontology library by using a pre-dependence matrix algorithm to obtain a concept pre-dependence graph. An encoding module is configured to perform six-layer cognitive level decomposition and encoding processing on teaching concepts according to the concept pre-dependence graph to obtain a multi-cognitive level semantic representation matrix. A reorganization module is configured to perform dynamic programming reorganization processing on a learner cognitive ability vector and the multi-cognitive level semantic representation matrix under cognitive load constraints to obtain a cognitive load adaptive content sequence. An update module is configured to construct a timestamp-labeled incremental update fusion space based on the cognitive load adaptive content sequence, perform knowledge graph evolution processing by using a relationship strength decay model, and obtain a dynamic personalized learning path. 7.A knowledge graph-based teaching content integration device, characterized by The computer program, when executed on the processor, causes the processor to implement the knowledge graph-based teaching content integration method according to any one of claims 1-5.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program, when executed on the processor, causes the processor to implement the knowledge graph-based teaching content integration method according to any one of claims 1-5.

Images