A guidance method and system for intelligent educational devices based on learning algorithms

Through in-depth analysis and personalized processing of user learning data, intelligent education equipment builds personalized knowledge learning paths, solving the problem that existing equipment cannot meet individual differentiated needs, and achieving efficient learning path customization and learning quality improvement.

CN118627611BActive Publication Date: 2024-11-22CLP DIGITAL TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411116916.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-15
Publication Date
2024-11-22
Estimated Expiration
2044-08-15

AI Technical Summary

Technical Problem

Existing intelligent educational equipment is difficult to analyze students' learning process in depth, and cannot meet individual differentiated learning needs, resulting in inefficient learning.

Method used

By obtaining user learning course outline data, structuring the knowledge ontology, generating personalized knowledge learning trees, extracting historical learning knowledge points, matching review questions, monitoring question answering behavior, inferring cognitive status, calculating knowledge point weights, and building metacognition Portrait, analyze knowledge vulnerabilities, and provide repair strategies and personalized learning tasks recommendations.

Benefits of technology

It has realized a customized learning path for each user, improve learning efficiency, enhance knowledge mastery, meet individual differentiated needs, and improve learning quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118627611B_ABST
    Figure CN118627611B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of artificial intelligence technology, and in particular to a guidance method and system for intelligent educational equipment based on a learning algorithm. The method comprises the following steps: reconstructing a personalized knowledge tree according to the user's learning course outline data to generate a personalized knowledge learning tree; performing real-time cognitive inference of the user's answering according to the personalized knowledge learning tree, and performing knowledge mastery score processing to generate knowledge point mastery level data; performing knowledge gap analysis according to the knowledge point mastery level data, and performing knowledge tracking processing to obtain knowledge gap repair strategy data; performing intelligent intervention guidance strategy processing according to the knowledge gap repair strategy data to obtain intelligent guidance effect evaluation data. The present invention realizes adaptive personalized learning through accurate perception of the user's real-time learning status.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a guidance method and system for intelligent educational equipment based on a learning algorithm. Background Art

[0002] With the rapid development of information technology and artificial intelligence, the application of intelligent educational devices in the field of education is becoming more and more extensive, providing students with a new learning experience and support. These devices usually consist of two parts: hardware and software. In terms of hardware, they include portable devices such as tablets and smart phones, which are not only easy to carry, but also provide a variety of learning methods and tools. The software system is its core, with advanced artificial intelligence functions such as intelligent interaction and adaptive learning guidance. Through these functions, the software can adjust the teaching content and methods according to the students' learning progress and needs, and effectively support the teaching and learning activities of teachers and students. However, many intelligent educational equipment products on the market (such as educational apps, etc.) usually evaluate students' mastery through fixed question banks, chapter exercises and knowledge point assessments, lack in-depth analysis of students' learning process, and it is difficult to meet diverse learning needs and differentiated requirements of individual learners, resulting in low learning efficiency. Summary of the invention

[0003] Based on this, the present invention provides a guidance method and system for an intelligent educational device based on a learning algorithm to solve at least one of the above technical problems.

[0004] To achieve the above purpose, a guidance method of an intelligent educational device based on a learning algorithm comprises the following steps:

[0005] Step S1: obtaining user learning course outline data; performing knowledge ontology structuring processing according to the user learning course outline data to generate knowledge ontology structure data; performing personalized knowledge tree reconstruction according to the knowledge ontology structure data to generate personalized knowledge learning tree;

[0006] Step S2: extract historical learning knowledge points according to the personalized knowledge learning tree to obtain historical learning knowledge point data; match knowledge point review questions according to the historical learning knowledge point data to generate recommended review question data; use the historical learning knowledge point data to perform knowledge point mapping processing on the recommended review question data to generate knowledge point-question mapping table data; perform interactive learning operations according to the recommended review question data to obtain answering behavior trajectory data;

[0007] Step S3: Based on the answering behavior trajectory data, the user's real-time answering status is monitored, and the answering cognitive inference is performed to generate cognitive status category data; the cognitive status category data is processed with knowledge point weights using the knowledge point-question mapping table data to generate knowledge point cognitive weight data; the knowledge mastery score is processed according to the knowledge point cognitive weight data to generate knowledge point mastery level data;

[0008] Step S4: construct a user metacognitive portrait based on cognitive state category data and personalized knowledge learning tree; perform knowledge gap analysis based on knowledge point mastery level data to generate a knowledge mastery gap matrix; perform knowledge tracking processing based on the knowledge mastery gap matrix, and generate knowledge gap repair strategies to obtain knowledge gap repair strategy data;

[0009] Step S5: The user performs knowledge gap repair processing through the knowledge gap repair strategy data to generate user knowledge repair feedback data; personalized learning task recommendations are performed through the user knowledge repair feedback data and the user metacognitive portrait to obtain adaptive learning task recommendation data; intelligent intervention guidance strategy processing is performed based on the adaptive learning task recommendation data to obtain intelligent guidance effect evaluation data.

[0010] The present invention can structure the course knowledge system and form unified knowledge ontology structure data by acquiring the course syllabus data of the user and performing knowledge ontology structured processing. Based on the knowledge ontology structure data, a personalized knowledge learning tree can be reconstructed, and the knowledge learning path can be tailored for each user to meet the personalized learning needs. The personalized knowledge learning tree can guide the user to learn relevant knowledge points in an orderly and systematic manner and improve learning efficiency. The knowledge points that the user has learned are extracted from the personalized knowledge learning tree to construct historical learning knowledge point data. Based on the historical learning knowledge point data, the review questions can be accurately matched, and the recommended review question data can be generated to help the user consolidate the knowledge learned. Through the knowledge point-question mapping table data, the user's mastery of the knowledge points corresponding to each question can be tracked. The answering behavior trajectory data generated during the user's answering process can record the user's learning process. Based on the user's answering behavior trajectory data, the user's answering status can be monitored in real time, and the user's cognitive status category can be inferred, such as mastery, doubtful understanding, and no mastery at all. Combined with the knowledge point-question mapping table data, the cognitive weight corresponding to each knowledge point can be calculated, and then the mastery level of the knowledge point can be obtained. The mastery level data can quantify the user's mastery of each knowledge point. Based on the user's cognitive state category data and personalized knowledge learning tree, a metacognitive portrait of the user can be constructed to fully reflect the user's learning state and cognitive ability characteristics. By analyzing the knowledge point mastery level data, it is possible to find out which knowledge points the user has loopholes in, and generate a knowledge mastery loophole matrix. Based on the loophole matrix, the user's knowledge loopholes can be tracked, and corresponding repair strategies can be generated to guide the user to make up for the knowledge loopholes in a targeted manner. The user performs targeted knowledge repair learning according to the knowledge loophole repair strategy, and generates repair feedback data, which can evaluate the repair effect. Combined with the user's metacognitive portrait, personalized follow-up learning tasks can be recommended to meet the user's actual learning needs. Through intelligent intervention guidance strategy processing, the user's learning process can be intelligently guided and intervened, the learning quality can be improved, and adaptive personalized learning can be achieved. Therefore, the guidance method of a smart educational device based on a learning algorithm of the present invention starts from the course outline, uses natural language processing, knowledge ontology construction and other technologies, constructs a knowledge ontology structure and a knowledge component tree, and then combines the user's learning data for personalized adjustment, and tailors a suitable knowledge learning tree for each user. Utilize the user's learning log data to analyze the user's learning behavior characteristics and learning preferences to make personalized adjustments to the knowledge tree, analyze the user's answers and learning cognitive status in real time, and comprehensively control the user's knowledge mastery. Tailor the most appropriate learning path for each learner to achieve accurate knowledge transfer and effective learning guidance.

[0011] Preferably, step S1 comprises the following steps:

[0012] Step S11: Acquire user learning course outline data and user learning log data;

[0013] Step S12: Performing semantic analysis on the user's course syllabus data to generate course knowledge semantic network data;

[0014] Step S13: Perform knowledge ontology structuring processing according to the course knowledge semantic network data to generate knowledge ontology structure data;

[0015] Step S14: identifying knowledge point association relationships of knowledge ontology structure data through course knowledge semantic network data to generate knowledge point association data;

[0016] Step S15: Analyze the knowledge point concept hierarchy of the knowledge ontology structure data through the knowledge point association data, and construct a knowledge tree to generate knowledge component tree data;

[0017] Step S16: Use the user learning log data to reconstruct the knowledge component tree data into a personalized knowledge tree to generate a personalized knowledge learning tree.

[0018] The present invention obtains user learning course outline data and user learning log data, provides initial information of user learning content for intelligent education equipment, and determines the scope and goal of learning. The learning content in the course outline is semantically analyzed to identify key concepts, relationships and structures in the learning content. The association relationship between knowledge is established, and the knowledge is structured according to the course knowledge semantic network data, so that the device can understand the knowledge system more deeply, and plan personalized learning paths according to the logical relationship between knowledge. According to the course knowledge semantic network data, the association relationship between knowledge points in the knowledge ontology structure is analyzed, such as: inclusion relationship, parallel relationship, causal relationship, etc. The knowledge points are conceptually analyzed, and the knowledge points are organized according to the hierarchical relationship to form a structured knowledge tree. According to the user learning log data, the user's learning habits, learning style and learning progress are analyzed, and the knowledge component tree is personalized reconstructed, so that the device can provide a customized learning path for each user, meet the user's personalized needs, and improve learning efficiency and effect.

[0019] Preferably, step S16 comprises the following steps:

[0020] Step S161: labeling the knowledge component importance weights of the knowledge component tree data using the preset knowledge importance weights to generate an important knowledge component tree;

[0021] Step S162: extracting user learning behavior features from the user learning log data to generate user learning behavior feature data, wherein the user learning behavior feature data includes learning content data, learning sequence data, and learning duration data;

[0022] Step S163: Plan important learning nodes for the important knowledge component tree using the learning sequence data to generate important learning node data;

[0023] Step S164: performing knowledge content mapping processing on the important knowledge component tree through the learning content data to obtain user-knowledge component interaction data;

[0024] Step S165: Performing user learning preference analysis based on the learning time data through the user-knowledge component interaction data to generate user learning preference data;

[0025] Step S166: Calculate the knowledge point relevance of the important knowledge component tree according to the user learning preference data to generate knowledge point relevance matrix data;

[0026] Step S167: Perform knowledge point clustering analysis based on the knowledge point correlation matrix data, and use the important learning node data to reconstruct the important knowledge component tree into a personalized learning tree to generate a personalized knowledge learning tree.

[0027] The invention sets importance weights for different knowledge points based on teaching experience and expert opinions, reflecting the importance of different knowledge points in learning. The important knowledge component tree can highlight the key and difficult knowledge in the course, and the intelligent education device can arrange the teaching focus according to these key knowledge. The learning behavior characteristics of the user are extracted from the user learning log data. The device is enabled to understand the user's learning habits, learning style and learning progress. According to the user's learning sequence data, the sequential relationship of the user's learning content is analyzed, and the key nodes in the learning content are identified, so that the device can guide the user to learn more effectively. The knowledge points learned by the user are mapped with the knowledge components in the important knowledge component tree, and the learning interaction between the user and each knowledge component is recorded. According to the user-knowledge component interaction data and the learning time data, the user's learning preference for different knowledge points is analyzed, and the user's learning interest and learning needs can be understood more accurately. According to the user learning preference data, the user's learning relationship for different knowledge points is analyzed, and the logical relationship between knowledge can be understood more accurately. According to the knowledge point correlation matrix data, the personalized learning tree is reconstructed, and the importance, correlation and learning preference of the knowledge point are comprehensively considered, and the optimal learning path is tailored for each user.

[0028] Preferably, step S2 comprises the following steps:

[0029] Step S21: Mark the user's historical learning knowledge nodes through the personalized knowledge learning tree to obtain the user's historical learning node data;

[0030] Step S22: extracting learning knowledge points according to the user's historical learning node data to obtain historical learning knowledge point data;

[0031] Step S23: matching knowledge point review questions according to historical learning knowledge point data to generate recommended review question data;

[0032] Step S24: using the historical learning knowledge point data to perform knowledge point mapping processing on the recommended review question data to generate knowledge point-question mapping table data;

[0033] Step S25: using the intelligent educational device to initialize the user's learning progress, and performing visual guidance processing according to the recommended review question data to obtain visual review question data;

[0034] Step S26: The user performs interactive learning operations on the visualized review question data and captures the behavior trajectory to generate answering behavior trajectory data.

[0035] The present invention marks the knowledge nodes that the user has learned according to the learning trajectory of the user in the personalized knowledge learning tree, and can understand the user's learning progress and learning content. According to the user's historical learning node data, the knowledge points that the user has learned are extracted, and the user's mastery of different knowledge points can be understood. According to the historical learning knowledge point data, relevant review questions are recommended to the user, and targeted review exercises can be provided. The recommended review question data is matched with the historical learning knowledge point data, and a mapping relationship between the knowledge point and the corresponding exercise questions is established, which can accurately evaluate the user's mastery of different knowledge points, and targeted knowledge point explanations and exercise recommendations are performed. According to the user's historical learning node data, personalized learning content recommendations and learning path planning are performed, which can more effectively guide the user to learn, and the learning content is adjusted according to the learning progress. The user performs interactive learning operations on the visualized review question data through the intelligent education device, thereby analyzing the user's learning characteristics and weak links.

[0036] Preferably, step S23 includes the following steps:

[0037] Step S231: performing knowledge point theme analysis on the historical learning knowledge point data to generate knowledge point theme data;

[0038] Step S232: performing knowledge topic semantic retrieval on a preset exercise question bank through knowledge point topic vector data to generate topic retrieval exercise question data;

[0039] Step S233: using a preset word embedding model to perform topic embedding processing on the topic retrieval exercise data, and using a latent Dirichlet allocation algorithm to perform text sequence topic analysis to obtain exercise topic vector data, wherein the topic embedding processing is performed by converting the exercise stem and the option text into a word vector sequence;

[0040] Step S234: performing bidirectional soft matching of the subject matter vector data of the exercises through the subject matter data of the knowledge points to generate a subject matter matching probability matrix;

[0041] Step S235: performing similarity thresholding processing on the subject matching probability matrix. When the subject similarity of the exercises in the subject matching probability matrix is ​​greater than the preset subject matching similarity threshold, the exercises are marked as initial matching exercises; when the subject similarity of the exercises in the subject matching probability matrix is ​​less than or equal to the preset subject matching similarity threshold, the exercises are marked as unmatched exercises and are eliminated;

[0042] Step S236: Perform context knowledge matching scoring on the initial matching exercises using the knowledge point theme data to generate context knowledge point scoring data;

[0043] Step S237: weighted sorting of the initial matching exercises is performed using the contextual knowledge point scoring data to obtain recommended review question data.

[0044] The present invention performs thematic analysis on historical learning knowledge point data, classifies knowledge points according to themes, and can more accurately understand the logical relationship between knowledge points. According to the knowledge point theme data, the preset exercise question bank is searched for theme semantics to find exercises related to the knowledge point theme. Using the preset word embedding model, the exercise question stem and the option text are converted into a word vector sequence, which can more accurately identify the theme of the exercise question, and match the exercise questions according to the theme. The word vector sequence of the exercise question is analyzed to more deeply understand the theme of the exercise question. The knowledge point theme data and the exercise question theme vector data are soft-matched in both directions to more accurately identify the relationship between the knowledge point theme and the exercise question theme, and the exercise question is recommended according to the similarity. According to the preset theme matching similarity threshold, the theme matching probability matrix is ​​screened, and the exercises related to the knowledge point theme can be more effectively screened out, thereby improving the review efficiency and learning effect. According to the knowledge point theme data, the context knowledge matching score is performed, and the knowledge points of the exercise question can be more accurately identified, and the exercise questions are sorted according to the relevance. The initial matching exercises are weighted and sorted, and the exercises with higher relevance to the knowledge points are placed at the front to form recommended review question data.

[0045] Preferably, step S3 comprises the following steps:

[0046] Step S31: extracting the behavior time sequence according to the answering behavior trajectory data to generate the answering behavior time sequence data; calculating the answering speed according to the answering behavior time sequence data to generate the answering speed data;

[0047] Step S32: mining the answering behavior pattern of the answering behavior time series data through the answering speed data to obtain the answering behavior pattern data;

[0048] Step S33: extracting answering effects according to the answering behavior pattern data to generate answering effect data;

[0049] Step S34: Based on the answering behavior trajectory data, the intelligent education device is used to monitor the user's answering status in real time, and answering cognition inference is performed to generate answering cognition inference data;

[0050] Step S35: performing cognitive state clustering processing on the cognitive inference data of the answer to generate cognitive state category data;

[0051] Step S36: performing knowledge point association analysis on the cognitive state category data through the knowledge point-topic mapping table data, and performing knowledge point weight calculation to generate knowledge point cognitive weight data;

[0052] Step S37: using the answering effect data to perform answering result weighting processing on the knowledge point cognitive weight data to generate knowledge point cognitive weighted data;

[0053] Step S38: Calculate the knowledge mastery score of the knowledge point cognitive weighted data using the cognitive state category data, and perform mastery level classification processing to generate knowledge point mastery level data.

[0054] The present invention extracts behavior time series according to answering behavior trajectory data, records the timestamp of each behavior in the process of answering questions by the user, and reflects the speed characteristics of the user's answering. By mining the answering behavior pattern of the answering behavior time series data through the answering speed data, the answering behavior pattern data can be obtained, which reveals the behavior pattern and regularity of the user in the process of answering questions. The answering effect is extracted according to the answering behavior pattern data, which reflects the user's answering accuracy and error situation, evaluates the user's knowledge mastery, and adjusts the teaching difficulty and progress in a targeted manner to improve the adaptability and effectiveness of teaching. The user's real-time answering status is monitored, and the answering cognitive inference is performed, which can reflect the changes in the user's cognitive state in the process of answering questions, such as concentration, thinking in trouble, etc., and timely discover the user's learning difficulties. The answering cognitive inference data is clustered by cognitive state, and the user's cognitive state is divided into different categories, such as mastery, doubtful understanding, and no mastery at all, etc. Based on these category data, the user's knowledge mastery can be more accurately evaluated. The knowledge point association analysis is performed on the cognitive state category data, which can quantify the user's mastery of each knowledge point and discover the user's knowledge loopholes. The answering results are weighted by using the answering effect data to process the knowledge point cognitive weight data, taking into account the user's cognitive state, and can more comprehensively reflect the user's mastery of each knowledge point. The knowledge point mastery level data can clearly show the user's mastery level of each knowledge point. Knowledge points with lower mastery levels need to be reviewed in detail; knowledge points with higher mastery levels should be reviewed less appropriately to improve teaching efficiency.

[0055] Preferably, step S34 includes the following steps:

[0056] Step S341: Based on the answering behavior trajectory data, the user's real-time answering status is collected using the intelligent education device to generate real-time answering status data, wherein the real-time answering status data includes answering facial expression data and eye movement trajectory data;

[0057] Step S342: performing facial expression feature analysis based on the answering facial expression data, and performing emotion label matching to generate facial expression label data;

[0058] Step S343: capturing the gaze point of the eyeball movement trajectory data to generate answer gaze point data; performing answer time series aggregation analysis on the answer gaze point data using the eyeball movement trajectory data to generate answer sight focus data;

[0059] Step S344: Calculate the duration of the gaze focus according to the gaze focus data of the answer, and generate gaze focus duration data; evaluate the concentration level according to the gaze focus duration data, and perform question relevance matching to generate answer concentration level evaluation data;

[0060] Step S345: using the answering concentration evaluation data and the facial expression label data to perform transfer learning processing on the preset recurrent neural network model, and performing cognitive state inference model training to obtain a cognitive state inference model;

[0061] Step S346: Use the cognitive state inference model to perform question cognitive state inference processing on the question answering concentration evaluation data and the facial expression label data to generate question answering cognitive inference data.

[0062] The present invention collects the user's answering status data in real time, including answering facial expression data and eye movement trajectory data, which can comprehensively reflect the user's answering status. Facial expression feature analysis is performed based on the answering facial expression data, and the user's facial expression categories during the answering process are marked, such as doubt, concentration, confusion, etc. The eye movement trajectory data is captured for gaze points to reflect the distribution and changes of the user's line of sight focus during the answering process, and to understand the user's focus area. The length of time the user's line of sight stays in different areas is calculated based on the answering line of sight focus data, so as to understand the user's concentration level and more accurately understand the user's learning status. The cognitive state inference model is used to perform question cognitive state inference processing on the answering concentration evaluation data and the facial expression label data, which can accurately reflect the user's cognitive state on each question, such as concentration, confusion, doubts about understanding, etc.

[0063] Preferably, step S4 comprises the following steps:

[0064] Step S41: labeling the knowledge point mastery level of the personalized knowledge learning tree using the knowledge point mastery level data to obtain a labeled knowledge learning tree;

[0065] Step S42: constructing a user metacognitive portrait based on cognitive state category data and annotated knowledge learning tree;

[0066] Step S43: using the annotated knowledge learning tree to perform knowledge gap analysis on the knowledge point mastery level data to generate a knowledge mastery gap matrix;

[0067] Step S44: training a Bayesian network model using a preset massive educational resource library and annotated knowledge learning tree to construct a knowledge tracking model;

[0068] Step S45: The knowledge mastery vulnerability matrix is ​​transmitted to the knowledge tracking model to generate a knowledge vulnerability repair strategy, and obtain knowledge vulnerability repair strategy data.

[0069] The present invention uses the knowledge point mastery level data to mark the knowledge point mastery level of the personalized knowledge learning tree, marks the user's mastery level of each knowledge point, and grasps the user's knowledge structure and weaknesses. According to the cognitive state category data and the marked knowledge learning tree, the user's learning habits, learning style and learning strategy are analyzed to build a user metacognitive portrait. The knowledge point mastery level data is used to analyze the knowledge gaps, which reflects the user's mastery level on different knowledge points, quickly discovers the user's knowledge gaps, and focuses on and strengthens the weak knowledge points to improve the accuracy and effectiveness of teaching. According to the preset massive educational resource library and the marked knowledge learning tree, a knowledge tracking model is constructed, which tracks resources according to the relevant knowledge weaknesses and arranges repair strategies. The knowledge mastery gap matrix is ​​transmitted to the knowledge tracking model, and according to the user's knowledge gap situation, for example: recommending relevant learning resources, providing knowledge point explanations, arranging exercises, etc. It can more effectively help users make up for knowledge gaps in learning and improve learning efficiency and learning effects.

[0070] Preferably, step S5 comprises the following steps:

[0071] Step S51: The user performs knowledge vulnerability repair processing according to the knowledge vulnerability repair strategy data to generate user knowledge repair feedback data;

[0072] Step S52: performing knowledge tree updating processing on the user metacognitive portrait through the user knowledge repair feedback data to generate an updated user metacognitive portrait;

[0073] Step S53: obtaining the user's current learning target data; performing personalized learning task recommendations for the user's current learning target data by updating the user's metacognitive portrait, and performing adaptive learning difficulty adjustment to obtain adaptive learning task recommendation data;

[0074] Step S54: performing real-time learning state monitoring based on the adaptive learning task recommendation data, and performing learning cognitive inference to generate real-time metacognitive state label data;

[0075] Step S55: performing intelligent intervention guidance strategy processing according to the real-time metacognitive state tag data to generate intelligent intervention guidance strategy data, wherein the intelligent intervention guidance strategy data includes learning state prompts, knowledge point explanations, and exercise question insertions;

[0076] Step S56: Evaluate learning efficiency based on the intelligent intervention guidance strategy data to obtain intelligent guidance effect evaluation data.

[0077] According to the knowledge gap repair strategy provided by the device, users of the present invention study and practice, such as watching relevant videos, reading relevant teaching materials, and taking exercises, etc., to help users effectively make up for knowledge gaps in learning. According to the user's knowledge repair feedback data, the knowledge tree in the user's metacognitive portrait is updated, such as adjusting the knowledge point mastery level, updating the learning path, etc. It can more accurately reflect the user's learning situation. The user's learning goals are obtained, such as learning a certain knowledge point, completing a certain task, etc. According to the updated user metacognitive portrait and the user's learning goals, personalized learning tasks are recommended to the user, such as recommending relevant learning resources, arranging exercises, and providing guidance on learning methods. It can provide targeted learning content to improve learning efficiency and learning interest. According to the adaptive learning task recommendation data, the user's learning status is monitored in real time, the user's learning problems can be discovered in time, and the changes in the user's cognitive state during the learning process can be reflected in real time. According to the real-time metacognitive status tag data, intelligent intervention guidance is provided to users, such as providing learning status prompts, explaining knowledge points, inserting exercises, etc. These strategies can tailor personalized learning guidance for users, such as prompting distraction, explaining difficult knowledge, inserting related exercises, etc. These strategies can be used to intervene and assist the user's learning process in real time to improve learning efficiency and quality. Based on the data of intelligent intervention guidance strategies, learning efficiency is evaluated to reflect the actual role of intelligent guidance strategies in improving learning efficiency. Based on this data, the guidance strategy is continuously optimized and improved, the guidance effect is continuously improved, a closed-loop feedback is formed, and continuous iterative optimization of the strategy is achieved.

[0078] The present invention also provides a guidance system for an intelligent educational device based on a learning algorithm, which executes the guidance method for an intelligent educational device based on a learning algorithm as described above. The guidance system for an intelligent educational device based on a learning algorithm comprises:

[0079] The knowledge learning tree construction module obtains the user's learning course outline data; performs knowledge ontology structure processing based on the user's learning course outline data to generate knowledge ontology structure data; performs personalized knowledge tree reconstruction based on the knowledge ontology structure data to generate a personalized knowledge learning tree;

[0080] The knowledge point review recommendation module is used to extract historical learning knowledge points according to the personalized knowledge learning tree to obtain historical learning knowledge point data; match knowledge point review questions according to the historical learning knowledge point data to generate recommended review question data; use the historical learning knowledge point data to perform knowledge point mapping processing on the recommended review question data to generate knowledge point-question mapping table data; perform interactive learning operations according to the recommended review question data to obtain answering behavior trajectory data;

[0081] The cognitive state assessment module is used to monitor the user's real-time answering state based on the answering behavior trajectory data, and to make cognitive inferences about answering questions to generate cognitive state category data; to perform knowledge point weight processing on the cognitive state category data using the knowledge point-question mapping table data to generate knowledge point cognitive weight data; to perform knowledge mastery score processing based on the knowledge point cognitive weight data to generate knowledge point mastery level data;

[0082] The knowledge leak tracking module is used to build a user metacognitive portrait based on cognitive state category data and personalized knowledge learning tree; perform knowledge leak analysis based on knowledge point mastery level data to generate a knowledge mastery leak matrix; perform knowledge tracking processing based on the knowledge mastery leak matrix, and generate knowledge leak repair strategies to obtain knowledge leak repair strategy data;

[0083] The intelligent learning guidance module is used for users to perform knowledge loophole repair processing through knowledge loophole repair strategy data to generate user knowledge repair feedback data; to recommend personalized learning tasks through user knowledge repair feedback data and user metacognitive portraits to obtain adaptive learning task recommendation data; to perform intelligent intervention guidance strategy processing based on adaptive learning task recommendation data to obtain intelligent guidance effect evaluation data. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Figure 1 A schematic diagram of the steps of a guidance method and system for an intelligent educational device based on a learning algorithm according to the present invention;

[0085] Figure 2 for Figure 1 Detailed implementation steps of step S4 in FIG.

[0086] Figure 3 for Figure 1 Detailed implementation steps of step S5;

[0087] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0088] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0089] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0090] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0091] To achieve this, please refer to Figures 1 to 3 The present invention provides a guiding method of an intelligent educational device based on a learning algorithm, comprising the following steps:

[0092] Step S1: obtaining user learning course outline data; performing knowledge ontology structuring processing according to the user learning course outline data to generate knowledge ontology structure data; performing personalized knowledge tree reconstruction according to the knowledge ontology structure data to generate personalized knowledge learning tree;

[0093] Step S2: extract historical learning knowledge points according to the personalized knowledge learning tree to obtain historical learning knowledge point data; match knowledge point review questions according to the historical learning knowledge point data to generate recommended review question data; use the historical learning knowledge point data to perform knowledge point mapping processing on the recommended review question data to generate knowledge point-question mapping table data; perform interactive learning operations according to the recommended review question data to obtain answering behavior trajectory data;

[0094] Step S3: Based on the answering behavior trajectory data, the user's real-time answering status is monitored, and the answering cognitive inference is performed to generate cognitive status category data; the cognitive status category data is processed with knowledge point weights using the knowledge point-question mapping table data to generate knowledge point cognitive weight data; the knowledge mastery score is processed according to the knowledge point cognitive weight data to generate knowledge point mastery level data;

[0095] Step S4: construct a user metacognitive portrait based on cognitive state category data and personalized knowledge learning tree; perform knowledge gap analysis based on knowledge point mastery level data to generate a knowledge mastery gap matrix; perform knowledge tracking processing based on the knowledge mastery gap matrix, and generate knowledge gap repair strategies to obtain knowledge gap repair strategy data;

[0096] Step S5: The user performs knowledge gap repair processing through the knowledge gap repair strategy data to generate user knowledge repair feedback data; personalized learning task recommendations are performed through the user knowledge repair feedback data and the user metacognitive portrait to obtain adaptive learning task recommendation data; intelligent intervention guidance strategy processing is performed based on the adaptive learning task recommendation data to obtain intelligent guidance effect evaluation data.

[0097] In the embodiment of the present invention, reference Figure 1 The above is a schematic flow chart of the steps of a method for guiding an intelligent educational device based on a learning algorithm of the present invention. In this embodiment, the method for guiding an intelligent educational device based on a learning algorithm includes the following steps:

[0098] Step S1: obtaining user learning course outline data; performing knowledge ontology structuring processing according to the user learning course outline data to generate knowledge ontology structure data; performing personalized knowledge tree reconstruction according to the knowledge ontology structure data to generate personalized knowledge learning tree;

[0099] In an embodiment of the present invention, course syllabus data is obtained from a course management system, including information such as course name, chapter structure, and knowledge point content. At the same time, the user's learning log data is obtained from a learning record system, including information such as the courses, chapters, knowledge points, learning time, duration, and learning effect that have been learned. The course syllabus data is semantically analyzed using natural language processing technology to extract the relationship between knowledge points and knowledge points, and a course knowledge semantic network is constructed. Based on the course knowledge semantic network, the knowledge is structured using ontology construction technology to form knowledge ontology structure data. According to the knowledge ontology structure data, a knowledge tree is constructed to bring together closely related knowledge points to form knowledge components and generate knowledge component tree data. The user's learning log data is analyzed to identify the user's weak links. According to the user's weak links, the knowledge component tree data is reconstructed and adjusted, and the weak link knowledge components are moved upward so that they can obtain more learning resources and attention, and a personalized knowledge learning tree is generated.

[0100] Step S2: extract historical learning knowledge points according to the personalized knowledge learning tree to obtain historical learning knowledge point data; match knowledge point review questions according to the historical learning knowledge point data to generate recommended review question data; use the historical learning knowledge point data to perform knowledge point mapping processing on the recommended review question data to generate knowledge point-question mapping table data; perform interactive learning operations according to the recommended review question data to obtain answering behavior trajectory data;

[0101] In an embodiment of the present invention, the knowledge points that the user has learned are marked according to the personalized knowledge learning tree. The knowledge point content description is extracted from the marked nodes to form a historical learning knowledge point data set. Based on the historical learning knowledge point data set, the corresponding review questions are accurately matched from the question bank. For example, for "Newton's First Law", questions about the uniform motion of the test object are selected. The historical learning knowledge point data is mapped and associated with the recommended review question data to generate knowledge point-question mapping table data to clarify the knowledge points corresponding to each review question. Based on the recommended review question data, the review questions are presented on the device interface in the form of graphics, text, animation, etc. The user can view the content of the question and try to answer it. In the process of the user answering, the user's operation trajectory is captured to form the answering behavior trajectory data.

[0102] Step S3: Based on the answering behavior trajectory data, the user's real-time answering status is monitored, and the answering cognitive inference is performed to generate cognitive status category data; the cognitive status category data is processed with knowledge point weights using the knowledge point-question mapping table data to generate knowledge point cognitive weight data; the knowledge mastery score is processed according to the knowledge point cognitive weight data to generate knowledge point mastery level data;

[0103] In an embodiment of the present invention, a timestamp sequence is extracted from the user's answering behavior trajectory data, and the answering speed data is calculated to reflect the user's answering rhythm and efficiency. The answering behavior time series data is analyzed using a sequence pattern mining algorithm, and the typical behavior pattern of the user in the answering process is found in combination with the answering speed data, and the answering behavior pattern data is generated. The answering behavior pattern is associated with the final answering result, and the answering effect caused by different behavior patterns is analyzed. Combined with a pre-established cognitive model, the user's current cognitive state is inferred to generate answering cognitive inference data. The answering cognitive inference data is clustered and similar cognitive states are classified into one category. The user's cognitive state category data is linked to the corresponding knowledge point, and the weight score of the knowledge point is calculated according to the frequency of occurrence of the cognitive state of each knowledge point, and the knowledge point cognitive weight data is generated. The knowledge point cognitive weight data is combined with the user's answering effect data to adjust the weight value. According to the cognitive state category data, different scoring strategies are designed and applied to the knowledge point cognitive weight data, so as to generate a mastery score for each knowledge point. The knowledge point score is mapped to different mastery levels, and finally the knowledge point mastery level data is formed.

[0104] Step S4: construct a user metacognitive portrait based on cognitive state category data and personalized knowledge learning tree; perform knowledge gap analysis based on knowledge point mastery level data to generate a knowledge mastery gap matrix; perform knowledge tracking processing based on the knowledge mastery gap matrix, and generate knowledge gap repair strategies to obtain knowledge gap repair strategy data;

[0105] In an embodiment of the present invention, according to the cognitive state category data, the number of occurrences of each cognitive state is counted and mapped to the annotated knowledge learning tree. Based on cognitive state analysis, combined with the annotated knowledge learning tree, a metacognitive portrait of the user is drawn. Each node on the annotated knowledge learning tree is scanned to identify the knowledge points marked as "not mastered" or "basically mastered" as the user's knowledge gap. The position of these vulnerability knowledge points in the knowledge tree, the relationship with other knowledge points and other information are analyzed to evaluate the importance of each vulnerability knowledge point and generate a knowledge mastery gap matrix. The cause-and-effect relationship between knowledge points is extracted from the educational resource library to construct an initial Bayesian network structure. The parameters of the Bayesian network are trained and optimized using the data in the annotated knowledge learning tree. Each vulnerability knowledge point with a higher weight score in the vulnerability matrix is ​​traversed and input as an evidence node into the Bayesian network knowledge tracking model. Based on reasoning analysis, a series of repair strategies are generated for each vulnerability knowledge point, including pre-requisite knowledge points that need to be learned first, suitable teaching resources, exercise questions, etc., to form a knowledge gap repair strategy data set.

[0106] Step S5: The user performs knowledge gap repair processing through the knowledge gap repair strategy data to generate user knowledge repair feedback data; personalized learning task recommendations are performed through the user knowledge repair feedback data and the user metacognitive portrait to obtain adaptive learning task recommendation data; intelligent intervention guidance strategy processing is performed based on the adaptive learning task recommendation data to obtain intelligent guidance effect evaluation data.

[0107] In an embodiment of the present invention, the user learns according to the knowledge gap repair strategy data to generate user knowledge repair feedback data. The feedback data is analyzed to evaluate the user's mastery of the gap knowledge points and update the metacognitive portrait. According to the updated metacognitive portrait, the mastery of the knowledge points related to the user's learning goals is analyzed. Corresponding learning tasks are recommended, including learning resources, exercises, etc., to generate adaptive learning task recommendation data. The user starts learning according to the adaptive learning task recommendation data, and the intelligent education device monitors the user's learning behavior trajectory in real time and generates real-time metacognitive state label data. According to the label data, a personalized intelligent intervention guidance strategy is automatically generated, and the strategy execution effect is continuously monitored to generate intelligent guidance effect evaluation data.

[0108] Preferably, step S1 comprises the following steps:

[0109] Step S11: Acquire user learning course outline data and user learning log data;

[0110] Step S12: Performing semantic analysis on the user's course syllabus data to generate course knowledge semantic network data;

[0111] Step S13: Perform knowledge ontology structuring processing according to the course knowledge semantic network data to generate knowledge ontology structure data;

[0112] Step S14: identifying knowledge point association relationships of knowledge ontology structure data through course knowledge semantic network data to generate knowledge point association data;

[0113] Step S15: Analyze the knowledge point concept hierarchy of the knowledge ontology structure data through the knowledge point association data, and construct a knowledge tree to generate knowledge component tree data;

[0114] Step S16: Use the user learning log data to reconstruct the knowledge component tree data into a personalized knowledge tree to generate a personalized knowledge learning tree.

[0115] In the embodiment of the present invention, for example, the syllabus data of the corresponding course is obtained from the course management system, including information such as the course name, chapter structure, and knowledge point content. At the same time, the user's learning log data can also be obtained from the learning record system, including the courses, chapters, knowledge points that the user has learned, as well as information such as learning time, learning duration, and learning effect. The course syllabus data is semantically analyzed using natural language processing technology to extract information such as course knowledge points and the relationship between knowledge points. For example, from the sentence "Newton's second law is a law that describes the relationship between the resultant force on an object when it moves and the acceleration of the object", concepts such as "Newton's second law", "object", "resultant force", and "acceleration" are extracted, and the semantic relationship between them is identified to build a course knowledge semantic network. Based on the course knowledge semantic network, the ontology construction technology is used to structure the knowledge. For example, "Newton's second law" is taken as a class, and "object", "resultant force", and "acceleration" are taken as attributes of the class. By performing this structured representation on the knowledge, the hierarchical relationship and association relationship between the knowledge concepts can be clearly portrayed to form knowledge ontology structure data. For example, from semantic information such as "mechanics is a discipline that studies the motion of objects and their causes, including kinematics and dynamics", "kinematics studies the laws of motion of objects without considering the causes of motion", and "dynamics studies the causes of motion of objects, including Newton's laws of motion", we can identify the inclusion relationship between "mechanics" and "kinematics" and "dynamics", the parallel relationship between "kinematics" and "dynamics", and the subordinate relationship between "Newton's laws of motion" and "dynamics". For example, "mechanics" is taken as the root node, "kinematics" and "dynamics" are taken as the child nodes of "mechanics", and "Newton's laws of motion" are taken as the child node of "dynamics". By characterizing this hierarchical relationship of all knowledge point concepts, a knowledge tree is constructed. On the basis of the knowledge tree, knowledge points with close associations are gathered together to form knowledge components, thereby generating knowledge component tree data. For example, if it is found from the log data that users spend a long time learning "Newton's laws of motion" and have poor learning results, it can be judged that this knowledge point is the user's weak link. Based on this analysis, the knowledge component tree data is reconstructed and adjusted, and the knowledge component "Newton's Laws of Motion" is moved upward and placed closer to the root node so that it can obtain more learning resources and attention.

[0116] Preferably, step S16 comprises the following steps:

[0117] Step S161: labeling the knowledge component importance weights of the knowledge component tree data using the preset knowledge importance weights to generate an important knowledge component tree;

[0118] Step S162: extracting user learning behavior features from the user learning log data to generate user learning behavior feature data, wherein the user learning behavior feature data includes learning content data, learning sequence data, and learning duration data;

[0119] Step S163: Plan important learning nodes for the important knowledge component tree using the learning sequence data to generate important learning node data;

[0120] Step S164: performing knowledge content mapping processing on the important knowledge component tree through the learning content data to obtain user-knowledge component interaction data;

[0121] Step S165: Performing user learning preference analysis based on the learning time data through the user-knowledge component interaction data to generate user learning preference data;

[0122] Step S166: Calculate the knowledge point relevance of the important knowledge component tree according to the user learning preference data to generate knowledge point relevance matrix data;

[0123] Step S167: Perform knowledge point clustering analysis based on the knowledge point correlation matrix data, and use the important learning node data to reconstruct the important knowledge component tree into a personalized learning tree to generate a personalized knowledge learning tree.

[0124] In an embodiment of the present invention, each knowledge component in the knowledge component tree is scored for importance according to a set of pre-set knowledge importance evaluation criteria. For example, for the course "Mechanics", "Newton's Laws of Motion" as a core knowledge point can be given a higher weight score, while some auxiliary knowledge points are given lower scores. After marking the weights of all knowledge components in this way, an important knowledge component tree is generated. For example, from the log data, it can be learned that the knowledge points learned by the user are "force", "motion", "acceleration", etc. The order of learning is "force" first, then "motion" and then "acceleration", and the learning time of each knowledge point is 30 minutes, 20 minutes and 10 minutes respectively. Combined with the learning knowledge component tree, some important learning nodes are planned. For example, if the learning order of most users is "kinematics" first and then "dynamics", "kinematics" can be determined as an important entry node; "Newton's Laws of Motion" as a core knowledge point can also be planned as an important node. The knowledge content actually learned by the user is mapped to the important knowledge component tree. For example, if the user has learned the knowledge points of "force", "motion" and "acceleration", the corresponding knowledge component nodes are found on the important knowledge component tree, and the interaction data between the user and these nodes are recorded to form a user-knowledge component interaction data set. For example, if the user spends a long time on the knowledge component of "Newton's Second Law", it can be judged that the user is interested in this knowledge point or has certain learning difficulties. If the user spends a short time on the knowledge component of "Introduction to Mechanics", it means that the user is not very interested in this knowledge point. Based on the user's learning preference data, the correlation between different knowledge points is calculated. For example, if the user has a high learning preference for the two knowledge points of "Newton's Second Law" and "Law of Conservation of Momentum", the relevant knowledge points of these two knowledge points on the knowledge tree can be extracted. This can be achieved by calculating the correlation scores between all knowledge points and generating a knowledge point correlation matrix. Each element value in the matrix represents the degree of correlation between the corresponding two knowledge points. Clustering algorithms (such as K-means) are used to cluster knowledge points, and knowledge points with high relevance are grouped into the same category. The structure of the important knowledge component tree is adjusted and optimized with reference to important learning node data, making learning nodes more prominent and related knowledge points closer. Through this personalized reconstruction, a personalized knowledge learning tree that best suits each user is generated. The tree structure and knowledge point layout are optimized based on the user's actual learning situation and preferences, which is conducive to improving learning efficiency and effectiveness.

[0125] Preferably, step S2 comprises the following steps:

[0126] Step S21: Mark the user's historical learning knowledge nodes through the personalized knowledge learning tree to obtain the user's historical learning node data;

[0127] Step S22: extracting learning knowledge points according to the user's historical learning node data to obtain historical learning knowledge point data;

[0128] Step S23: matching knowledge point review questions according to historical learning knowledge point data to generate recommended review question data;

[0129] Step S24: using the historical learning knowledge point data to perform knowledge point mapping processing on the recommended review question data to generate knowledge point-question mapping table data;

[0130] Step S25: using the intelligent educational device to initialize the user's learning progress, and performing visual guidance processing according to the recommended review question data to obtain visual review question data;

[0131] Step S26: The user performs interactive learning operations on the visualized review question data and captures the behavior trajectory to generate answering behavior trajectory data.

[0132] In an embodiment of the present invention, the knowledge points that the user has learned before are marked on the personalized knowledge learning tree. For example, if the user has learned "Newton's First Law" and "Newton's Second Law" before, the corresponding nodes are specially marked on the tree. The knowledge points that the user has actually learned are extracted from the user's historical learning node data. For example, from the tree marked with the "Newton's First Law" and "Newton's Second Law" nodes, the content descriptions of these two knowledge points can be extracted to form a historical learning knowledge point data set as the basis of the knowledge that the user has mastered. Based on the user's historical learning knowledge point data, the corresponding review questions are accurately matched from the question bank. For example, for the knowledge point "Newton's First Law", questions about the uniform motion of the test object can be selected from the question bank; for the knowledge point "Newton's Second Law", questions about the relationship between the calculation force and the acceleration can be selected. The historical learning knowledge point data is mapped and associated with the recommended review question data. For example, a mapping table is established, in which one column is the knowledge point name and the other column is the corresponding review question number, thereby generating knowledge point-question mapping table data. This mapping table can clearly identify the knowledge points corresponding to each review question. According to the user's historical learning situation, the current learning progress is initialized. For example, the knowledge points that have been learned are marked as "mastered", and the knowledge points that have not been learned are marked as "to be learned". Then, the recommended review question data can be visualized, so that the review questions can be vividly displayed on the device interface in the form of pictures, texts, animations, etc., to generate visual review question data and provide users with intuitive learning content. Interactive learning with the visual review question data on the interface of the smart education device. For example, users can view the animated demonstration, text description and other content of the question, and then try to answer the question. During the user's answering process, the smart education device can capture the user's operation trajectory, including behavioral data such as dwell time, click location, and input content, to form answering behavior trajectory data.

[0133] Preferably, step S23 includes the following steps:

[0134] Step S231: performing knowledge point theme analysis on the historical learning knowledge point data to generate knowledge point theme data;

[0135] Step S232: performing knowledge topic semantic retrieval on a preset exercise question bank through knowledge point topic vector data to generate topic retrieval exercise question data;

[0136] Step S233: using a preset word embedding model to perform topic embedding processing on the topic retrieval exercise data, and using a latent Dirichlet allocation algorithm to perform text sequence topic analysis to obtain exercise topic vector data, wherein the topic embedding processing is performed by converting the exercise stem and the option text into a word vector sequence;

[0137] Step S234: performing bidirectional soft matching of the subject matter vector data of the exercises through the subject matter data of the knowledge points to generate a subject matter matching probability matrix;

[0138] Step S235: performing similarity thresholding processing on the subject matching probability matrix. When the subject similarity of the exercises in the subject matching probability matrix is ​​greater than the preset subject matching similarity threshold, the exercises are marked as initial matching exercises; when the subject similarity of the exercises in the subject matching probability matrix is ​​less than or equal to the preset subject matching similarity threshold, the exercises are marked as unmatched exercises and are eliminated;

[0139] Step S236: Perform context knowledge matching scoring on the initial matching exercises using the knowledge point theme data to generate context knowledge point scoring data;

[0140] Step S237: weighted sorting of the initial matching exercises is performed using the contextual knowledge point scoring data to obtain recommended review question data.

[0141] In an embodiment of the present invention, a topic model algorithm (such as latent Dirichlet allocation LDA) is used to perform topic analysis on the knowledge point data of the user's historical learning. For example, the text descriptions of knowledge points such as "Newton's first law", "Newton's second law", and "law of conservation of momentum" are pre-processed by word segmentation and stop word removal, and then input into the LDA model, which can automatically discover the topics covered by these knowledge points, such as "laws of motion", "force and motion", etc. These topics are semantically processed and input into the exercise question bank for topic exercise question retrieval. The topics of these questions are relatively close to the topics of the knowledge points learned by the user, and can be used as preliminary retrieval results to generate topic retrieval exercise question data. The question stem and option text in the topic retrieval exercise question data are converted into word vector sequences as input for topic embedding. Then, a topic model algorithm (such as LDA) is used to perform topic analysis on these word vector sequences, automatically discover the topics covered by the exercise question text, and represent each exercise question as a topic vector, thereby obtaining exercise question topic vector data. The knowledge point topic data and the exercise question topic vector data are bidirectionally soft matched. Specifically, the similarity scores between the topic vector of each knowledge point and the topic vectors of all exercises, as well as the similarity scores between the topic vector of each exercise and the topic vectors of all knowledge points are calculated respectively. These two scores can be used as matching probability values. Through this two-way matching, a topic matching probability matrix can be generated, and each element value in the matrix represents the topic matching probability between the corresponding knowledge point and the exercise. A topic matching similarity threshold is set, and each element value in the topic matching probability matrix (representing the topic similarity between the knowledge point and the exercise) is thresholded. If an element value is greater than the threshold, the corresponding exercise is marked as an initial matching exercise, and it is considered that the topic of the question has a high correlation with the current knowledge point; if an element value is less than or equal to the threshold, the corresponding exercise is marked as an unmatched exercise and is removed from the subsequent matching process. Through this thresholding process, a group of exercises that are closest to the knowledge point theme can be preliminarily screened out. Extract contextual knowledge information such as upper and lower concepts, axioms and theorems related to the subject from the subject data of the knowledge point, perform semantic matching on this information with the stem text of the initial matching exercises, and calculate the matching score to reflect the extent to which the exercises cover the entire content of the knowledge point. Sort exercises with higher scoring data at the front and lower scoring data at the back, thus generating a weighted sorted exercise list. This weighted sorted exercise list is the recommended review question data, in which the exercise objectives are targeted at the subject of the knowledge point learned by the user and cover the contextual information of the knowledge point, making it an ideal choice for review exercises.

[0142] Preferably, step S3 comprises the following steps:

[0143] Step S31: extracting the behavior time sequence according to the answering behavior trajectory data to generate the answering behavior time sequence data; calculating the answering speed according to the answering behavior time sequence data to generate the answering speed data;

[0144] Step S32: mining the answering behavior pattern of the answering behavior time series data through the answering speed data to obtain the answering behavior pattern data;

[0145] Step S33: extracting answering effects according to the answering behavior pattern data to generate answering effect data;

[0146] Step S34: Based on the answering behavior trajectory data, the intelligent education device is used to monitor the user's answering status in real time, and answering cognition inference is performed to generate answering cognition inference data;

[0147] Step S35: performing cognitive state clustering processing on the cognitive inference data of the answer to generate cognitive state category data;

[0148] Step S36: performing knowledge point association analysis on the cognitive state category data through the knowledge point-topic mapping table data, and performing knowledge point weight calculation to generate knowledge point cognitive weight data;

[0149] Step S37: using the answering effect data to perform answering result weighting processing on the knowledge point cognitive weight data to generate knowledge point cognitive weighted data;

[0150] Step S38: Calculate the knowledge mastery score of the knowledge point cognitive weighted data using the cognitive state category data, and perform mastery level classification processing to generate knowledge point mastery level data.

[0151] In an embodiment of the present invention, a timestamp sequence of user operations, such as the time of viewing the question, the time of making a selection, the time of clicking to confirm, etc., is extracted from the user's answering behavior trajectory data, thereby generating answering behavior time series data. Based on these time series data, the time taken by the user to answer the question, the time interval of each operation, etc. are calculated, and then the answering speed data is generated to reflect the user's answering rhythm and efficiency. The answering behavior time series data is analyzed using a sequence pattern mining algorithm (such as a prefix tree, a statistical model, etc.), and combined with the user's answering speed data, some typical behavior patterns of the user in the answering process are found. For example, some users will answer immediately after seeing the question, while other users will repeatedly check the stem options; some users will pause for a long time in the middle link, etc. By identifying and extracting these behavior patterns, answering behavior pattern data is generated. The user's answering behavior pattern is associated with the final answer result (correct or wrong), and the answering effect caused by different behavior patterns is analyzed. For example, it is found that those users who answer immediately have a higher rate of wrong answers, while those who repeatedly check the questions have a higher rate of correct answers. Combined with the pre-established cognitive model, the user's current cognitive state is inferred. For example, if a user stays on a question of a certain knowledge point for a long time and operates frequently, it can be inferred that the user has some difficulty in understanding this knowledge point; if the user pauses significantly during the answering process, it means that he or she is confused. Through this real-time cognitive inference, cognitive inference data for answering questions is generated. Cluster analysis is performed on the cognitive inference data for answering questions, and similar cognitive states are classified into one category. For example, the two states of "difficulty in understanding" and "knowledge loss" are classified into a "knowledge deficiency" category; "distracted attention" and "distracted" states are classified into a "lack of concentration" category, etc. The user's cognitive state category data is linked to the corresponding knowledge point. For example, if a user is identified as being in a "knowledge deficiency" state when answering a certain question, it can be inferred that the user has insufficient knowledge of the knowledge point corresponding to the question. Furthermore, according to the frequency of occurrence of the cognitive state of each knowledge point, the weight score of the knowledge point is calculated to generate the cognitive weight data of the knowledge point. The higher the weight, the lower the user's mastery of the knowledge point. The cognitive weight data of the knowledge point is combined with the user's answering effect data to adjust the weight value. For example, if a user often appears in the "lack of knowledge" state on questions about a certain knowledge point, but eventually answers correctly, the weight value of the knowledge point can be appropriately reduced; conversely, if the user does not show obvious cognitive problems in the process of answering questions, but eventually answers incorrectly, the weight value of the knowledge point should be appropriately increased. According to the cognitive status category data, different scoring strategies are designed and applied to the cognitive weighted data of knowledge points to generate a mastery score for each knowledge point. For example, the "lack of knowledge" state has a low score, the "lack of concentration" state has a medium score, and the "no obvious cognitive problems" state has a high score. The score can be calculated in combination with the knowledge point weight value.Afterwards, the knowledge point scores are mapped to different mastery levels, such as "mastered", "basic mastery", "needs to be strengthened", etc., and finally the knowledge point mastery level data is formed.

[0152] Preferably, step S34 includes the following steps:

[0153] Step S341: Based on the answering behavior trajectory data, the user's real-time answering status is collected using the intelligent education device to generate real-time answering status data, wherein the real-time answering status data includes answering facial expression data and eye movement trajectory data;

[0154] Step S342: performing facial expression feature analysis based on the answering facial expression data, and performing emotion label matching to generate facial expression label data;

[0155] Step S343: capturing the gaze point of the eyeball movement trajectory data to generate answer gaze point data; performing answer time series aggregation analysis on the answer gaze point data using the eyeball movement trajectory data to generate answer sight focus data;

[0156] Step S344: Calculate the duration of the gaze focus according to the gaze focus data of the answer, and generate gaze focus duration data; evaluate the concentration level according to the gaze focus duration data, and perform question relevance matching to generate answer concentration level evaluation data;

[0157] Step S345: using the answering concentration evaluation data and the facial expression label data to perform transfer learning processing on the preset recurrent neural network model, and performing cognitive state inference model training to obtain a cognitive state inference model;

[0158] Step S346: Use the cognitive state inference model to perform question cognitive state inference processing on the question answering concentration evaluation data and the facial expression label data to generate question answering cognitive inference data.

[0159] In an embodiment of the present invention, the intelligent education device can monitor the user in real time through an integrated camera and an eye tracker. During the user's answering process, the camera can capture the user's facial expression changes, analyze the expression image, extract the displacement and deformation information of the facial feature points, and form the answering facial expression data; at the same time, the eye tracker can record the movement trajectory of the user's eyeball in real time, including the displacement of the line of sight, the length of stay, etc., to form the eyeball movement trajectory data. The answering facial expression data is analyzed using computer vision and pattern recognition technology. For example, the facial key points are extracted through a pre-trained facial recognition model, and the geometric relationship between the points is calculated to obtain a facial feature vector; then the vector is input into a classification model (such as a support vector machine) to classify the facial expression and match the emotional labels such as "happy", "doubtful", and "anxious". The eyeball movement trajectory data is analyzed to detect the user's gaze position and gaze time during the answering process, thereby capturing a series of answering gaze point data. Afterwards, these gaze point data are aggregated and analyzed in time series to identify the focus area of ​​the user's line of sight, such as long-term gaze at the question stem, options, auxiliary instructions, etc., to generate answering sight focus data. The duration of the user's stay in each visual focus area is counted to generate visual focus stay duration data. For example, if the user stays longer in the question stem area and shorter in the option area, it can be judged that the user's concentration level on the question is high. These duration data are matched with the relevant areas of the question (question stem, options, etc.) to evaluate the user's concentration level on different question elements, and finally generate the answering concentration evaluation data. The answering concentration evaluation data and facial expression label data are used as training samples to perform transfer learning on the pre-set recurrent neural network model (such as long short-term memory network LSTM). Specifically, these multimodal data can be serialized and used as the input of the network; the manually annotated cognitive state labels are used as supervision signals and as the output targets of the network. During the training process, the network can automatically learn the mapping rules from answering behavior to cognitive state. The current user's answering concentration evaluation data and facial expression label data are input into the trained cognitive state inference model, and the model's reasoning ability is used to infer the user's cognitive state in real time. For example, if the cognitive state output by the model is "knowledge loss", it can be judged that the user does not have sufficient knowledge of the question; if the output is "distracted", it can be judged that the user is distracted when answering the question.

[0160] Preferably, step S4 comprises the following steps:

[0161] Step S41: labeling the knowledge point mastery level of the personalized knowledge learning tree using the knowledge point mastery level data to obtain a labeled knowledge learning tree;

[0162] Step S42: constructing a user metacognitive portrait based on cognitive state category data and annotated knowledge learning tree;

[0163] Step S43: using the annotated knowledge learning tree to perform knowledge gap analysis on the knowledge point mastery level data to generate a knowledge mastery gap matrix;

[0164] Step S44: training a Bayesian network model using a preset massive educational resource library and annotated knowledge learning tree to construct a knowledge tracking model;

[0165] Step S45: The knowledge mastery vulnerability matrix is ​​transmitted to the knowledge tracking model to generate a knowledge vulnerability repair strategy, and obtain knowledge vulnerability repair strategy data.

[0166] As an example of the present invention, refer to Figure 2 As shown, Figure 1 Detailed implementation steps of step S4 in the embodiment are shown in the flowchart. In this embodiment, step S4 includes:

[0167] Step S41: labeling the knowledge point mastery level of the personalized knowledge learning tree using the knowledge point mastery level data to obtain a labeled knowledge learning tree;

[0168] In the embodiment of the present invention, different colors or icons are used to mark each node of the personalized knowledge learning tree according to the mastery level of the knowledge point represented by the node. For example, for the knowledge tree of the course "Mechanics", if the user is evaluated as "mastered" for the knowledge point "Newton's First Law", the corresponding node is marked with green; if the user's mastery level of "Newton's Second Law" is "basic mastery", a yellow mark can be used; if the user's mastery level of "Law of Conservation of Momentum" is "not mastered", a red mark can be used.

[0169] Step S42: constructing a user metacognitive portrait based on cognitive state category data and annotated knowledge learning tree;

[0170] In an embodiment of the present invention, the cognitive state category data generated during this learning process is analyzed to understand which cognitive states the user mainly has in this learning, and the distribution of these states on different knowledge points. For example, the number of occurrences of each cognitive state (such as "knowledge loss", "distraction", etc.) in the cognitive state category data is counted, and it is mapped to the annotated knowledge learning tree to see which knowledge points these states are more concentrated on. Based on this cognitive state analysis, the learning state that the user will have in subsequent learning is predicted. At the same time, combined with the annotated knowledge learning tree, a metacognitive portrait of the user is drawn, and the metacognitive portrait of the user can be displayed to the user for visual viewing, so that teachers and learners can see the weak links and the mastered parts in the knowledge tree at a glance.

[0171] Step S43: using the annotated knowledge learning tree to perform knowledge gap analysis on the knowledge point mastery level data to generate a knowledge mastery gap matrix;

[0172] In the embodiment of the present invention, each node on the labeled knowledge learning tree is scanned to identify the knowledge points marked as "not mastered" or "basically mastered", and these knowledge points are regarded as the knowledge gaps currently existing in the user. For example, in the knowledge tree of the "mechanics" course, if the "Newton's Second Law" node is marked as "not mastered" in red, and the "law of conservation of momentum" node is marked as "basically mastered" in yellow, these two knowledge points are identified as the user's knowledge gaps. Further analyze the position of these gap knowledge points in the knowledge tree, the relationship with other knowledge points and other information, and evaluate the importance of each gap knowledge point. For example, "Newton's Second Law" is a core knowledge point, which is in a key position in the entire knowledge tree and is closely related to many other knowledge points, so the vulnerability weight of this knowledge point will be evaluated as high; and although "law of conservation of momentum" is also an important knowledge point, its position in the knowledge tree is relatively remote and has a low degree of correlation with other knowledge points, so its vulnerability weight will be evaluated as relatively low. Generate a knowledge mastery vulnerability matrix. Each element value in the matrix represents the vulnerability weight score of the corresponding knowledge point. The higher the score, the more serious the vulnerability of the knowledge point is, and it needs to be given priority attention and repaired.

[0173] Step S44: training a Bayesian network model using a preset massive educational resource library and annotated knowledge learning tree to construct a knowledge tracking model;

[0174] In the embodiment of the present invention, a massive educational resource library is built into the database, including the course outlines, teaching resources, exercise banks and other contents of various subjects. The cause-and-effect relationship between knowledge points is extracted from the educational resource library. For example, mastering the concepts of "force" and "mass" is a prerequisite for understanding "Newton's Second Law". The dependency relationship between these knowledge points can be constructed into an initial Bayesian network structure. The parameters of the Bayesian network are trained and optimized using the data in the labeled knowledge learning tree. Specifically, for each knowledge point node, the mastery state of the node in the training data is counted and used as the conditional probability value of the node; for each dependent edge, the pairing of the mastery state of the source node and the target node is counted and used as the conditional probability value. Through the learning of a large amount of training data, the Bayesian network can automatically capture the mapping relationship between knowledge points and between knowledge points and teaching resources, forming a comprehensive knowledge structure model. The knowledge tracking model can intelligently recommend relevant learning resources to users based on the user's learning situation, helping users to make up for knowledge gaps.

[0175] Step S45: The knowledge mastery vulnerability matrix is ​​transmitted to the knowledge tracking model to generate a knowledge vulnerability repair strategy, and obtain knowledge vulnerability repair strategy data.

[0176] In the embodiment of the present invention, each vulnerability knowledge point with a high weight score in the vulnerability matrix is ​​traversed and input into the Bayesian network knowledge tracking model as an evidence node. Taking the vulnerability knowledge point of "Newton's Second Law" as an example, the node is found in the Bayesian network and its mastery status is set to "not mastered". Then, the knowledge tracking model will automatically deduce the mastery of other related knowledge points according to the probability propagation rules of the network, so as to find the root knowledge point that causes the "Newton's Second Law" to be not mastered. For example, the model finds that the mastery status of the pre-knowledge points such as "force", "mass", and "acceleration" is also poor, so it is necessary to give priority to supplementing and consolidating these basic knowledge in order to effectively repair the vulnerability of "Newton's Second Law". Based on this reasoning analysis, a series of repair strategies can be generated for each vulnerability knowledge point, including pre-knowledge points that need to be learned first, suitable teaching resources, exercises, etc. These strategy data will be aggregated to form a knowledge vulnerability repair strategy data set, which provides strong support for accurate personalized teaching.

[0177] Preferably, step S5 comprises the following steps:

[0178] Step S51: The user performs knowledge vulnerability repair processing according to the knowledge vulnerability repair strategy data to generate user knowledge repair feedback data;

[0179] Step S52: performing knowledge tree updating processing on the user metacognitive portrait through the user knowledge repair feedback data to generate an updated user metacognitive portrait;

[0180] Step S53: obtaining the user's current learning target data; performing personalized learning task recommendations for the user's current learning target data by updating the user's metacognitive portrait, and performing adaptive learning difficulty adjustment to obtain adaptive learning task recommendation data;

[0181] Step S54: performing real-time learning state monitoring based on the adaptive learning task recommendation data, and performing learning cognitive inference to generate real-time metacognitive state label data;

[0182] Step S55: performing intelligent intervention guidance strategy processing according to the real-time metacognitive state tag data to generate intelligent intervention guidance strategy data, wherein the intelligent intervention guidance strategy data includes learning state prompts, knowledge point explanations, and exercise question insertions;

[0183] Step S56: Evaluate learning efficiency based on the intelligent intervention guidance strategy data to obtain intelligent guidance effect evaluation data.

[0184] As an example of the present invention, refer to Figure 3 As shown, Figure 1 Detailed implementation steps of step S5 in the flowchart, in this example, step S5 includes:

[0185] Step S51: The user performs knowledge vulnerability repair processing according to the knowledge vulnerability repair strategy data to generate user knowledge repair feedback data;

[0186] In an embodiment of the present invention, the intelligent education device will present the knowledge loophole repair strategy data to the user, and the user can learn the corresponding knowledge points and teaching resources according to these strategies. Taking the "Mechanics" course as an example, if the user previously had a large loophole in the "Newton's Second Law" knowledge point, the intelligent education device will push the repair strategy for this knowledge point. For example, the key analysis video, related concept learning materials, basic exercises and other resources of "Newton's Second Law" are displayed on the interface, and these resources are arranged according to a certain learning path to guide users to learn and review systematically. Users can watch videos, read materials, do exercises, etc. in an orderly manner according to the recommended repair strategy. In this process, the intelligent education device will record the user's learning behavior trajectory in real time, such as video viewing time, material reading times, exercise answering, etc., and these data constitute the user's knowledge repair feedback data. For example, if the user has a high correct rate in the basic exercises of "Newton's Second Law", but makes many mistakes in the deep application questions, it means that the user has a certain grasp of the basic concept of the knowledge point, but is still insufficient in flexible application. These feedback data can reflect the effect of the user's repair learning of the knowledge point.

[0187] Step S52: performing knowledge tree updating processing on the user metacognitive portrait through the user knowledge repair feedback data to generate an updated user metacognitive portrait;

[0188] In an embodiment of the present invention, after the user learns and generates feedback data according to the knowledge gap repair strategy, the user's knowledge repair feedback data is analyzed to evaluate whether the user's mastery of the knowledge points that previously had gaps has been improved. Taking "Newton's Second Law" as an example, if the feedback data shows that the user has a high correct rate on the basic exercises of this knowledge point and sufficient video viewing time, it will be judged that the user has a certain mastery of this knowledge point. In this case, the node corresponding to the knowledge point of "Newton's Second Law" will be found in the original metacognitive portrait, and its previous red "not mastered" mark will be updated to a yellow "basic mastery" mark. Similarly, if the feedback data shows that the user also has a good performance on the deep application questions of this knowledge point, the mark will be further updated to the green "mastered" state, generating a new updated user metacognitive portrait, which can more accurately and comprehensively reflect the user's current actual knowledge mastery level and cognitive state than the previous version.

[0189] Step S53: obtaining the user's current learning target data; performing personalized learning task recommendations for the user's current learning target data by updating the user's metacognitive portrait, and performing adaptive learning difficulty adjustment to obtain adaptive learning task recommendation data;

[0190] In an embodiment of the present invention, after the knowledge loopholes are repaired and the metacognitive portrait is updated, the target data of the user's current learning is obtained. This can be a learning outline pre-set by a teacher or a course system, or it can be a learning direction selected by the user. For example, the user's learning goal this time is to master the knowledge point of "law of conservation of momentum". In the updated metacognitive portrait, the mastery of other knowledge points related to this knowledge point is analyzed. If the portrait shows that the user has basically mastered the pre-knowledge points such as "momentum" and "kinetic energy", but has a cognitive state label that is difficult to understand in the concept of "impulse", the user will be recommended entry-level learning tasks of the concept of "impulse", such as watching related concept videos, doing simple calculation exercises, etc. If the portrait shows that the user is already proficient in all pre-knowledge points, the deep learning resources of "law of conservation of momentum" are directly recommended, such as detailed analysis of formula derivation, experimental videos, intermediate application questions, etc. While recommending learning tasks, the difficulty level of the tasks is automatically adjusted according to the overall knowledge level of the user. For example, for beginners, the recommended tasks will focus more on basic concepts and calculations; while for users with advanced academic skills, more exploratory and open learning content will be added.

[0191] Step S54: performing real-time learning state monitoring based on the adaptive learning task recommendation data, and performing learning cognitive inference to generate real-time metacognitive state label data;

[0192] In the embodiment of the present invention, when the user starts learning according to the adaptive learning task recommendation data, the user's learning behavior trajectory, facial expression, eye movement and other data are monitored in real time, and these data are input into the previously trained cognitive state inference model. The model can infer the user's current cognitive state, such as "focused learning", "knowledge loss", "attention dispersed", etc. based on the user's multimodal behavior data, and generate corresponding real-time metacognitive state label data. These label data can reflect the user's cognitive weaknesses in the process of learning new knowledge points.

[0193] Step S55: performing intelligent intervention guidance strategy processing according to the real-time metacognitive state tag data to generate intelligent intervention guidance strategy data, wherein the intelligent intervention guidance strategy data includes learning state prompts, knowledge point explanations, and exercise question insertions;

[0194] In the embodiment of the present invention, a personalized intelligent intervention guidance strategy is automatically generated based on the real-time metacognitive state tag data. If the tag data shows that the user is in a "knowledge missing" state, the intelligent education device will insert a text explanation or multimedia explanation of the corresponding knowledge point on the learning interface; if the user is in a "dispersed" state, the intelligent education device will pop up a learning status prompt to guide the user to concentrate; if the user is in a "focused learning" state for a long time, the intelligent education device will insert relevant exercises in a timely manner to test and deepen the knowledge mastery. All these personalized guidance strategies will be encoded as intelligent intervention guidance strategy data.

[0195] Step S56: Evaluate learning efficiency based on the intelligent intervention guidance strategy data to obtain intelligent guidance effect evaluation data.

[0196] In an embodiment of the present invention, during the execution of the intelligent intervention guidance strategy, the intelligent education device will continuously monitor the user's learning behavior data and analyze the impact of the guidance strategy on the user's learning efficiency. For example, if after inserting the knowledge point explanation, the user's eye movement trajectory shows that he has been fixated on the explanation content for a long time, it means that the explanation strategy is effective; if after the learning status prompt pops up, the user's concentration is improved, it means that the prompt strategy is effective. By analyzing these effects, the effectiveness of different guidance strategies is evaluated, and the evaluation results are encoded as intelligent guidance effect evaluation data. The effect evaluation data of all different strategies will be summarized to form an intelligent guidance effect evaluation data set. This data set will serve as an important reference for self-iteration and strategy optimization of intelligent education devices.

[0197] The present invention also provides a guidance system for an intelligent educational device based on a learning algorithm, which executes the guidance method for an intelligent educational device based on a learning algorithm as described above. The guidance system for an intelligent educational device based on a learning algorithm comprises:

[0198] The knowledge learning tree construction module obtains the user's learning course outline data; performs knowledge ontology structure processing based on the user's learning course outline data to generate knowledge ontology structure data; performs personalized knowledge tree reconstruction based on the knowledge ontology structure data to generate a personalized knowledge learning tree;

[0199] The knowledge point review recommendation module is used to extract historical learning knowledge points according to the personalized knowledge learning tree to obtain historical learning knowledge point data; match knowledge point review questions according to the historical learning knowledge point data to generate recommended review question data; use the historical learning knowledge point data to perform knowledge point mapping processing on the recommended review question data to generate knowledge point-question mapping table data; perform interactive learning operations according to the recommended review question data to obtain answering behavior trajectory data;

[0200] The cognitive state assessment module is used to monitor the user's real-time answering state based on the answering behavior trajectory data, and to make cognitive inferences about answering questions to generate cognitive state category data; to perform knowledge point weight processing on the cognitive state category data using the knowledge point-question mapping table data to generate knowledge point cognitive weight data; to perform knowledge mastery score processing based on the knowledge point cognitive weight data to generate knowledge point mastery level data;

[0201] The knowledge leak tracking module is used to build a user metacognitive portrait based on cognitive state category data and personalized knowledge learning tree; perform knowledge leak analysis based on knowledge point mastery level data to generate a knowledge mastery leak matrix; perform knowledge tracking processing based on the knowledge mastery leak matrix, and generate knowledge leak repair strategies to obtain knowledge leak repair strategy data;

[0202] The intelligent learning guidance module is used for users to perform knowledge loophole repair processing through knowledge loophole repair strategy data to generate user knowledge repair feedback data; to recommend personalized learning tasks through user knowledge repair feedback data and user metacognitive portraits to obtain adaptive learning task recommendation data; to perform intelligent intervention guidance strategy processing based on adaptive learning task recommendation data to obtain intelligent guidance effect evaluation data.

[0203] The beneficial effect of this application is that, by structuring the course outline into knowledge ontology, the course content can be decomposed into more fine-grained knowledge points and the association relationship between knowledge points can be established. On this basis, according to the user's learning goals and existing knowledge base, a personalized knowledge learning tree is constructed for each user, and the learning content is reasonably sorted and organized to guide the user to learn in the best order and improve learning efficiency. By extracting the knowledge points that the user has learned in the past, review questions can be recommended in a targeted manner to help users consolidate the knowledge they have learned and improve their knowledge mastery. The user's answering behavior can be tracked in real time, and the user's cognitive state can be inferred based on indicators such as answering time and answering accuracy, such as whether the knowledge points are understood, whether the problem-solving methods are mastered, etc., to help users find learning problems in time and adjust learning strategies. By analyzing the knowledge point mastery level data, the user's learning loopholes can be discovered in time, and learning resources can be recommended in a targeted manner to help users overcome learning difficulties and improve learning effects. By analyzing the knowledge mastery loophole matrix, the user's knowledge loopholes can be accurately located to help users understand their own learning situation and formulate learning plans in a targeted manner. Based on the knowledge mastery gap matrix, smart education devices can provide users with personalized knowledge gap repair strategies to guide users to learn in a targeted manner and improve learning efficiency. Users learn according to the knowledge gap repair strategies provided by the device and generate knowledge repair feedback data to reflect the user's acceptance of the knowledge repair strategy and learning effect. Based on the user's knowledge repair feedback data and metacognitive portrait, personalized learning tasks are recommended to users to help users consolidate learning content and expand knowledge. Intelligent guidance and intervention of the user's learning process improves learning quality and realizes adaptive personalized learning.

[0204] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0205] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for guiding an intelligent educational device based on a learning algorithm, characterized in that: The following steps are involved: Step S1: Obtaining user learning course outline data; Perform knowledge ontology structuring processing based on the user's course syllabus data to generate knowledge ontology structure data; Reconstruct the personalized knowledge tree according to the knowledge ontology structure data to generate a personalized knowledge learning tree; Step S2: extracting historical learning knowledge points according to the personalized knowledge learning tree to obtain historical learning knowledge point data; Matching knowledge point review questions according to historical learning knowledge point data to generate recommended review question data; using historical learning knowledge point data to perform knowledge point mapping processing on recommended review question data to generate knowledge point-question mapping table data; performing interactive learning operations according to recommended review question data to obtain answering behavior trajectory data; wherein step S2 is specifically as follows: Step S21: Mark the user's historical learning knowledge nodes through the personalized knowledge learning tree to obtain the user's historical learning node data; Step S22: extracting learning knowledge points according to the user's historical learning node data to obtain historical learning knowledge point data; Step S23: matching knowledge point review questions according to historical learning knowledge point data to generate recommended review question data; wherein step S23 is specifically as follows: Step S231: performing knowledge point theme analysis on the historical learning knowledge point data to generate knowledge point theme data; Step S232: performing knowledge topic semantic retrieval on a preset exercise question bank through knowledge point topic vector data to generate topic retrieval exercise question data; Step S233: using a preset word embedding model to perform topic embedding processing on the topic retrieval exercise data, and using a latent Dirichlet allocation algorithm to perform text sequence topic analysis to obtain exercise topic vector data, wherein the topic embedding processing is performed by converting the exercise stem and the option text into a word vector sequence; Step S234: performing bidirectional soft matching of the subject matter vector data of the exercises through the subject matter data of the knowledge points to generate a subject matter matching probability matrix; Step S235: performing similarity thresholding processing on the subject matching probability matrix. When the subject similarity of the exercises in the subject matching probability matrix is ​​greater than the preset subject matching similarity threshold, the exercises are marked as initial matching exercises; when the subject similarity of the exercises in the subject matching probability matrix is ​​less than or equal to the preset subject matching similarity threshold, the exercises are marked as unmatched exercises and are eliminated; Step S236: Perform context knowledge matching scoring on the initial matching exercises using the knowledge point theme data to generate context knowledge point scoring data; Step S237: performing weighted sorting processing on the initial matching exercises list according to the context knowledge point scoring data to obtain recommended review questions data; Step S24: using the historical learning knowledge point data to perform knowledge point mapping processing on the recommended review question data to generate knowledge point-question mapping table data; Step S25: using the intelligent educational device to initialize the user's learning progress, and performing visual guidance processing according to the recommended review question data to obtain visual review question data; Step S26: the user performs interactive learning operations on the visualized review question data, and captures the behavior trajectory to generate answering behavior trajectory data; Step S3: Based on the answering behavior trajectory data, the user's real-time answering status is monitored, and the answering cognitive inference is performed to generate cognitive status category data; the cognitive status category data is processed with knowledge point weights using the knowledge point-question mapping table data to generate knowledge point cognitive weight data; the knowledge mastery score is processed according to the knowledge point cognitive weight data to generate knowledge point mastery level data; wherein step S3 is specifically as follows: Step S31: extracting the behavior time sequence according to the answering behavior trajectory data to generate the answering behavior time sequence data; calculating the answering speed according to the answering behavior time sequence data to generate the answering speed data; Step S32: mining the answering behavior pattern of the answering behavior time series data through the answering speed data to obtain the answering behavior pattern data; Step S33: extracting answering effects according to the answering behavior pattern data to generate answering effect data; Step S34: Based on the answering behavior trajectory data, the intelligent education device is used to monitor the user's answering status in real time, and answering cognition inference is performed to generate answering cognition inference data; Step S35: performing cognitive state clustering processing on the cognitive inference data of the answer to generate cognitive state category data; Step S36: performing knowledge point association analysis on the cognitive state category data through the knowledge point-topic mapping table data, and performing knowledge point weight calculation to generate knowledge point cognitive weight data; Step S37: using the answering effect data to perform answering result weighting processing on the knowledge point cognitive weight data to generate knowledge point cognitive weighted data; Step S38: Calculate the knowledge mastery score of the knowledge point cognitive weighted data using the cognitive state category data, and perform mastery level classification processing to generate knowledge point mastery level data; Step S4: construct a user metacognitive portrait based on cognitive state category data and personalized knowledge learning tree; perform knowledge gap analysis based on knowledge point mastery level data to generate a knowledge mastery gap matrix; perform knowledge tracking processing based on the knowledge mastery gap matrix, and generate knowledge gap repair strategies to obtain knowledge gap repair strategy data; Step S5: The user performs knowledge gap repair processing through the knowledge gap repair strategy data to generate user knowledge repair feedback data; personalized learning task recommendations are performed through the user knowledge repair feedback data and the user metacognitive portrait to obtain adaptive learning task recommendation data; intelligent intervention guidance strategy processing is performed based on the adaptive learning task recommendation data to obtain intelligent guidance effect evaluation data.

2. The method for guiding an intelligent educational device based on a learning algorithm according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Acquire user learning course outline data and user learning log data; Step S12: Performing semantic analysis on the user's course syllabus data to generate course knowledge semantic network data; Step S13: Perform knowledge ontology structuring processing according to the course knowledge semantic network data to generate knowledge ontology structure data; Step S14: identifying knowledge point association relationships of knowledge ontology structure data through course knowledge semantic network data to generate knowledge point association data; Step S15: Analyze the knowledge point concept hierarchy of the knowledge ontology structure data through the knowledge point association data, and construct a knowledge tree to generate knowledge component tree data; Step S16: Use the user learning log data to reconstruct the knowledge component tree data into a personalized knowledge tree to generate a personalized knowledge learning tree.

3. The intelligent educational device guidance method based on learning algorithm according to claim 2 is characterized in that: Step S16 includes the following steps: Step S161: labeling the knowledge component importance weights of the knowledge component tree data using the preset knowledge importance weights to generate an important knowledge component tree; Step S162: extracting user learning behavior features from the user learning log data to generate user learning behavior feature data, wherein the user learning behavior feature data includes learning content data, learning sequence data, and learning duration data; Step S163: Plan important learning nodes for the important knowledge component tree using the learning sequence data to generate important learning node data; Step S164: performing knowledge content mapping processing on the important knowledge component tree through the learning content data to obtain user-knowledge component interaction data; Step S165: Performing user learning preference analysis based on the learning time data through the user-knowledge component interaction data to generate user learning preference data; Step S166: Calculate the knowledge point relevance of the important knowledge component tree according to the user learning preference data to generate knowledge point relevance matrix data; Step S167: Perform knowledge point clustering analysis based on the knowledge point correlation matrix data, and use the important learning node data to reconstruct the important knowledge component tree into a personalized learning tree to generate a personalized knowledge learning tree.

4. The intelligent educational device guidance method based on learning algorithm according to claim 1, characterized in that: Step S34 includes the following steps: Step S341: Based on the answering behavior trajectory data, the user's real-time answering status is collected using the intelligent education device to generate real-time answering status data, wherein the real-time answering status data includes answering facial expression data and eye movement trajectory data; Step S342: performing facial expression feature analysis based on the answering facial expression data, and performing emotion label matching to generate facial expression label data; Step S343: capturing the gaze point of the eyeball movement trajectory data to generate answer gaze point data; performing answer time series aggregation analysis on the answer gaze point data using the eyeball movement trajectory data to generate answer sight focus data; Step S344: Calculate the duration of the gaze focus according to the gaze focus data of the answer, and generate gaze focus duration data; evaluate the concentration level according to the gaze focus duration data, and perform question relevance matching to generate answer concentration level evaluation data; Step S345: using the answering concentration evaluation data and the facial expression label data to perform transfer learning processing on the preset recurrent neural network model, and performing cognitive state inference model training to obtain a cognitive state inference model; Step S346: Use the cognitive state inference model to perform question cognitive state inference processing on the question answering concentration evaluation data and the facial expression label data to generate question answering cognitive inference data.

5. The intelligent educational device guidance method based on learning algorithm according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: labeling the knowledge point mastery level of the personalized knowledge learning tree using the knowledge point mastery level data to obtain a labeled knowledge learning tree; Step S42: constructing a user metacognitive portrait based on cognitive state category data and annotated knowledge learning tree; Step S43: using the annotated knowledge learning tree to perform knowledge gap analysis on the knowledge point mastery level data to generate a knowledge mastery gap matrix; Step S44: training a Bayesian network model using a preset massive educational resource library and annotated knowledge learning tree to construct a knowledge tracking model; Step S45: The knowledge mastery vulnerability matrix is ​​transmitted to the knowledge tracking model to generate a knowledge vulnerability repair strategy, and obtain knowledge vulnerability repair strategy data.

6. The method for guiding an intelligent educational device based on a learning algorithm according to claim 1, characterized in that: Step S5 includes the following steps: Step S51: The user performs knowledge vulnerability repair processing according to the knowledge vulnerability repair strategy data to generate user knowledge repair feedback data; Step S52: performing knowledge tree updating processing on the user metacognitive portrait through the user knowledge repair feedback data to generate an updated user metacognitive portrait; Step S53: obtaining the user's current learning target data; performing personalized learning task recommendations for the user's current learning target data by updating the user's metacognitive portrait, and performing adaptive learning difficulty adjustment to obtain adaptive learning task recommendation data; Step S54: performing real-time learning state monitoring based on the adaptive learning task recommendation data, and performing learning cognitive inference to generate real-time metacognitive state label data; Step S55: performing intelligent intervention guidance strategy processing according to the real-time metacognitive state tag data to generate intelligent intervention guidance strategy data, wherein the intelligent intervention guidance strategy data includes learning state prompts, knowledge point explanations, and exercise question insertions; Step S56: Evaluate learning efficiency based on the intelligent intervention guidance strategy data to obtain intelligent guidance effect evaluation data.

7. An intelligent educational device guidance system based on a learning algorithm, characterized in that: For executing the intelligent education device guidance method based on learning algorithm as claimed in claim 1, the intelligent education device guidance system based on learning algorithm comprises: The knowledge learning tree construction module obtains the user's learning course outline data; performs knowledge ontology structure processing based on the user's learning course outline data to generate knowledge ontology structure data; performs personalized knowledge tree reconstruction based on the knowledge ontology structure data to generate a personalized knowledge learning tree; The knowledge point review recommendation module is used to extract historical learning knowledge points according to the personalized knowledge learning tree to obtain historical learning knowledge point data; match knowledge point review questions according to the historical learning knowledge point data to generate recommended review question data; use the historical learning knowledge point data to perform knowledge point mapping processing on the recommended review question data to generate knowledge point-question mapping table data; perform interactive learning operations according to the recommended review question data to obtain answering behavior trajectory data; the knowledge point review recommendation module is used to: Step S21: Mark the user's historical learning knowledge nodes through the personalized knowledge learning tree to obtain the user's historical learning node data; Step S22: extracting learning knowledge points according to the user's historical learning node data to obtain historical learning knowledge point data; Step S23: matching knowledge point review questions according to historical learning knowledge point data to generate recommended review question data; wherein step S23 is specifically as follows: Step S231: performing knowledge point theme analysis on the historical learning knowledge point data to generate knowledge point theme data; Step S232: performing knowledge topic semantic retrieval on a preset exercise question bank through knowledge point topic vector data to generate topic retrieval exercise question data; Step S233: using a preset word embedding model to perform topic embedding processing on the topic retrieval exercise data, and using a latent Dirichlet allocation algorithm to perform text sequence topic analysis to obtain exercise topic vector data, wherein the topic embedding processing is performed by converting the exercise stem and the option text into a word vector sequence; Step S234: performing bidirectional soft matching of the subject matter vector data of the exercises through the subject matter data of the knowledge points to generate a subject matter matching probability matrix; Step S235: performing similarity thresholding processing on the subject matching probability matrix. When the subject similarity of the exercises in the subject matching probability matrix is ​​greater than the preset subject matching similarity threshold, the exercises are marked as initial matching exercises; when the subject similarity of the exercises in the subject matching probability matrix is ​​less than or equal to the preset subject matching similarity threshold, the exercises are marked as unmatched exercises and are eliminated; Step S236: Perform context knowledge matching scoring on the initial matching exercises using the knowledge point theme data to generate context knowledge point scoring data; Step S237: performing weighted sorting processing on the initial matching exercises list according to the context knowledge point scoring data to obtain recommended review questions data; Step S24: using the historical learning knowledge point data to perform knowledge point mapping processing on the recommended review question data to generate knowledge point-question mapping table data; Step S25: using the intelligent educational device to initialize the user's learning progress, and performing visual guidance processing according to the recommended review question data to obtain visual review question data; Step S26: the user performs interactive learning operations on the visualized review question data, and captures the behavior trajectory to generate answering behavior trajectory data; The cognitive state evaluation module is used to monitor the user's real-time answering state based on the answering behavior trajectory data, and to perform answering cognitive inference to generate cognitive state category data; to perform knowledge point weight processing on the cognitive state category data using the knowledge point-question mapping table data to generate knowledge point cognitive weight data; to perform knowledge mastery score processing based on the knowledge point cognitive weight data to generate knowledge point mastery level data; the cognitive state evaluation module is used to: Step S31: extracting the behavior time sequence according to the answering behavior trajectory data to generate the answering behavior time sequence data; calculating the answering speed according to the answering behavior time sequence data to generate the answering speed data; Step S32: mining the answering behavior pattern of the answering behavior time series data through the answering speed data to obtain the answering behavior pattern data; Step S33: extracting answering effects according to the answering behavior pattern data to generate answering effect data; Step S34: Based on the answering behavior trajectory data, the intelligent education device is used to monitor the user's answering status in real time, and answering cognition inference is performed to generate answering cognition inference data; Step S35: performing cognitive state clustering processing on the cognitive inference data of the answer to generate cognitive state category data; Step S36: performing knowledge point association analysis on the cognitive state category data through the knowledge point-topic mapping table data, and performing knowledge point weight calculation to generate knowledge point cognitive weight data; Step S37: using the answering effect data to perform answering result weighting processing on the knowledge point cognitive weight data to generate knowledge point cognitive weighted data; Step S38: Calculate the knowledge mastery score of the knowledge point cognitive weighted data using the cognitive state category data, and perform mastery level classification processing to generate knowledge point mastery level data; The knowledge leak tracking module is used to build a user metacognitive portrait based on cognitive state category data and personalized knowledge learning tree; perform knowledge leak analysis based on knowledge point mastery level data to generate a knowledge mastery leak matrix; perform knowledge tracking processing based on the knowledge mastery leak matrix, and generate knowledge leak repair strategies to obtain knowledge leak repair strategy data; The intelligent learning guidance module is used for users to perform knowledge loophole repair processing through knowledge loophole repair strategy data to generate user knowledge repair feedback data; to recommend personalized learning tasks through user knowledge repair feedback data and user metacognitive portraits to obtain adaptive learning task recommendation data; to perform intelligent intervention guidance strategy processing based on adaptive learning task recommendation data to obtain intelligent guidance effect evaluation data.