Method and device for generating personalized wisdom education textbook based on large model

Abstract

The invention discloses a personalized intelligent education textbook generation method and device based on a large model, and the method comprises the steps: obtaining a knowledge point set, learning data of a learner, a learning target, and eye movement tracking data; determining a cognitive style vector of the learner according to the learning data and the eye movement tracking data; according to the cognitive style vector and the learning target, a learning path diagram of the learner for the knowledge point set is constructed, and the learning path diagram comprises learning nodes and jump relations among the multiple learning nodes; determining cognitive matching parameters of the knowledge points based on the cognitive style vector and the knowledge point attribute vector of each knowledge point in the knowledge point set, and determining presentation forms of the knowledge points according to the cognitive matching parameters; and based on the learning path diagram and the presentation form, generating a personalized intelligent education textbook through a preset large model. The learning efficiency of students is improved.

Description

Technical Field

[0001] This application relates to the field of intelligent education, and specifically relates to a method and device for generating personalized intelligent education textbooks based on large models. Background Art

[0002] With the in-depth development of education informatization, intelligent education has become an important means to improve teaching quality and learning effects. Especially in the context of the rapid popularization of online education, how to provide suitable teaching content for different learners has become an important issue of concern in the education field.

[0003] In the prior art, the method for generating personalized intelligent education textbooks usually recommends learning content based on the learning weaknesses of learners. By analyzing learners' learning data, such as test scores, answering records, error rates, etc., it identifies the knowledge points that they have not mastered or have not mastered firmly, and accordingly pushes corresponding remedial content. This "checking for deficiencies and filling in the gaps" method can improve learning efficiency to a certain extent.

[0004] However, the method of recommending learning content based on the learning weaknesses of learners in the prior art usually relies on learning result data such as test scores and answering correct rates, and it is difficult to fully reflect their actual needs in the process of knowledge acquisition. Even if different learners face the same knowledge weaknesses, there may be significant differences in the way they understand knowledge points, resulting in the lack of pertinence of the content of personalized intelligent education textbooks and reducing the learning efficiency of students. Summary of the Invention

[0005] This application provides a method and device for generating personalized intelligent education textbooks based on large models, which are used to improve the learning efficiency of students.

[0006] In the first aspect of this application, a method for generating personalized intelligent education textbooks based on large models is provided, which is applied to a server. The method includes: Obtain a set of knowledge points, learners' learning data, learning goals, and eye movement tracking data; determine the cognitive style vector of the learner according to the learning data and the eye movement tracking data, and the cognitive style vector is used to represent the information processing preference characteristics of the learner; construct a learning path map for the learner for the set of knowledge points according to the cognitive style vector and the learning goals, and the learning path map includes learning nodes and jump relationships between multiple learning nodes; determine the cognitive matching parameters of the knowledge points based on the cognitive style vector and the knowledge point attribute vectors of each knowledge point in the set of knowledge points, and determine the presentation form of the knowledge points according to the cognitive matching parameters; generate a personalized intelligent education textbook through a preset large model based on the learning path map and the presentation form.

[0007] Optionally, determining the cognitive style vector of the learner according to the learning data and the eye movement tracking data specifically includes: Extract behavioral feature data from the learning data; identify the content interaction features of the learner when processing the first knowledge point based on the page stay duration, repeated access frequency, and interaction operation sequence in the behavioral feature data, where the first knowledge point is any type of knowledge point; determine the cognitive processing features during the learning process according to the saccade rate, fixation transfer pattern, and pupil response characteristics in the eye movement tracking data; determine the cognitive style vector of the learner based on the content interaction features and cognitive processing features.

[0008] Optionally, determining the cognitive style vector of the learner based on the content interaction features and cognitive processing features specifically includes: determining the corresponding relationship between the content interaction features and cognitive processing features of the learner when learning different types of knowledge points through a preset behavior-cognition association model to obtain the learner's information processing feature set; determining the cognitive style vector according to the information processing feature set.

[0009] Optionally, determining the cognitive style vector according to the information processing feature set specifically includes: Analyze the information processing feature set to obtain the learner's information acquisition intensity score, information processing depth score, knowledge construction mode index score, and skill transfer efficiency score; calculate the dimensional feature values according to the cognitive feature parameters to obtain the cognitive style vector, where the cognitive feature parameters include the information acquisition intensity score, information processing depth score, knowledge construction mode index score, and skill transfer efficiency score.

[0010] Optionally, constructing a learning path map for the learner for the knowledge point set according to the cognitive style vector and learning objective, specifically including: According to the learning objective, screen out the second knowledge points that meet the preset requirements from the knowledge point set, and use the second knowledge points as learning nodes; determine the logical dependency relationship between the second knowledge points based on the preset knowledge graph; determine the jump relationship between the learning nodes based on the cognitive style vector and the logical dependency relationship to obtain the learning path map.

[0011] Optionally, determining the jump relationship between the learning nodes based on the cognitive style vector and the logical dependency relationship to obtain the learning path map specifically includes: Calculate the cognitive load value of each learning node according to the dimensional feature value of the cognitive style vector; construct the knowledge association intensity matrix between the learning nodes; combine the cognitive load value and the knowledge association intensity matrix to calculate the jump probability between the learning nodes; construct the learning path map according to the jump probability, and the learning path map includes the necessary learning path and the alternative learning path.

[0012] Optionally, determining the cognitive matching parameter of the knowledge point based on the cognitive style vector and the knowledge point attribute vector of each knowledge point in the knowledge point set, and determining the presentation form of the knowledge point according to the cognitive matching parameter, specifically including: Feature matching is performed between the cognitive style vector and the knowledge point attribute vector based on a preset feature matching rule to obtain a feature matching matrix; a cognitive matching parameter is calculated according to the feature matching matrix; a presentation form label is generated based on the cognitive matching parameter, and the presentation form of the knowledge point is determined according to the presentation form label.

[0013] In a second aspect of the present application, a personalized intelligent education textbook generation system based on a large model is provided, including: An acquisition module, configured to acquire a knowledge point set, learning data of a learner, a learning goal, and eye movement tracking data; A first determination module, configured to determine a cognitive style vector of the learner according to the learning data and the eye movement tracking data, where the cognitive style vector is used to represent the information processing preference characteristics of the learner; A construction module, configured to construct a learning path map of the learner for the knowledge point set according to the cognitive style vector and the learning goal, where the learning path map includes learning nodes and jump relationships between multiple learning nodes; A second determination module, configured to determine a cognitive matching parameter of the knowledge point based on the cognitive style vector and the knowledge point attribute vector of each knowledge point in the knowledge point set, and determine the presentation form of the knowledge point according to the cognitive matching parameter; A generation module, configured to generate a personalized intelligent education textbook through a preset large model based on the learning path map and the presentation form.

[0014] In a third aspect of the present application, an electronic device is provided, including a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are both used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes the method as described in any one of the above.

[0015] In a fourth aspect of the present application, a computer-readable storage medium is provided. The computer-readable storage medium stores instructions, and when the instructions are executed, the method as described in any one of the above is executed.

[0016] In summary, one or more technical solutions provided by the present application have at least the following technical effects or advantages: 1. By using the learning data and eye-tracking data of learners, the cognitive style vector of learners is determined, so as to accurately depict the individual characteristics of learners in the information processing process, enabling the system to construct a personalized learning path map based on the cognitive style vector, and further determining the cognitive adaptation parameters and presentation forms of knowledge points, achieving a high degree of matching between the teaching material content and the cognitive characteristics of learners, effectively improving the pertinence and adaptability of learning content. At the same time, combined with the learning objectives, the system organizes the logical structure between knowledge points based on the learning path map, clarifies the jump relationship of learning nodes, ensures that the arrangement of teaching material content conforms to the cognitive habits and learning progress of learners, and enhances the scientificity and coherence of the learning path. On this basis, a personalized intelligent education teaching material is generated using a pre-set large model, dynamically adapting to the needs of different cognitive styles and learning objectives, thereby improving the learning efficiency of students.

[0017] 2. By extracting behavioral feature data such as page stay duration, repeated access frequency, and interaction operation sequence from the learning data, and combining the saccade rate, fixation transfer pattern, and pupil response characteristics in the eye-tracking data, the content interaction characteristics and cognitive processing characteristics of learners when processing the first knowledge point are comprehensively identified. On this basis, the corresponding relationship between content interaction characteristics and cognitive processing characteristics under different types of knowledge points is further analyzed, and an information processing feature set of learners is constructed to effectively depict the information acquisition and processing behaviors of learners in the learning process. Subsequently, based on the information processing feature set, the system quantitatively obtains cognitive feature parameters such as information acquisition intensity score, information processing depth score, knowledge construction mode index score, and skill transfer efficiency score, and calculates the dimensional feature values of different dimensions accordingly, and finally forms a multi-dimensional cognitive style vector. Through the construction of the cognitive style vector, not only the modeling accuracy of the system for learners' cognitive characteristics is improved, but also a reliable basis is provided for subsequent learning path planning and personalized teaching material generation, thus achieving a deep match between the teaching material content and learners' cognitive habits.

[0018] 3. By combining the learning objectives, the second knowledge points that meet the requirements are screened out from the knowledge point set, and the logical dependence relationship between the second knowledge points is determined based on the pre-set knowledge graph to establish a preliminary learning node system. On this basis, the cognitive style vector of learners is further introduced. By analyzing the dimensional feature values of the cognitive style vector, the cognitive load values corresponding to each learning node are calculated, and at the same time, a knowledge association intensity matrix between learning nodes is constructed to quantify the knowledge relationship intensity between nodes. Combining the cognitive load value and the association intensity, the system can dynamically calculate the jump probability between learning nodes, thereby generating a learning path map that includes the main learning path and alternative learning paths. The learning path map not only reflects the logic of the knowledge structure, but also fully adapts to the cognitive characteristics of learners, achieving the personalization and adaptability of path organization. Thus, while ensuring the logical rigor between knowledge points, this solution improves the matching degree of path generation to learners' cognitive abilities. Description of the Drawings

[0019] Figure 1 is a schematic diagram of the system architecture of an embodiment of a personalized intelligent education textbook generation method or a personalized intelligent education textbook generation system based on a large model in an embodiment of the present application; Figure 2 is a schematic flowchart of a personalized intelligent education textbook generation method based on a large model in an embodiment of the present application; Figure 3 is another schematic flowchart of a personalized intelligent education textbook generation method based on a large model in an embodiment of the present application; Figure 4 is a schematic diagram of the structure of a personalized intelligent education textbook generation system based on a large model in an embodiment of the present application; Figure 5 is a schematic diagram of the structure of an electronic device in an embodiment of the present application.

[0020] Description of the reference numerals: 401, acquisition module; 402, first determination module; 403, construction module; 404, second determination module; 405, generation module; 501, processor; 502, communication bus; 503, user interface; 504, network interface; 505, memory. Detailed Embodiments

[0021] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this specification. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments.

[0022] In the description of the embodiments of the present application, words such as "for example" or "for instance" are used to indicate examples, illustrations or explanations. Any embodiment or design solution described as "for example" or "for instance" in the embodiments of the present application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Rather, the use of words such as "for example" or "for instance" is intended to present related concepts in a specific manner.

[0023] In the description of the embodiments of the present application, the term "plural" means two or more. For example, plural systems refer to two or more systems, and plural screen terminals refer to two or more screen terminals. In addition, the terms "first" and "second" are used only for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of such features. The terms "comprise", "include", "have" and their variants all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0024] Figure 1 An exemplary system architecture 100 is shown, which can apply an embodiment of a large model-based personalized intelligent education textbook generation method or a large model-based personalized intelligent education textbook generation system of the present application.

[0025] As Figure 1 shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, etc.

[0026] Users can use the terminal devices 101, 102, 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications may be installed on the terminal devices 101, 102, 103, such as model training applications, video recognition applications, web browser applications, social platform software, etc.

[0027] The terminal devices 101, 102, 103 may be hardware or software. When the terminal devices 101, 102, 103 are hardware, they may be various electronic devices with a display screen, including but not limited to smart phones, tablet computers, e-book readers, MP3 (Moving Picture Experts Group Audio Layer III) players, MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers, and desktop computers, etc. When the terminal devices 101, 102, 103 are software, they may be installed in the above-listed electronic devices. The electronic devices may be implemented as multiple software or software modules (such as multiple software or software modules for providing distributed services), or may be implemented as a single software or software module. No specific limitation is made here.

[0028] When the terminals 101, 102, and 103 are hardware, a video capture device can also be installed thereon. The video capture device can be various devices that can implement the function of capturing video, such as cameras, sensors, and so on. Users can use the video capture devices on the terminals 101, 102, and 103 to capture video.

[0029] The server 105 can be a server that provides various services, such as a background server for processing the data displayed on the terminal devices 101, 102, and 103. The background server can analyze and process the received data, and can feedback the processing results (such as recognition results) to the terminal devices.

[0030] It should be noted that the server can be hardware or software. When the server is hardware, it can be implemented as a distributed server cluster composed of multiple servers, or as a single server. When the server is software, it can be implemented as multiple software or software modules (such as multiple software or software modules for providing distributed services), or as a single software or software module. No specific limitation is made here.

[0031] It should be understood that Figure 1 the numbers of the terminal devices, the network, and the server in

[0032] Figure 2 are only illustrative. According to the implementation requirements, there can be any number of terminal devices, the network, and the server. In particular, in the case where the target data does not need to be obtained remotely, the above system architecture may not include the network, but only includes terminal devices or servers.

[0033] Please refer to Figure 2 , a personalized intelligent education textbook generation method based on a large model in an embodiment of the present application is applied to a server, and the method includes: S201. Obtain a knowledge point set, learning data of a learner, a learning goal, and eye movement tracking data; In step 201, the server first receives learner data uploaded from a learner client terminal, and the learner data includes a knowledge point set, learning data, a learning goal, and eye movement tracking data.

[0034] Specifically, the knowledge point set is a part of a knowledge graph pre-constructed and maintained by the system, and the knowledge graph covers knowledge units of different granularities in multiple subject fields. In the implementation process, the server filters out a knowledge point set related to the learning topic according to the current learning task or course module, providing a knowledge basis for subsequent construction of a learning path map and matching of cognitive styles.

[0035] When obtaining the learning data of learners, considering that there may be learners who are using the system for the first time and have not yet formed complete learning behavior data, a dual-path data completion strategy is adopted in this embodiment: for users with existing historical learning records, the historical interaction behavior characteristics of the learners are directly extracted from the learning data of the learners, including page residence time, module access frequency, interaction click sequence, scroll depth, etc. These data can be used to extract the content interaction characteristics of the learners when processing specific knowledge points in the future; for learners using the system for the first time, information such as the initial learning preferences, learning style tendencies, attention distribution habits, and acceptance levels of different teaching content forms of the learners is obtained by guiding the learners to fill out a structured questionnaire. Then, the questionnaire data is used as the initial learning data input to form a preliminary learner behavior portrait. The questionnaire design is based on existing educational psychology models and cognitive style classification criteria to ensure the usability and representativeness of the initial data of the learners, and it is dynamically supplemented and corrected through continuous collection of real-time interaction behaviors during the subsequent learning process.

[0036] In addition, the system also obtains the learning goals set by the learners currently. The learning goals can be actively input by the learners or generated by the system in combination with the learners' course progress and assessment results. Synchronously, the server also receives the eye movement tracking data collected by the learners through the eye movement tracking device during the learning process. The data includes saccade rate, fixation transfer pattern, pupil response characteristics, etc., which are used to identify the cognitive processing characteristics of the learners during the knowledge acquisition process.

[0037] S202. Determine the cognitive style vector of the learner according to the learning data and the eye movement tracking data. The cognitive style vector is used to represent the information processing preference characteristics of the learner; In step S202, the system jointly analyzes the learning data and the eye movement tracking data of the learner, extracts the behavior characteristics and cognitive processing characteristics of the learner during the learning process, constructs a multi-dimensional information processing feature set based on a preset behavior-cognition association model, and finally generates a cognitive style vector for characterizing the information processing preference characteristics of the learner, providing a cognitive basis for the subsequent construction of the learning path and the personalized matching of the knowledge point presentation form.

[0038] For details, please refer to Figure 3 , Figure 3 which is another process schematic diagram of a personalized intelligent education textbook generation method based on a large model in the embodiment of the present application.

[0039] S301. Extract behavioral feature data from the learning data; To model the content interaction features in the learner's cognitive style vector, the system extracts the behavioral feature data from the acquired learning data. The learning data contains the interaction information generated by the learner when completing various learning tasks on the intelligent education platform. The interaction information can reflect the operation mode, learning rhythm, and content attention method of the learner when facing different knowledge points, and has obvious individual differences. Therefore, it can be used as an important basis for identifying information processing preference features.

[0040] In specific operations, the server first performs a structured analysis on the learning data, identifies the behavioral events related to knowledge point interactions, and based on the set behavioral feature extraction model, performs quantitative calculations on the following key dimensions: First, the page residence time, by counting the average residence duration of the learner on each knowledge point content page, reflects the attention depth of the learner to different knowledge contents; Second, the click behavior feature, by analyzing the frequency and distribution position of click operations within the page, judges the usage preference and exploration tendency of the learner for modular content; Third, the scroll behavior feature, by calculating the page scroll depth and scroll frequency, infers the attention degree and browsing method of the learner to information integrity; Fourth, the jump path and repeated access behavior, by identifying the jump sequence between knowledge points of the learner and whether there is a situation of accessing the same content multiple times, evaluates the knowledge integration strategy of the learner and the recognition requirement for specific content. Among them, the behavioral feature extraction model is obtained by training with a large amount of historical learning data and expert-annotated samples. During the modeling process, the system combines machine learning algorithms (such as random forest, support vector machine, or convolutional neural network) to model the correlation between different behavioral features and the learner's cognitive performance, and determines the key feature dimensions and extraction methods after multiple rounds of verification and optimization.

[0041] S302. Based on the page residence duration, repeated access frequency, and interaction operation sequence in the behavioral feature data, identify the content interaction features of the learner when processing the first knowledge point, where the first knowledge point is any type of knowledge point; In step S302, to accurately identify the content interaction features of the learner when processing the first knowledge point, the system analyzes the interaction behavior of the learner by calling the interaction feature recognition model based on the page residence duration, repeated access frequency, and interaction operation sequence in the behavioral feature data. The interaction feature recognition model is obtained by training with a large amount of historical learning log data. During the model construction process, the system introduces expert-annotated interaction feature samples and uses supervised learning algorithms (such as random forest or long short-term memory network) to model the mapping relationship between various behavioral features and specific interaction feature labels, and finally obtains a prediction model that can identify the interaction features of any knowledge point.

[0042] In specific implementation, the system first extracts the page residence duration, repeated access frequency, and complete interaction operation sequence associated with the first knowledge point from the behavioral feature data, and performs normalization processing on the above data. Subsequently, the system combines the preset interaction feature recognition model to comprehensively analyze the page residence duration, repeated access frequency, and complete interaction operation sequence: if the page residence duration is significantly higher than the overall average, and there are multiple repeated access behaviors, and high-complexity patterns such as "jump - return - enter again" or "browse - pause - scroll - click" appear in the operation sequence, then the system recognizes that the learner has content interaction characteristics with high processing depth and active exploration tendency on the knowledge point. On the contrary, if the residence time is short, there is a lack of repeated access, and the operation sequence is relatively simple, it is recognized as a quick browsing type of interaction characteristic.

[0043] S303. Determine the cognitive processing characteristics during the learning process according to the saccade rate, fixation migration pattern, and pupil response characteristics in the eye movement tracking data; In this embodiment, to further explore the information processing method and cognitive load level of the learner during the learning process, the system determines the cognitive processing characteristics of the learner during the learning process according to the saccade rate, fixation migration pattern, and pupil response characteristics in the eye movement tracking data. Eye movement behavior is widely regarded as an explicit cognitive processing index, which can relatively truly reflect the mental state and processing depth of an individual in the links of information reception, understanding, integration, etc. Therefore, analyzing the learner can provide key support for constructing a refined cognitive style vector.

[0044] Specifically, the system first uses the eye movement acquisition device to obtain the eye movement tracking data of the learner in real time during the completion of the learning task, and performs corresponding matching with the behavioral feature data based on the time synchronization mechanism. Subsequently, the system calls the pre-trained cognitive processing feature recognition model to perform combined analysis on the saccade rate, fixation migration pattern, and pupil response characteristics. The cognitive processing feature recognition model is constructed based on the existing eye movement experiment data, and the eye movement feature annotations under different cognitive load levels are included in the training samples. The system captures the temporal change and spatial distribution law of the eye movement features by introducing a fusion architecture of convolutional neural network and recurrent neural network, and then realizes the automatic recognition and classification of the cognitive processing features.

[0045] In practical applications, if the saccade rate of the learner is low, the fixation migration path is concentrated, and the pupil shows a continuous enlargement trend, the system accordingly recognizes that the learner is in a high cognitive load state, and the cognitive processing characteristics are manifested as "deep integration type"; on the contrary, if the saccade rate is high, the fixation jumps frequently, and the pupil change amplitude is small, it is recognized as the cognitive processing characteristics of "surface browsing type". The recognition result not only provides support for the multi-dimensional modeling of the subsequent cognitive style vector in the physiological dimension, but also provides an important perceptual basis for the dynamic adjustment of personalized learning resources.

[0046] S304. Determine the corresponding relationship between the content interaction characteristics and cognitive processing characteristics of learners when learning different types of knowledge points through a preset behavior-cognition association model, and obtain the information processing characteristic set of the learners. To accurately depict the information processing method of learners, the system determines the corresponding relationship between the content interaction characteristics and cognitive processing characteristics of learners when learning different types of knowledge points through a preset behavior-cognition association model, and generates an information processing characteristic set accordingly. Step S304 aims to integrate the data at the behavioral layer and the physiological layer, construct a unified individual cognitive processing portrait, and improve the accuracy of subsequent cognitive style modeling and the matching degree of personalized teaching resources.

[0047] Specifically, the system first retrieves the content interaction characteristics and cognitive processing characteristics of the identified learners on multiple knowledge points and classifies and summarizes them according to the knowledge point types. Subsequently, the system calls the preset behavior-cognition association model to model the corresponding relationship between the interaction characteristics and cognitive processing characteristics presented in each type of knowledge point. The behavior-cognition association model is trained based on a joint sample of large-scale learning behaviors and eye movement data. By introducing a graph neural network or a multimodal alignment learning algorithm, a similarity mapping and semantic correspondence relationship between interaction behaviors and eye movement characteristics is established. During the model training process, expert-annotated high-cognitive-consistency samples are introduced as supervision signals to improve the generalization ability and cross-knowledge-type adaptability of the model.

[0048] During operation, when the system identifies that the content interaction characteristic under a certain knowledge point is "key point repeated type" and the corresponding cognitive processing characteristic is "deep integration type", the corresponding relationship can be incorporated into the information processing characteristic set of the learners. By establishing such corresponding relationships on multiple knowledge points, the system finally forms an information processing characteristic set containing multiple groups of interaction-cognition pairing patterns, which is used to reflect the processing methods generally adopted by learners under different knowledge structures.

[0049] S305. Analyze the information processing characteristic set to obtain the information acquisition intensity score, information processing depth score, knowledge construction mode index score, and skill transfer efficiency score of the learners. After obtaining the information processing characteristic set of the learners, the system analyzes the performance of the learners in the cognitive processes of knowledge acquisition, processing, construction, and transfer based on the information processing characteristic set, and further generates an information acquisition intensity score, an information processing depth score, a knowledge construction mode index score, and a skill transfer efficiency score, which are used to quantify the information processing ability of the learners in different cognitive dimensions. Among them, the knowledge construction mode index score is a structured scoring result, which includes knowledge index scores of multiple sub-dimensions and is used to depict the specific ability performance of the learners in knowledge organization and structured processing.

[0050] Among them, the information acquisition intensity score is used to quantify the ability and willingness of learners to actively acquire information during the learning process. The information acquisition intensity score is calculated from behavioral characteristics such as the content browsing density, page jump breadth, gaze distribution range, and information collection time shown by learners in the task. A high score usually indicates that learners have a high degree of exploration and external information dependence, reflecting that the cognitive style of learners tends to be "perceptive" or "empirical".

[0051] The information processing depth score represents the ability and tendency of learners to deeply process the received information. The information processing depth score is generated based on characteristics such as the gaze concentration, pupil dilation persistence, semantic backtracking behavior, and implicit reasoning paths between knowledge points shown by learners in the task. Those with high scores usually have strong abstract induction abilities and tend to have a "reflective" or "structured" cognitive style.

[0052] The knowledge construction mode index score is used to evaluate the performance of learners in organizing, integrating, and reconstructing knowledge content. The knowledge construction mode index score is not a single value but consists of multiple sub-indicators, including but not limited to: Structure perception ability score: Measures the ability of learners to recognize and utilize the knowledge structure; Generic integration ability score: Measures the ability of learners to classify and integrate across knowledge points; Concept abstraction ability score: Judges whether learners can form high-order concepts from examples.

[0053] The system generates the knowledge construction mode index score through weighted integration of these sub-scores, which is used to reflect whether the knowledge construction method of learners tends to be "networked", "linear", or "concept mapping type".

[0054] The skill transfer efficiency score is used to measure the ability of learners to transfer existing knowledge or skills to new tasks and new scenarios. The score is generated based on indicators such as the similarity between the strategies adopted by learners in the new task and the learned strategies, the number of transfer attempts, the transfer success rate, and the strategy stability. Those with high scores usually have a "transferable" cognitive style and are characterized by strong strategy generalization ability and outstanding problem-solving ability.

[0055] During the implementation process, the system first aggregates the interaction features and cognitive processing features in the information processing feature set according to knowledge points, and extracts highly relevant feature variables for scoring, including the hierarchical depth of the page access path, the jump frequency between similar knowledge points, the occurrence times of content integration behaviors, and the continuity of pupil response and fixation pattern. Subsequently, the system calls the pre-trained scoring model to model the four types of scoring dimensions respectively. The pre-trained scoring model is designed by experts with a scoring index system, and is obtained through the fusion training of multi-factor regression and fuzzy logic rules by combining real learning logs and eye movement data samples, and has strong interpretability and scalability.

[0056] For the scoring of the knowledge construction mode index, the system divides the scoring of the knowledge construction mode index into multiple sub-index scores, including knowledge structure perception score, generic integration ability score, cross-topic association score, and concept abstraction ability score, etc. Each sub-score is independently calculated by the corresponding feature combination input into the model and then weighted and summarized. For example, if a learner has a high-frequency cross-level jump behavior between multiple knowledge points, and the interaction operation shows "first scan widely - then focus - then integrate", and the fixation path shows obvious clustering and backtracking characteristics, then the system determines that the learner has a high generic integration ability score and knowledge structure perception score. If a learner shows active invocation and analogy behavior of old knowledge points when dealing with new knowledge points, the concept abstraction ability score will increase accordingly. Finally, the system outputs the scoring results of the learner under the four main dimensions, and attaches each sub-score item in the scoring of the knowledge construction mode index to form a complete quantitative index of information processing ability.

[0057] S306. Calculate the dimension feature value according to the cognitive feature parameter to obtain the cognitive style vector, where the cognitive feature parameter includes the information acquisition intensity score, the information processing depth score, the knowledge construction mode index score, and the skill transfer efficiency score.

[0058] In this embodiment, after obtaining the cognitive feature parameters of the learner, the system generates a cognitive style vector with dynamic evolution ability by constructing a cognitive evolution map and performing structured graph embedding processing. It breaks through the traditional static scoring and weighting method, and can express the evolution trend of the learner's cognitive strategy at different learning stages, so as to more accurately depict the deep structure of the individual cognitive style.

[0059] Cognitive characteristic parameters may include: information acquisition intensity score (reflecting the learner's active information acquisition ability), information processing depth score (measuring the learner's ability to perform in-depth processing such as reasoning, integration, and abstraction on information), knowledge construction mode index score (indicating the learner's performance in knowledge organization and structural integration, which includes multiple sub-dimension scores, such as structure perception ability, generic integration ability, etc.), and skill transfer efficiency score (reflecting the efficiency and strategy stability of the learner in transferring the learned knowledge to new tasks). In this embodiment, the introduced Cognitive Evolution Graph refers to a model that represents the evolution of a learner's cognitive ability over the learning process in the form of a graph structure. The Cognitive Evolution Graph is composed of multiple stage-based cognitive state nodes and the transition relationships between the cognitive state nodes, and is used to express the change process of the learner's cognitive behavior on multiple tasks and multiple knowledge points. The Cognitive Evolution Graph G can be expressed as: G=(V, E), where V is the node in the Cognitive Evolution Graph, representing the learner's cognitive state within a preset time period. E is the edge of the Cognitive Evolution Graph, representing the transition relationship between two cognitive states, and recording the directionality and amplitude of the changes in each score during the transition process.

[0060] Subsequently, the system inputs the constructed Cognitive Evolution Graph into the graph embedding model for high-dimensional vectorization processing. The adopted embedding model is a Structure-aware Graph Neural Network, and the core is to simultaneously consider node attributes, edge weight semantics, and local sub-graph structures, so as to capture the multi-scale evolution law of cognitive behavior. Map the Cognitive Evolution Graph to the cognitive style vector V CS : V CS =f GraphEmbed (G; θ), where f GraphEmbed is the graph embedding function, and θ is the model parameter, which is obtained through training with a large number of cognitive evolution samples.

[0061] S203. According to the cognitive style vector and the learning objective, construct a learning path graph for the learner for the knowledge point set. The learning path graph includes learning nodes and the jump relationships between multiple learning nodes; According to the learning objective, screen out the second knowledge points that meet the preset requirements from the knowledge point set, and use the second knowledge points as learning nodes; based on the preset knowledge graph, determine the logical dependency relationship between the second knowledge points; based on the cognitive style vector and the logical dependency relationship, determine the jump relationship between the learning nodes to obtain the learning path graph.

[0062] In this embodiment, in order to achieve personalized learning path planning that is more in line with the individual cognitive characteristics of learners, the system selects the second knowledge point that meets the preset requirements from the knowledge point set based on the learning objectives, and uses the knowledge point corresponding to the learning objective as the starting node (i.e., learning node) of the learning path map, and then combines the logical dependency relationship of the knowledge points constructed in the knowledge graph with the learner's cognitive style vector to dynamically construct an adaptive learning path map.

[0063] First, the system receives the learning goal set for the current learning task. The learning goal is usually set by the teacher or learner or automatically generated by the system according to the course progress. The specific form is a unique identifier corresponding to the knowledge point. The knowledge point is called a learning node in this embodiment. In order to clarify all the knowledge content involved in the learning task, the system selects a second knowledge point from the preset knowledge point set that has a direct or indirect logical association with the learning goal and meets the task granularity and cognitive span requirements.

[0064] During the screening process, the system can screen candidate knowledge points based on judgment criteria in two dimensions to determine whether to include the knowledge points in the subset of knowledge points involved in the current learning task. Among them, the first dimension is the knowledge path judgment at the structural level, that is, to determine whether there is a valid logical path relationship between the candidate knowledge point and the second knowledge point in the knowledge graph. Based on the pre-built knowledge graph structure, the system performs path retrieval operations on each pair of candidate knowledge points and target knowledge points, using a depth-first or breadth-first search algorithm, limiting the maximum path length, and determining whether the following two types of paths exist: 1. The candidate knowledge point is a preceding node of the target knowledge point, and there is a direct or indirect preceding dependency path; 2. The candidate knowledge point is a subsequent expansion node of the target knowledge point, and can logically undertake the learning outcomes of the target knowledge point. If any of the path existence conditions are met, the system regards the structural relationship as meeting the requirements, where the target knowledge point is any second knowledge point.

[0065] The second dimension is the compatibility judgment of cognitive load and cognitive style. The system first calls the cognitive load prediction model to calculate the cognitive load value of the candidate knowledge point under the specific learner's cognitive style vector. The cognitive load prediction model can use a multi-input regression network, and the input includes the knowledge point attribute vector (complexity, abstract level, number of pre-dependencies, etc.) of the candidate knowledge point and the current learner's cognitive style vector (such as processing depth preference, transfer ability, structural integration ability, etc.). The cognitive load value output by the model will be compared with the cognitive load adaptation threshold range preset by the system. If the cognitive load value of the candidate knowledge point is within the current acceptable range of the learner (that is, it does not exceed the upper limit of the learner's cognitive processing ability, nor is it lower than the lower limit of the learner's excitation threshold), it is considered that the cognitive adaptability meets the requirements.

[0066] Only when the candidate knowledge points simultaneously meet the above two judgment conditions, that is: (1) there is an effective path relationship with the target knowledge point in the knowledge graph; (2) the cognitive load under the current learner's cognitive style condition is within the adaptation threshold range, will the system include the candidate knowledge points in the subset of knowledge points involved in this task as one of the basic nodes for subsequent path graph construction and teaching material generation.

[0067] After completing the screening of learning nodes and the extraction of logical dependency relationships, in order to further construct a personalized learning path graph that adapts to the individual cognitive characteristics of learners, the system determines the jump relationships between learning nodes based on the cognitive style vector and the logical dependency relationships extracted from the knowledge graph, and generates a learning path graph including a core path and auxiliary paths accordingly.

[0068] Specifically, according to the dimensional characteristic values of the cognitive style vector, calculate the cognitive load value of each learning node; construct a knowledge association strength matrix between learning nodes; combine the cognitive load value and the knowledge association strength matrix to calculate the jump probability between learning nodes; construct a learning path graph according to the jump probability, and the learning path graph includes a necessary learning path and an alternative learning path.

[0069] In this embodiment, in order to achieve a deep adaptation of the learning path to the individual differences of learners, the system further introduces a cognitive style vector on the basis of the existing learning node set and the logical dependency relationship of the learning node set, and calculates the jump relationship between learning nodes in a model-based manner, so as to construct a learning path graph with both knowledge logic and cognitive adaptability.

[0070] First, based on the cognitive style vector of the learner, the system calculates the cognitive load value corresponding to each learning node in the current learning task through the constructed cognitive load prediction model. The cognitive load prediction model is a multi-input feature-driven model, and the inputs include the dimensional values of the learner's cognitive style vector (such as information processing depth, knowledge construction preference, transfer efficiency, etc.), as well as the structural characteristics of the learning node itself, such as content complexity, knowledge abstraction level, prior knowledge requirements, etc. The model can adopt a multi-layer perceptron, a graph neural network or a structure based on the attention mechanism to fuse the input features and output the cognitive load value of each node under a specific learner. Compared with the traditional static estimation method, the cognitive load prediction model has the ability to respond to cognitive dynamic changes and higher prediction accuracy.

[0071] Subsequently, the system constructs a knowledge association strength matrix between learning nodes based on the knowledge graph structure. The system generates the association strength value between node pairs through a weighted calculation method based on various factors such as the topological position, connection path length, semantic label similarity, co-occurrence frequency, etc. of two learning nodes in the knowledge graph. The association strength represents the tightness of the two nodes in knowledge logic and serves as a basic reference for the feasibility of path jumps.

[0072] After obtaining the cognitive load values of each node and the knowledge association strength between node pairs, the system further calculates the jump probabilities between learning nodes through a jump probability modeling module. The jump probability modeling module takes the knowledge association strength and the cognitive load of the target node as joint inputs based on a heuristic strategy or a trained scoring model, and outputs jump probability values. Specifically, the jump probability modeling module tends to assign higher jump probabilities to jump paths with "high knowledge association and moderate cognitive load", while automatically reducing the jump possibility for paths with too high a load on the target node or too low an association degree. This can effectively control the density distribution of high-load nodes in the path and improve the cognitive bearability of the overall path.

[0073] After completing the jump probability calculation, the system constructs a learning path graph according to the jump probabilities. The learning path graph takes the learning nodes corresponding to the learning objectives as the starting point and constructs a directed connection relationship between nodes according to the size of the jump probabilities, forming a structured path graph.

[0074] In the process of generating the learning path graph, the system screens and classifies the connection relationships between learning nodes according to the values of the jump probabilities. To ensure the controllability and cognitive adaptability of the path structure, the system sets a jump probability threshold interval and divides it into "necessary learning paths" and "alternative learning paths" accordingly.

[0075] Among them, the system regards the paths with jump probabilities greater than or equal to the main threshold as paths that meet the main conditions and defines them as necessary learning paths. The main threshold is set by the system according to empirical rules, training data statistics, or task scenarios, and is usually set as a fixed value or a dynamic value within the range of 0.65 to 0.8, indicating that the jump relationship has high adaptability and execution feasibility under the current cognitive style and knowledge logic. The necessary paths form the main guiding paths in the learning path graph and are the path structures that the system preferentially recommends in the teaching task.

[0076] At the same time, the system marks the paths with jump probabilities lower than the main threshold but higher than the secondary threshold (such as in the range of 0.4 - 0.65) as alternative learning paths. Although the alternative learning paths are slightly inferior to the main paths in terms of adaptability, they still have certain logical feasibility and cognitive bearability, and are usually used in the following scenarios: one is to replace the path when the learner encounters cognitive blockage during the execution of the main path task; the second is for guiding strategy diversity to support different-style learners in path exploration; the third is as the basis for path reconstruction of the cognitive intervention module to dynamically adjust the learning guidance strategy.

[0077] S204. Determine the cognitive matching parameters of the knowledge points based on the cognitive style passing and the knowledge point attribute vectors of each knowledge point in the knowledge point set, and determine the presentation form of the knowledge points according to the cognitive matching parameters; Feature match the cognitive style vector and the knowledge point attribute vector based on a preset feature matching rule to obtain a feature matching matrix; calculate a cognitive matching parameter according to the feature matching matrix; generate a presentation form label based on the cognitive matching parameter, and determine the presentation form of the knowledge point according to the presentation form label.

[0078] In this embodiment, to achieve dynamic adaptation between the knowledge point presentation method and the learner's individual cognitive characteristics, the system proposes a feature matching mechanism based on the cognitive style vector and the knowledge point attribute vector, thereby determining the cognitive matching parameter of each knowledge point, and generating an adapted presentation form label based on this, and finally realizing the display of knowledge content in a way that conforms to the learner's cognitive preference, which can effectively improve the knowledge transfer efficiency, enhance the immersion and acceptance of the learning process, and avoid learning obstacles or efficiency losses caused by the mismatch between the content presentation method and the learner's cognitive style.

[0079] The cognitive style vector consists of multiple dimensions, usually including information acquisition method preference (such as image type or symbol type), information processing depth (such as shallow processing or structural processing), knowledge construction tendency (such as inductive or deductive), and transfer strategy characteristics, etc. At the same time, the system also extracts the knowledge point attribute vector corresponding to each knowledge point from the knowledge point set, and the knowledge point attribute vector reflects the inherent characteristics of the knowledge point in terms of content form, structural complexity, abstraction level, carrier type (such as text, diagram, animation, etc.).

[0080] The system matches the cognitive style vector and the knowledge point attribute vector according to the preset feature matching rule to form a feature matching matrix. The feature matching matrix is used to quantify the adaptation degree of the learner's cognitive tendency and the knowledge point content characteristics in each dimension. Among them, each dimension matching is calculated using a defined matching function, and the matching function can be: cosine similarity function, inverse Euclidean distance function, discrete rule mapping function, etc. For example, in the dimension of "information acquisition preference" and "content carrier type", if the learner prefers the image type and the knowledge point carrier is a graphical structure, the output value of the dimension matching function is high matching (such as 0.9); otherwise, it is low matching (such as 0.3). The system completes the matching for all dimensions to form an N×1 feature matching matrix.

[0081] The system further normalizes the feature matching matrix, and combines the weight factors of each dimension (learned from historical data or configured by instructional design experts), and performs weighted aggregation to obtain a cognitive matching parameter representing the overall adaptation degree. This parameter is a scalar value, usually in the range of [0, 1]. The higher the value, the more matching the current knowledge point is presented in its original carrier form with the learner's cognitive characteristics, and the higher the cognitive friendliness.

[0082] Subsequently, the system sets multiple cognitive matching threshold intervals by combining historical learning data analysis and cognitive psychology theory to distinguish the adaptation degree between learners at different levels and knowledge points. First, the system uses clustering analysis and piecewise regression methods to identify key turning points based on the relationship between cognitive matching parameters and learning effect performance (such as accuracy rate, learning duration) in a large number of samples, and divides them into three intervals: low matching, medium matching, and high matching; Subsequently, combined with the "optimal learning interval" principle in cognitive load theory, the interval boundaries are optimized and adjusted to finally form a stable threshold interval configuration. For example, the low matching interval is set as [0, 0.45), the medium matching interval is [0.45, 0.75), and the high matching interval is [0.75, 1].

[0083] According to the cognitive matching threshold interval of cognitive matching parameters, the system generates corresponding presentation form labels in cooperation with the preset presentation form decision rules. The construction method of the decision rules is as follows: on the one hand, based on a large number of historical learning records, extract the mapping relationship between "cognitive style + knowledge point attributes → presentation effect", and train a candidate form recommendation model; on the other hand, combined with cognitive psychology principles and the experience of teaching experts, proofread and summarize to form a set of structured decision rule tables. Example rules include: when the cognitive matching parameter ≥ 0.85 and the weight of the image preference dimension is greater than 0.4, generate the "structural diagram" label; when the matching parameter is in the intermediate value interval and the processing depth dimension is relatively high, generate the "context case" or "reasoning and deduction" label.

[0084] After generating the presentation form labels, the system searches for matching content templates or material modules in the multimodal content resource library. Before the system is deployed, this resource library is composed of two types of content construction methods: one is structured collection from standard textbooks and online course resources, and the one-to-one correspondence between content granularity and knowledge points is established through manual annotation; the other is generative content driven by content templates, such as "structural diagram templates", "animation explanation script templates", etc., which automatically generate graphic or script content in combination with the semantics of knowledge points. The system calls the corresponding templates or resource modules according to the label matching, automatically generates or loads the presentation content, and finally determines the personalized presentation form of this knowledge point.

[0085] S205. Generate personalized intelligent education textbooks based on the learning path map and presentation form.

[0086] In this embodiment, to achieve the deep integration of teaching content and learners' cognitive characteristics, based on the construction of the learning path map and the determination of the presentation form of knowledge points, the system further introduces a pre-set large model. Through the content understanding and reconstruction ability of the large-scale generative model, it automatically generates a personalized intelligent education textbook that is consistent with the current learning path and adapts to the content expression method. The core of the method is to fuse the structured path guidance information and personalized content presentation preferences and input them into the large model, realizing the intelligent generation from "knowledge structure + presentation method" to "complete textbook content", significantly improving the efficiency and adaptability of textbook development.

[0087] Specifically, the system first obtains the learning path map that should be covered at the current stage of the target learner. The learning path map has been generated according to the cognitive style vector and the jump probability model, and includes the logical sequence and hierarchical structure of learning nodes. Each learning node corresponds to a specific knowledge point, and for each knowledge point, the corresponding presentation form label has been determined according to the aforementioned cognitive matching parameters, indicating the content expression strategy that the system should adopt, such as text explanation, graphical illustration, scenario case, or animation guidance, etc.

[0088] The system encodes the learning path map and the presentation form information structurally and inputs them as guiding prompts into the pre-set large model. The large model is a pre-trained generative model with large-scale language generation ability, which can automatically complete the organization, expression, and personalized presentation of textbook content based on the input structured path and label information. In specific implementation, the system constructs a multi-modal prompt template for each learning node, including the following contents: the path position where the learning node is located, the front and back dependencies, the knowledge attributes of the knowledge point, the cognitive load level, the recommended presentation form label, etc. The system inputs the prompt template into the large model and generates a structurally reasonable and appropriately expressed knowledge content segment through the context understanding and content generation ability of the model.

[0089] During the generation process, the large model not only refers to the information of the current node but also combines the content of the previous nodes in the path map, making the generated content have logical coherence and knowledge progression. At the same time, according to the presentation form label, the large model automatically selects a suitable text style and expression method. For example, it generates a content segment with vivid illustrations and texts under the graphic illustration label, and generates a contextualized teaching content close to the learners' experience under the case label, ensuring that the content expression method is consistent with the learners' cognitive preferences.

[0090] The system integrates, typesets, and optimizes the generated content segments of each knowledge point, and finally forms a complete personalized intelligent education textbook. The textbook strictly follows the node sequence defined by the learning path map in terms of content order, and strictly matches the presentation form label of each node in terms of expression method, achieving a high degree of integration of content structure rationality, expression method personalization, and cognitive load adaptability as a whole.

[0091] In some embodiments, the system further includes a knowledge mastery evaluation module and a learning behavior prediction module, which are used to further improve the diagnostic accuracy and intervention prospectiveness in the personalized teaching process. Among them, the knowledge mastery evaluation module is used to dynamically evaluate the mastery degree of learners on each knowledge point based on learning performance information such as interaction data, answer results, and task completion records generated by learners during the execution of the personalized intelligent education teaching materials, combined with the cognitive load levels and jump sequences of each knowledge node in the learning path map. The system divides the mastery degree of knowledge points into multiple levels according to a preset evaluation model, and identifies the weak mastery nodes and advantageous areas in the current learning path, generating knowledge mastery labels for subsequent path adjustment, resource recommendation, and cognitive intervention strategy generation.

[0092] The learning behavior prediction module is used to construct a learning behavior feature vector based on the learner's historical learning path, cognitive style vector, knowledge mastery label, and behavior sequence log, and predict potential problems such as cognitive bottlenecks, path jump anomalies, or learning rhythm fluctuations that the learner may face in the future learning process through an integrated prediction model. The prediction model can adopt a time series modeling architecture, including but not limited to recurrent neural networks, temporal convolutional networks, or Transformer structures, which can predict the subsequent behavior evolution trajectory of the learner based on the existing learning path and performance trends, and output learning risk warning labels.

[0093] In practical applications, the system can jointly determine whether there is an abnormality in the current state of the learner according to the risk label of learning behavior prediction and the knowledge mastery label. If the preset intervention conditions are met, it triggers the dynamic reconstruction module of the path map and the teaching material regeneration module to adjust the subsequent learning path or the form of knowledge presentation, so as to realize the real-time optimization and dynamic adaptation of the learning process.

[0094] Please refer to Figure 4 , which is a schematic structural diagram of a personalized intelligent education teaching material generation system based on a large model provided by an embodiment of this application. A personalized intelligent education teaching material generation system 400 specifically includes: An acquisition module 401, which is used to acquire a knowledge point set, the learning data of learners, learning objectives, and eye movement tracking data; A first determination module 402, which is used to determine the cognitive style vector of the learner according to the learning data and the eye movement tracking data, and the cognitive style vector is used to represent the information processing preference characteristics of the learner; A construction module 403, which is used to construct a learning path map for the learner for the knowledge point set according to the cognitive style vector and the learning objectives, and the learning path map includes learning nodes and jump relationships between multiple learning nodes; The second determination module 404 is configured to determine the cognitive matching parameter of each knowledge point in the knowledge point set based on the cognitive style vector and the knowledge point attribute vector of each knowledge point, and determine the presentation form of the knowledge point according to the cognitive matching parameter; The generation module 405 is configured to generate a personalized intelligent education textbook through a preset large model based on the learning path map and the presentation form.

[0095] Optionally, the first determination module 402 is specifically configured to: extract behavioral feature data from the learning data; based on the page residence time, repeated access frequency, and interaction operation sequence in the behavioral feature data, identify the content interaction feature of the learner when processing the first knowledge point, where the first knowledge point is any type of knowledge point; determine the cognitive processing feature during the learning process according to the saccade rate, fixation transfer pattern, and pupil response feature in the eye movement tracking data; determine the cognitive style vector of the learner according to the content interaction feature and the cognitive processing feature.

[0096] Optionally, the first determination module 402 is further specifically configured to: determine the corresponding relationship between the content interaction feature and the cognitive processing feature of the learner when learning different types of knowledge points through a preset behavior-cognition association model, and obtain the information processing feature set of the learner; determine the cognitive style vector according to the information processing feature set.

[0097] Optionally, the first determination module 402 is further specifically configured to: analyze the information processing feature set to obtain the information acquisition intensity score, information processing depth score, knowledge construction mode index score, and skill transfer efficiency score of the learner; calculate the dimensional feature value according to the cognitive feature parameter to obtain the cognitive style vector, where the cognitive feature parameter includes the information acquisition intensity score, information processing depth score, knowledge construction mode index score, and skill transfer efficiency score.

[0098] Optionally, the construction module 403 is specifically configured to: screen out the second knowledge points that meet the preset requirements from the knowledge point set according to the learning goal, and use the second knowledge points as learning nodes; Based on the preset knowledge graph, determine the logical dependency relationship between the second knowledge points; determine the jump relationship between the learning nodes based on the cognitive style vector and the logical dependency relationship to obtain the learning path map.

[0099] Optionally, the construction module 403 is further specifically configured to: calculate the cognitive load value of each learning node according to the dimensional feature value of the cognitive style vector; construct the knowledge association intensity matrix between the learning nodes; combine the cognitive load value and the knowledge association intensity matrix to calculate the jump probability between the learning nodes; construct the learning path map according to the jump probability, where the learning path map includes the necessary learning path and the alternative learning path.

[0100] Optionally, the second determination module 404 is specifically configured to: perform feature matching on the cognitive style vector and the knowledge point attribute vector based on a preset feature matching rule to obtain a feature matching matrix; calculate a cognitive matching parameter according to the feature matching matrix; generate a presentation form label based on the cognitive matching parameter, and determine the presentation form of the knowledge point according to the presentation form label.

[0101] It should be noted that when the device provided in the above embodiment realizes its functions, only the division of the above functional modules is used for illustration. In actual applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be seen in the method embodiment, which will not be repeated here.

[0102] This embodiment also discloses an electronic device. Referring to Figure 5 , the electronic device may include: at least one processor 501, at least one communication bus 502, a user interface 503, a network interface 504, and at least one memory 505.

[0103] Among them, the communication bus 502 is used to realize the connection and communication between these components.

[0104] Among them, the user interface 503 may include a display screen (Display) and a camera (Camera). Optionally, the user interface 503 may further include a standard wired interface and a wireless interface.

[0105] Among them, the network interface 504 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface).

[0106] Among them, the processor 501 may include one or more processing cores. The processor 501 connects various parts within the entire server through various interfaces and lines. By running or executing instructions, programs, code sets, or instruction sets stored in the memory 505, and by calling the data stored in the memory 505, it performs various functions of the server and processes data. Optionally, the processor 501 may be implemented in at least one hardware form of digital signal processing (DSP), field-programmable gate array (FPGA), or programmable logic array (PLA). The processor 501 may integrate a combination of one or more of a central processing unit (CPU), a graphics processing unit (GPU), and a modem, etc. Among them, the CPU mainly processes the operating system, user interface, application programs, etc.; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communication. It can be understood that the above-mentioned modem may not be integrated into the processor 501 and may be implemented separately by a single chip.

[0107] Among them, the memory 505 may include random access memory (RAM) and may also include read-only memory. Optionally, the memory 505 includes a non-transitory computer-readable storage medium. The memory 505 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 505 may include a program storage area and a data storage area. Among them, the program storage area may store instructions for implementing the operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store the data involved in the above-mentioned various method embodiments. Optionally, the memory 505 may also be at least one storage device located far from the aforementioned processor 501. As Figure 5 shown, the memory 505, as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program for a personalized intelligent education textbook generation method based on a large model.

[0108] In Figure 5In the electronic device shown, the user interface 503 is mainly used to provide an interface for the user to input and obtain the data input by the user; and the processor 501 can be used to call an application program stored in the memory 505 for a personalized intelligent education textbook generation method based on a large model. When executed by one or more processors 501, the electronic device executes the method of one or more of the above embodiments.

[0109] It should be noted that, for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that this application is not limited by the described action sequence, because according to this application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0110] In the above embodiments, the descriptions of the respective embodiments have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0111] In the several embodiments provided by this application, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of units is only a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed mutual coupling, direct coupling or communication connection can be through some service interfaces. The indirect coupling or communication connection of the device or unit can be in an electrical or other form.

[0112] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place, or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0113] In addition, the functional units in each embodiment of this application can be integrated into a processing unit, or each unit exists physically alone, or two or more units are integrated into one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.

[0114] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory 505 and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present application. The aforementioned memory 505 includes: various media such as USB flash drives, mobile hard disks, magnetic disks, or optical discs that can store program codes.

[0115] The above are only exemplary embodiments of the present disclosure and should not be used to limit the scope of the present disclosure. That is, any equivalent changes and modifications made in accordance with the teachings of the present disclosure still fall within the scope covered by the present disclosure. Those skilled in the art will readily think of other implementation manners of the present disclosure after considering the disclosure of the specification. This application aims to cover any variations, uses, or adaptive changes of the present disclosure, and these variations, uses, or adaptive changes follow the general principles of the present disclosure and include common general knowledge or conventional technical means in the technical field not described in the present disclosure. The specification and the embodiments are only regarded as exemplary, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A personalized intelligent education textbook generation method based on a large model, characterized in that Applied in a server, the method comprises: Obtain knowledge point sets, learners’ learning data, learning goals, and eye tracking data; Determine a cognitive style vector of the learner according to the learning data and the eye tracking data, wherein the cognitive style vector is used to represent the information processing preference characteristics of the learner; Constructing a learning path diagram for the learner for the knowledge point set according to the cognitive style vector and the learning goal, wherein the learning path diagram includes learning nodes and jump relationships between a plurality of the learning nodes; Determining cognitive matching parameters of the knowledge point based on the cognitive style vector and the knowledge point attribute vector of each knowledge point in the knowledge point set, and determining a presentation form of the knowledge point according to the cognitive matching parameters; Based on the learning path map and the presentation form, personalized smart education teaching materials are generated through a preset large model.

2. The method according to claim 1, wherein The determining the learner's cognitive style vector according to the learning data and the eye tracking data specifically includes: extracting behavioral characteristic data from the learning data; Based on the page dwell time, the frequency of repeated visits and the interactive operation sequence in the behavior feature data, identifying the content interaction feature of the learner when processing the first knowledge point, the first knowledge point being any type of the knowledge point; determining cognitive processing characteristics during the learning process based on the saccade rate, gaze migration pattern, and pupil response characteristics in the eye tracking data; The learner's cognitive style vector is determined according to the content interaction feature and the cognitive processing feature.

3. The method according to claim 2, wherein The determining the learner's cognitive style vector according to the content interaction feature and the cognitive processing feature specifically includes: Determining the correspondence between the content interaction feature and the cognitive processing feature when the learner learns different types of the knowledge points through a preset behavior-cognition association model, and obtaining the learner's information processing feature set; The cognitive style vector is determined according to the information processing feature set.

4. The method according to claim 3, wherein Determining the cognitive style vector according to the information processing feature set specifically includes: Analyze the information processing feature set to obtain the learner's information acquisition intensity score, information processing depth score, knowledge construction mode index score, and skill transfer efficiency score; The cognitive style vector is obtained by calculating dimension feature values according to cognitive feature parameters, wherein the cognitive feature parameters include the information acquisition intensity score, the information processing depth score, the knowledge construction mode index score and the skill transfer efficiency score.

5. The method according to claim 1, characterized in that, The step of constructing a learning path map for the learner for the knowledge point set according to the cognitive style vector and the learning goal specifically includes: According to the learning objective, selecting a second knowledge point that meets the preset requirements from the knowledge point set, and using the second knowledge point as the learning node; Based on a preset knowledge graph, determining a logical dependency relationship between the second knowledge points; The jump relationship between the learning nodes is determined based on the cognitive style vector and the logical dependency relationship to obtain the learning path graph.

6. The method according to claim 5, characterized in that, Determining the jump relationship between the learning nodes based on the cognitive style vector and the logical dependency relationship to obtain the learning path graph specifically includes: Calculating the cognitive load value of each learning node according to the dimensional eigenvalue of the cognitive style vector; Constructing a knowledge association strength matrix between the learning nodes; Combining the cognitive load value and the knowledge association strength matrix to calculate the jump probability between the learning nodes; Constructing a learning path graph according to the jump probability, where the learning path graph includes a necessary learning path and an alternative learning path.

7. The method according to claim 1, wherein Determining the cognitive matching parameter of the knowledge point based on the cognitive style vector and the knowledge point attribute vector of each knowledge point in the knowledge point set, and determining the presentation form of the knowledge point according to the cognitive matching parameter, specifically including: Performing feature matching on the cognitive style vector and the knowledge point attribute vector based on a preset feature matching rule to obtain a feature matching matrix; Calculating the cognitive matching parameter according to the feature matching matrix; Generating a presentation form label based on the cognitive matching parameter, and determining the presentation form of the knowledge point according to the presentation form label.

8. A personalized intelligent education textbook generation system based on a large model, characterized in that, Including: An acquisition module, configured to acquire a knowledge point set, learning data of a learner, a learning target, and eye movement tracking data; A first determination module, configured to determine the cognitive style vector of the learner according to the learning data and the eye movement tracking data, where the cognitive style vector is used to represent the information processing preference characteristics of the learner; A construction module, configured to construct a learning path graph of the learner for the knowledge point set according to the cognitive style vector and the learning target, where the learning path graph includes learning nodes and jump relationships between multiple learning nodes; A second determination module, configured to determine the cognitive matching parameter of the knowledge point based on the cognitive style vector and the knowledge point attribute vector of each knowledge point in the knowledge point set, and determine the presentation form of the knowledge point according to the cognitive matching parameter; A generation module, configured to generate a personalized intelligent education textbook based on the learning path graph and the presentation form through a preset large model.

9. A personalized intelligent education textbook generation device based on a large model, characterized in that, Including: One or more processors and a memory; The memory is coupled to the one or more processors, and the memory is used to store computer program code, where the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the personalized intelligent education textbook generation device based on the large model to execute the method according to any one of claims 1-7.

10. A computer-readable storage medium, comprising instructions, characterized in that, When the instruction runs on the personalized intelligent education textbook generation device based on the large model, it enables the personalized intelligent education textbook generation device based on the large model to execute the method according to any one of claims 1-7.

Images