Intelligent adjustment method, system and device for student learning and storage medium

Abstract

The invention relates to the technical field of education, in particular to an intelligent adjustment method, system and device for student learning and a storage medium. In order to solve the problems that in the prior art, electronic education equipment is difficult to meet individual requirements of students and is difficult to improve learning quality, the method comprises the following steps of: selecting whether to execute repeated learning operation or not according to stage attention in real time in a learning stage; in the test stage, corresponding consolidation learning operation is selected to be executed based on the answering time and accuracy of each test question; besides, after all the questions are answered, chapter and section absorption degrees are obtained, a graph structure is constructed with chapter and section contents and student personal information, student learning preference information is extracted, and learning comprehensive features are obtained; based on the learning comprehensive features, the student personal features and the to-be-recommended learning method features, obtaining a recommendation value; and finally, updating the learning content form of the next chapter based on the to-be-recommended learning method of which the recommendation value is greater than the recommendation threshold so as to meet the individual learning requirements of the students and improve the learning quality of the students.

Description

Technical Field

[0001] The present invention relates to the field of educational technology, and in particular to an intelligent adjustment method, system, device and storage medium for student learning. Background Art

[0002] Traditional education often uses teachers and textbooks to provide learning content, monitor students' learning attention, and evaluate students' learning effects. However, a teacher often has to face many students. Due to the teaching progress and other reasons, it is difficult to pay attention to the learning behavior of each student at the same time in the classroom. It is difficult to provide timely feedback on each student's problematic behavior. Due to time and energy issues, it is impossible to provide personalized teaching plans based on students' specific needs. At the same time, the current feedback on learning effects is often concentrated after homework and exams, which has a certain lag. When teachers find problems, students may have gone further on the wrong path.

[0003] In order to improve the quality and efficiency of education, electronic devices and online education platforms have gradually been introduced into teaching. However, these learning systems generally only provide standardized learning content, which makes it difficult to meet students' personalized needs. They are also unable to make timely adjustments to learning content and methods based on students' learning situation, making it difficult to improve students' learning quality. Summary of the invention

[0004] The purpose of the present invention is to provide an intelligent adjustment method, system, device and storage medium for student learning.

[0005] The technical solution of the present invention is as follows: An intelligent adjustment method for student learning includes the following operations: S1. During the process of students learning the content of the current chapter, based on the students' learning behavior at the current stage, the current stage attention is obtained; whether the current stage attention exceeds the attention threshold is determined; if the current stage attention is not greater than the first attention threshold, the current stage content is re-displayed on the learning desktop; if the current stage attention is between the first attention threshold and the second attention threshold, after the learning desktop displays a warning reminder, the next stage content is displayed on the learning desktop; if the current stage attention is not less than the second attention threshold, the next stage content is displayed on the learning desktop; the first attention threshold is less than the second attention threshold; after all the contents of the last stage of the current chapter are displayed, the learning attention of all stages is weighted and summed with the corresponding stage content weight to obtain the chapter comprehensive attention; if the chapter comprehensive attention is not greater than the comprehensive attention threshold, the learning content in the current chapter content whose stage content weight is greater than the content weight threshold is displayed on the learning desktop as the re-learned content, and after all the contents in the re-learned content are displayed, S2 is executed; if the chapter comprehensive attention is greater than the comprehensive attention threshold, S2 is executed; S2. Obtain the answering time and correct rate of each test question in the process of completing the test question bank of the current chapter, and perform the answering judgment operation; in the answering judgment operation, if the answering time of the current test question is less than the answering time threshold, and the correct rate is not less than the correct rate threshold, the next test question will be displayed on the learning desktop; if the correct rate of the current test question is less than the correct rate threshold, the learning content corresponding to the current test question will be displayed on the learning desktop; after all the learning content corresponding to the current test question is displayed, the current test question and the consolidation questions with a content similarity greater than the similarity threshold to the current test question will be displayed on the learning desktop, and the answering judgment operation will be performed; the consolidation questions are in the consolidation question bank of the current chapter; if the answering time of the current test question is not less than the answering time threshold, but the correct rate is not less than the correct rate threshold, the consolidation questions with a content similarity greater than the similarity threshold to the current test question will be displayed on the learning desktop, and the answering judgment operation will be performed; after all the questions are answered, the absorption degree of the current chapter is obtained based on the total answering time and the comprehensive correct rate of the first answer of the test question bank of the current chapter; S3. Based on the content of the current chapter, the absorption level of the current chapter and the personal information of the students, a graph structure is constructed. The graph structure is processed by graph convolution to obtain comprehensive learning features. The comprehensive learning features and the personal characteristics of the students, as well as the characteristics of the learning method to be recommended, are processed by probability mapping to obtain a recommendation value. Based on the learning method to be recommended whose recommendation value is greater than the recommendation threshold, the form of the learning content of the next chapter is updated.

[0006] The specific operation of obtaining the current learning attention in S1 is: obtaining the student learning behavior video of the current stage, identifying the student head turning event, the student line of sight deviation event, the student expression abnormal event, and the student body posture abnormal event from the student learning behavior video, and forming the current stage learning abnormal behavior set; the corresponding score of each event in the current stage learning abnormal behavior set, and the learning interface jump event score obtained based on the number of learning interface jumps and the jump interface type, after weighted summation processing, the difference between the total score and the score is the current learning attention.

[0007] The method for obtaining the content similarity in S2 is: perform word segmentation on the current test question and the consolidation question respectively to obtain the current test question word set and the consolidation question word set; take the ratio of the number of intersection elements of the current test question word set and the consolidation question word set to the number of union elements of the current test question word set and the consolidation question word set as the content similarity between the current test question and the consolidation question.

[0008] In S3, the operation of updating the form of the learning content of the next chapter based on the learning method to be recommended whose recommendation value is greater than the recommendation threshold is specifically as follows: if the learning method to be recommended whose recommendation value is greater than the recommendation threshold is a picture display type learning method, then in the form of the learning content of the next chapter, the picture display duration is greater than the first display duration; if the learning method to be recommended whose recommendation value is greater than the recommendation threshold is an audio-visual type learning method, then in the form of the learning content of the next chapter, the video display duration is greater than the first display duration; if the learning method to be recommended whose recommendation value is greater than the recommendation threshold is a question-solving type learning method, then in the form of the learning content of the next chapter, the sum of the question display duration and the answer display duration is greater than the first display duration.

[0009] After obtaining the absorption level of the next chapter, determine whether the difference between the absorption level of the next chapter and the absorption level of the current chapter is greater than the absorption level difference threshold; if not, do not update the learning content format of the next chapter; if greater, add the next chapter content and the absorption level of the next chapter to the current graph structure to obtain an updated graph structure for performing graph convolution processing operations.

[0010] The nodes in the updated graph structure are chapters and students, and the edges are the chapter absorption degree and the degree of association between chapters. The degree of association between chapters is to convert the contents of two chapters into vector form to obtain the first chapter vector and the second chapter vector. The cosine similarity between the first chapter vector and the second chapter vector is used as the degree of association between chapters.

[0011] The personal characteristics of S3 middle school students are obtained based on information on their age, gender, grade, and hobbies.

[0012] An intelligent adjustment system for student learning, used to implement the above-mentioned intelligent adjustment method for student learning, comprising: The learning stage adjustment module is used for obtaining the current stage attention degree based on the student learning behavior in the current stage during the process of students learning the current chapter content; judging whether the current stage attention degree exceeds the attention degree threshold; if the current stage attention degree is not greater than the first attention degree threshold, the current stage content is re-displayed on the learning desktop; if the current stage attention degree is between the first attention degree threshold and the second attention degree threshold, after the learning desktop displays a warning reminder, the next stage content is displayed on the learning desktop; if the current stage attention degree is not less than the second attention degree threshold, the next stage content is displayed on the learning desktop; the first attention threshold is less than the second attention threshold; after all the contents of the last stage of the current chapter are displayed, the learning attention degree of all stages is weighted and summed with the corresponding stage content weight to obtain the chapter comprehensive attention degree; if the chapter comprehensive attention degree is not greater than the comprehensive attention threshold, the learning content in the current chapter content whose stage content weight is greater than the content weight threshold is displayed on the learning desktop as the re-learned content, and after all the contents in the re-learned content are displayed, the test stage adjustment module is executed; if the chapter comprehensive attention degree is greater than the comprehensive attention threshold, the test stage adjustment module is executed; The test phase adjustment module is used to obtain the answering time and correct rate of each test question in the process of completing the test question bank of the current chapter, and perform the answering judgment operation; in the answering judgment operation, if the answering time of the current test question is less than the answering time threshold, and the correct rate is not less than the correct rate threshold, the next test question will be displayed on the learning desktop; if the correct rate of the current test question is less than the correct rate threshold, the learning content corresponding to the current test question will be displayed on the learning desktop; after all the learning content corresponding to the current test question is displayed, the current test question and the consolidation questions with a content similarity greater than the similarity threshold to the current test question will be displayed on the learning desktop, and the answering judgment operation will be performed; the consolidation questions are in the consolidation question bank of the current chapter; if the answering time of the current test question is not less than the answering time threshold, but the correct rate is not less than the correct rate threshold, the consolidation questions with a content similarity greater than the similarity threshold to the current test question will be displayed on the learning desktop, and the answering judgment operation will be performed; after all questions are answered, the absorption degree of the current chapter is obtained based on the total answering time and the comprehensive correct rate of the first answer of the test question bank of the current chapter; The learning content form update module is used to construct a graph structure based on the current chapter content, the current chapter absorption level and student personal information; the graph structure is processed by graph convolution to obtain comprehensive learning features; the comprehensive learning features and the student's personal features, as well as the features of the learning method to be recommended, are processed by probability mapping to obtain a recommendation value; based on the learning method to be recommended whose recommendation value is greater than the recommendation threshold, the learning content form of the next chapter is updated.

[0013] An intelligent adjustment device for student learning comprises a processor and a memory, wherein the processor implements the above-mentioned intelligent adjustment method for student learning when executing a computer program stored in the memory.

[0014] A computer-readable storage medium is used to store a computer program, wherein the computer program, when executed by a processor, implements the above-mentioned intelligent adjustment method for student learning.

[0015] The beneficial effects of the present invention are: The present invention provides an intelligent adjustment method for student learning. In the learning stage, the student's attention at each stage is obtained in real time to predict the student's learning effect in real time, and according to the comparison result of the stage attention and the attention threshold, whether to perform repeated learning operations is selected to ensure the learning quality of students in the learning stage; in the test stage, the student's mastery of the learning content is evaluated in real time by obtaining the answering time and accuracy of each test question, and the corresponding consolidation learning operation is selected according to the different mastery of the students, so as to continuously strengthen the mastery of knowledge by learning; in addition, after all questions are answered, the total answering time is calculated based on the total answering time. and the comprehensive accuracy rate of the first answer to the test questions to obtain the degree of chapter absorption; then, based on the current chapter content, the current chapter absorption degree and student personal information, a graph structure that can intuitively reflect the student's learning of the chapter content is constructed, and after graph convolution processing, the student's learning preference information in the process of learning the chapter content is extracted to obtain the comprehensive learning features; the comprehensive learning features are probabilistically mapped with the student's personal characteristics and the characteristics of the learning method to be recommended to obtain the recommendation value; finally, based on the recommended learning method whose recommendation value is greater than the recommendation threshold, the form of the learning content of the next chapter is updated to meet the students' personalized learning needs, thereby improving the learning quality. DETAILED DESCRIPTION

[0016] This embodiment provides an intelligent adjustment method for student learning, including the following operations: S1. During the process of students learning the content of the current chapter, based on the students' learning behavior at the current stage, the current stage attention is obtained; whether the current stage attention exceeds the attention threshold is determined; if the current stage attention is not greater than the first attention threshold, the current stage content is re-displayed on the learning desktop; if the current stage attention is between the first attention threshold and the second attention threshold, after the learning desktop displays a warning reminder, the next stage content is displayed on the learning desktop; if the current stage attention is not less than the second attention threshold, the next stage content is displayed on the learning desktop; the first attention threshold is less than the second attention threshold; after all the contents of the last stage of the current chapter are displayed, the learning attention of all stages is weighted and summed with the corresponding stage content weight to obtain the chapter comprehensive attention; if the chapter comprehensive attention is not greater than the comprehensive attention threshold, the learning content in the current chapter content whose stage content weight is greater than the content weight threshold is displayed on the learning desktop as the re-learned content, and after all the contents in the re-learned content are displayed, S2 is executed; if the chapter comprehensive attention is greater than the comprehensive attention threshold, S2 is executed; S2. Obtain the answering time and correct rate of each test question in the process of completing the test question bank of the current chapter, and perform the answering judgment operation; in the answering judgment operation, if the answering time of the current test question is less than the answering time threshold, and the correct rate is not less than the correct rate threshold, the next test question will be displayed on the learning desktop; if the correct rate of the current test question is less than the correct rate threshold, the learning content corresponding to the current test question will be displayed on the learning desktop; after all the learning content corresponding to the current test question is displayed, the current test question and the consolidation questions with a content similarity greater than the similarity threshold to the current test question will be displayed on the learning desktop, and the answering judgment operation will be performed; the consolidation questions are in the consolidation question bank of the current chapter; if the answering time of the current test question is not less than the answering time threshold, but the correct rate is not less than the correct rate threshold, the consolidation questions with a content similarity greater than the similarity threshold to the current test question will be displayed on the learning desktop, and the answering judgment operation will be performed; after all the questions are answered, the absorption degree of the current chapter is obtained based on the total answering time and the comprehensive correct rate of the first answer of the test question bank of the current chapter; S3. Based on the content of the current chapter, the absorption level of the current chapter and the personal information of the students, a graph structure is constructed. The graph structure is processed by graph convolution to obtain comprehensive learning features. The comprehensive learning features and the personal characteristics of the students, as well as the characteristics of the learning method to be recommended, are processed by probability mapping to obtain a recommendation value. Based on the learning method to be recommended whose recommendation value is greater than the recommendation threshold, the form of the learning content of the next chapter is updated.

[0017] S1. During the process of students learning the content of the current chapter, the current stage attention is obtained based on the students' learning behavior at the current stage; it is determined whether the current stage attention exceeds the attention threshold; if the current stage attention is not greater than the first attention threshold, the current stage content is re-displayed on the learning desktop; if the current stage attention is between the first attention threshold and the second attention threshold, after the learning desktop displays a warning reminder, the next stage content is displayed on the learning desktop; if the current stage attention is not less than the second attention threshold, the next stage content is displayed on the learning desktop; after all the contents of the last stage of the current chapter are displayed, the learning attention of all stages and the corresponding stage content weights are weighted and summed to obtain the chapter comprehensive attention; if the chapter comprehensive attention is not greater than the comprehensive attention threshold, the learning content in the current chapter content whose stage content weight is greater than the content weight threshold is displayed on the learning desktop as the re-learned content, and after all the contents in the re-learned content are displayed, S2 is executed; if the chapter comprehensive attention is greater than the comprehensive attention threshold, S2 is executed.

[0018] During the learning stage, the student's attention at each stage is obtained in real time to predict the student's learning effect at that stage in real time. Based on the comparison between the stage attention and the attention threshold, it is decided whether to perform repeated learning operations to ensure the student's learning quality during the learning stage.

[0019] First, when students are learning the content of the current chapter, the current attention level is obtained based on the students' learning behavior at the current stage, so as to observe the students' learning status at all times.

[0020] The specific operation of obtaining the current learning attention is as follows: using the camera to obtain the student learning behavior video of the current stage, identifying the student head turning event, the student line of sight deviation event, the student expression abnormal event, and the student body posture abnormal event from the student learning behavior video, forming a set of abnormal learning behaviors in the current stage; the corresponding score of each event in the set of abnormal learning behaviors in the current stage, and the learning interface jump event score obtained based on the number of learning interface jumps and the jump interface type, after weighted summation processing, the difference between the total score and the score is the current learning attention.

[0021] The current learning attention can be obtained through the following formula: , Q For current learning attention.

[0022] Above q 1 The score corresponding to the student's head turning event, w 1 is the score weight of the head turning event, and the number of head turning times of students in the current stage is obtained, recorded as n 1, set the head turning threshold to N 1 ,when n 1 ≤N 1 hour, q 1 =0, when n 1 > N 1 hour, q 1 =(n 1 -N 1 )·a 1 , a 1 is the head turning number penalty coefficient. The head turning event score weight, head turning number threshold, and head turning number penalty coefficient can be set according to historical experimental data. The specific setting method is the existing technology and will not be repeated here to save space.

[0023] Above q 2 The corresponding score for the student's sight deviation event, w 2 is the score weight of the sight deviation event, and the number of students’ sight deviations in the current stage is obtained, recorded as n 2 , set the line of sight deviation threshold to N 2 , record the time of each line of sight deviation as l i , set the sight deviation time threshold to L ,when n 2 ≤N 2 and l i ≤L hour, q 2 =0, when n 2 > N 2 or l i > L hour, q 2 The formula for obtaining is as follows: , a 2 The deduction coefficient for the number of sight deviations, b 2is the sight deviation time penalty coefficient. The sight deviation event score weight, sight deviation number threshold, sight deviation time threshold, sight deviation number penalty coefficient, and sight deviation time penalty coefficient can be set according to historical experimental data.

[0024] Above q 3 The corresponding score for students’ abnormal facial expressions, w 3 is the score weight of the abnormal expression event, and the number of abnormal expressions in the current stage is obtained, recorded as n 3 , set the threshold of abnormal expression times to N 3 ,when n 3 ≤N 3 hour, q 3 =0, when n 3 > N 3 hour, q 3 =(n 3 - N 3 )·a 3 , a 3 It is the deduction coefficient for the number of abnormal facial expressions. Abnormal facial expression events (such as silly smiles) can be classified and identified through the facial expression recognition model in deep learning. The score weight of abnormal facial expression events, the threshold of the number of abnormal facial expressions, and the deduction coefficient for the number of abnormal facial expressions can be set according to historical experimental data.

[0025] Above q 4 The corresponding scores for students’ abnormal body posture events, w 4 is the score weight of abnormal body posture events, and the number of abnormal body postures in the current stage is obtained, recorded as n 4 , set the threshold of abnormal body posture times to N 4 ,when n 4 ≤N 4 hour, q 4 =0, when n 4 > N 4 hour, q 4=(n 4 -N 4 )·a 4 , a 4 It is the deduction coefficient for the number of abnormal body postures. Abnormal body posture events (bending, hunching, body tilting, etc.) can be identified by detecting the position changes of key points of the human body (such as shoulders, spine, etc.) in the video through a deep learning model, and analyzing the change pattern of body posture. The score weight of abnormal body posture events, the threshold of the number of abnormal body postures, and the deduction coefficient for the number of abnormal body postures can be set according to historical experimental data.

[0026] Above q 5 Score the learning interface jump event. w 5 is the score weight of the learning interface jump event, and the number of learning interface jumps in the current stage is obtained, recorded as n 5 , classify all the current stage learning interface jump events, and count the number of non-learning related interface jumps as n 5,1 , the number of learning related interface jumps is n 5,2 =n 5 -n 5,1 , q 5 =n 5,1 ·a 5 -n 5,2 ·c 5 , a 5 is the interface jump deduction coefficient, c 5 It is the interface jump bonus coefficient. The learning interface jump event score weight, interface jump deduction coefficient, and interface jump bonus coefficient can be set according to historical experimental data.

[0027] The sum of the above-mentioned head turning event score weight, gaze deviation event score weight, expression abnormality event score weight, body posture abnormality event score weight and learning interface jump event score weight is 1.

[0028] Then, it is determined whether the current stage attention exceeds the attention threshold.

[0029] If the current stage's attention is not greater than the first attention threshold, it means that the student's attention is very inattentive during the current stage of learning. To ensure the quality of learning, the current stage content is re-displayed on the learning desktop so that the student can re-learn the current stage content.

[0030] If the current stage of attention is between the first attention threshold and the second attention threshold, in order to remind students to focus on learning, after the warning reminder is displayed on the learning desktop (the duration of the warning reminder is the first time), the content of the next stage will be displayed on the learning desktop.

[0031] If the current stage's attention is not less than the second attention threshold, it means that the student is paying more attention in the current stage of learning. The content of the next stage is displayed on the learning desktop, and the student can directly learn the content of the next stage.

[0032] The first attention threshold is smaller than the second attention threshold.

[0033] Next, when all the contents of the last stage in the current chapter are displayed, the learning attention of all stages and the weight of the corresponding stage content are weighted and summed to obtain the comprehensive attention of the chapter, which is used to comprehensively evaluate the students' attention level when learning the current chapter content and indirectly predict the students' learning effect on the current chapter content.

[0034] If the comprehensive attention of a chapter is not greater than the comprehensive attention threshold, it indicates that the student's learning effect in the current chapter may be poor. Therefore, in order to ensure the student's learning effect, the learning content in the current chapter whose stage content weight is greater than the content weight threshold is displayed on the learning desktop as re-learning content; after all the content in the re-learning content is displayed, the exercise detection operation in S2 is executed.

[0035] If the chapter comprehensive attention is greater than the comprehensive attention threshold, it indicates that the student's learning effect in the current chapter may be good. In order to save time, the test question detection operation in S2 can be directly executed.

[0036] S2. Obtain the answering time and accuracy of each test question in the process of completing the test question bank of the current chapter, and perform the answer judgment operation; after all questions are answered, obtain the absorption degree of the current chapter based on the total answering time and the comprehensive accuracy rate of the first answer of the test question bank of the current chapter.

[0037] During the testing phase, the time and accuracy of each test question are obtained to evaluate students' mastery of the learning content in real time. According to the different levels of mastery of students, corresponding consolidation learning operations are selected to continuously strengthen the mastery of knowledge during the testing phase. In addition, after all questions are answered, the effect of students' initial learning of the current chapter content is obtained based on the total answering time and the comprehensive accuracy of the first answer to the test questions, which is used for subsequent analysis of students' personal learning characteristics, so as to find suitable learning methods to be recommended.

[0038] First, the answering time and accuracy rate of each test question in the process of completing the test question bank of the current chapter are obtained, and the answer judgment operation is performed to evaluate the learning effect of the students in learning the content corresponding to the current question.

[0039] During the answer judgment operation, analyze students' learning shortcomings according to different situations.

[0040] If the answering time of the current test question is less than the answering time threshold, and the accuracy is not less than the accuracy threshold, it means that the student answers the question quickly and accurately, and has mastered the content corresponding to the current test question. The next test question can be displayed on the learning desktop, and the student can answer the next test question. The accuracy threshold can be set accordingly according to the question type. For example, the accuracy threshold for multiple-choice questions can be set to 100%.

[0041] If the accuracy of the current test question is less than the accuracy threshold, it means that the student has not understood the content corresponding to the current test question. To ensure the learning effect, the learning content corresponding to the current test question is displayed on the learning desktop for students to review and understand. After all the learning content corresponding to the current test question is displayed, the current test question and the consolidation questions with a similarity greater than the similarity threshold are displayed on the learning desktop for students to answer. The current test question and the consolidation question both perform the answer judgment operation to ensure that students can understand and master the corresponding and related knowledge points of the current test question. Among them, the consolidation questions are in the consolidation question bank of the current chapter to enrich the content of students' answers.

[0042] If the answering time for the current test question is not less than the answering time threshold, but the accuracy rate is not less than the accuracy rate threshold, it means that the student is hesitant in answering the question and takes a long time to answer the question due to lack of knowledge points or poor memory. In order to ensure that the student has mastered the knowledge points proficiently, the consolidation questions with a content similarity with the current test question greater than the similarity threshold will be displayed on the learning desktop, and the answer judgment operation will be performed.

[0043] The method for obtaining the above content similarity is: perform word segmentation on the current test question and the consolidation question respectively, extract the keywords, count the number and arrange them, and obtain the current test question word set and the consolidation question word set; take the ratio of the number of elements in the intersection of the current test question word set and the consolidation question word set (elements correspond to keywords) to the number of elements in the union of the current test question word set and the consolidation question word set as the content similarity between the current test question and the consolidation question.

[0044] Then, after all the questions (test questions in the test question bank of the current chapter and consolidation questions in the consolidation question bank of the current chapter) are answered, the absorption level of the current chapter is obtained based on the total answering time and the comprehensive accuracy of the first answer of the test question bank of the current chapter (the average accuracy of all test questions when answered for the first time). This is used to obtain the answering effect of students' first answer to the test questions after ensuring that students have mastered the content of the current chapter, thereby indirectly obtaining information on students' initial learning effect.

[0045] Specifically, the absorption degree of the current chapter is obtained by the following formula: , A The absorption level of the current chapter. W 1 and W 2 They are the answering time weight and the first answer comprehensive correctness weight, S T To standardize the time to answer the questions, P The overall correct rate of the first answer.

[0046] Standardized time score S T It is obtained by the following formula: , T is the total time to answer the questions, T c Answer the question reference time, T max The maximum time limit for answering questions.

[0047] S3. Based on the content of the current chapter, the absorption level of the current chapter and the personal information of the students, a graph structure is constructed. The graph structure is processed by graph convolution to obtain comprehensive learning features. The comprehensive learning features and the personal characteristics of the students, as well as the characteristics of the learning method to be recommended, are processed by probability mapping to obtain a recommendation value. Based on the learning method to be recommended whose recommendation value is greater than the recommendation threshold, the form of the learning content of the next chapter is updated.

[0048] Based on the current chapter content, the current chapter absorption level and students' personal information, a graph structure is constructed that can intuitively reflect students' learning status of the chapter content. After graph convolution processing, the students' learning preference information in the process of learning the chapter content is extracted to obtain comprehensive learning features. The comprehensive learning features are probabilistically mapped with the students' personal characteristics and the characteristics of the learning methods to be recommended to obtain the recommended values. Finally, based on the recommended learning methods whose recommended values ​​are greater than the recommended threshold, the form of the learning content of the next chapter is updated to meet students' personalized learning needs and improve learning quality.

[0049] First, a graph structure is constructed based on the current chapter content, the current chapter absorption level, and the student's personal information. The nodes in the graph structure are the current chapter and the student, and the edges are the current chapter absorption level, to reflect the student's personal learning situation in the current chapter content. The construction of the graph structure includes but is not limited to deep learning tools such as PyTorch Geometric tools or OGB tools (Open Graph Benchmark).

[0050] Then, the graph structure is processed by graph convolution to extract the deep information of students when learning the current chapter content, and obtain the comprehensive learning features that can reflect the students' learning preferences. Graph convolution processing can be achieved through graph convolutional network (GCN).

[0051] Next, the comprehensive learning features, the student's personal features, and the features of the learning method to be recommended are processed by probability mapping (which can be achieved through the softmax function) to evaluate the adaptability of the learning method to the student and obtain the recommended value.

[0052] Among them, the student's personal characteristics are obtained based on the student's age, gender, grade, historical scores of corresponding subjects and hobby information; the characteristics of the learning method to be recommended are obtained by embedding the picture display learning method, or the audio-visual learning method characteristics, or the question-solving learning method.

[0053] Finally, based on the learning methods to be recommended whose recommendation values ​​are greater than the recommendation threshold, the form of the learning content in the next chapter is updated. Specifically: if the learning method to be recommended whose recommendation value is greater than the recommendation threshold is a picture display type learning method, then the picture display duration in the learning content of the next chapter is greater than the first display duration; if the learning method to be recommended whose recommendation value is greater than the recommendation threshold is an audio-visual type learning method, then the learning content of the next chapter is a video display duration greater than the first display duration; if the learning method to be recommended whose recommendation value is greater than the recommendation threshold is a question-solving type learning method, then the learning content of the next chapter is a question answering display, and the sum of the question display and answer display duration is greater than the first display duration.

[0054] When there are more than one learning methods to be recommended whose recommendation value is greater than the recommendation threshold, the learning method to be recommended corresponding to the maximum recommendation value is used to update the form of the learning content of the next chapter.

[0055] In addition, in order to meet the ever-changing needs of students and to steadily improve the quality of students' learning, after obtaining the absorption level of the next chapter, determine whether the difference between the absorption level of the next chapter and the absorption level of the current chapter is greater than the absorption level difference threshold; if not, it means that the student needs have not changed much. In order to save computing efficiency, the learning content of the next chapter will not be updated; if greater, the content of the next chapter and the absorption level of the next chapter will be added to the current graph structure to obtain an updated graph structure for performing graph convolution processing operations.

[0056] In the updated graph structure, the nodes are chapters and students, and the edges are chapter absorption degree and inter-chapter correlation degree. The inter-chapter correlation degree is to convert the contents of two chapters into vector form to obtain the first chapter vector and the second chapter vector; the cosine similarity between the first chapter vector and the second chapter vector is used as the inter-chapter correlation degree.

[0057] The cosine similarity can be obtained by the following formula: , R is the first chapter vector x = (x 1 ,x 2 ,x 3 ...,x j ) and the second chapter vector y = (y 1 ,y 2 ,y 3 ...,y j ) The cosine similarity of x j is the first chapter vector j Dimensional features, y j is the vector in the second chapter i Dimensional features, J is the total number of dimensions.

[0058] This embodiment also provides an intelligent adjustment system for student learning, which is used to implement the above-mentioned intelligent adjustment method for student learning, including: The learning stage adjustment module is used for obtaining the current stage attention degree based on the student learning behavior in the current stage during the process of students learning the current chapter content; judging whether the current stage attention degree exceeds the attention degree threshold; if the current stage attention degree is not greater than the first attention degree threshold, the current stage content is re-displayed on the learning desktop; if the current stage attention degree is between the first attention degree threshold and the second attention degree threshold, after the learning desktop displays a warning reminder, the next stage content is displayed on the learning desktop; if the current stage attention degree is not less than the second attention degree threshold, the next stage content is displayed on the learning desktop; the first attention threshold is less than the second attention threshold; after all the contents of the last stage of the current chapter are displayed, the learning attention degree of all stages is weighted and summed with the corresponding stage content weight to obtain the chapter comprehensive attention degree; if the chapter comprehensive attention degree is not greater than the comprehensive attention threshold, the learning content in the current chapter content whose stage content weight is greater than the content weight threshold is displayed on the learning desktop as the re-learned content, and after all the contents in the re-learned content are displayed, the test stage adjustment module is executed; if the chapter comprehensive attention degree is greater than the comprehensive attention threshold, the test stage adjustment module is executed; The test phase adjustment module is used to obtain the answering time and correct rate of each test question in the process of completing the test question bank of the current chapter, and perform the answering judgment operation; in the answering judgment operation, if the answering time of the current test question is less than the answering time threshold, and the correct rate is not less than the correct rate threshold, the next test question will be displayed on the learning desktop; if the correct rate of the current test question is less than the correct rate threshold, the learning content corresponding to the current test question will be displayed on the learning desktop; after all the learning content corresponding to the current test question is displayed, the current test question and the consolidation questions with a content similarity greater than the similarity threshold to the current test question will be displayed on the learning desktop, and the answering judgment operation will be performed; the consolidation questions are in the consolidation question bank of the current chapter; if the answering time of the current test question is not less than the answering time threshold, but the correct rate is not less than the correct rate threshold, the consolidation questions with a content similarity greater than the similarity threshold to the current test question will be displayed on the learning desktop, and the answering judgment operation will be performed; after all questions are answered, the absorption degree of the current chapter is obtained based on the total answering time and the comprehensive correct rate of the first answer of the test question bank of the current chapter; The learning content form update module is used to construct a graph structure based on the current chapter content, the current chapter absorption level and student personal information; the graph structure is processed by graph convolution to obtain comprehensive learning features; the comprehensive learning features and the student's personal features, as well as the features of the learning method to be recommended, are processed by probability mapping to obtain a recommendation value; based on the learning method to be recommended whose recommendation value is greater than the recommendation threshold, the learning content form of the next chapter is updated.

[0059] This embodiment also provides an intelligent adjustment device for student learning, including a processor and a memory, wherein the processor implements the above-mentioned intelligent adjustment method for student learning when executing a computer program stored in the memory.

[0060] This embodiment also provides a computer-readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the above-mentioned intelligent adjustment method for student learning.

[0061] This embodiment provides an intelligent adjustment method for student learning. In the learning stage, the student's attention at each stage is obtained in real time to predict the student's learning effect in real time, and according to the comparison result of the stage attention and the attention threshold, whether to perform repeated learning operations is selected to ensure the learning quality of students in the learning stage; in the test stage, the student's mastery of the learning content is evaluated in real time by obtaining the answering time and accuracy of each test question, and the corresponding consolidation learning operation is selected according to the different mastery of the students, so as to continuously strengthen the mastery of knowledge through learning; in addition, after all questions are answered, the total answering time is calculated based on the total answering time. The comprehensive accuracy rate of the first answer to the test questions is used to obtain the chapter absorption degree; then, based on the current chapter content, the current chapter absorption degree and the student's personal information, a graph structure that can intuitively reflect the student's learning of the chapter content is constructed, and after graph convolution processing, the student's learning preference information in the process of learning the chapter content is extracted to obtain the comprehensive learning features; the comprehensive learning features are probabilistically mapped with the student's personal characteristics and the characteristics of the learning method to be recommended to obtain the recommendation value; finally, based on the learning method to be recommended whose recommendation value is greater than the recommendation threshold, the form of the learning content of the next chapter is updated to meet the students' personalized learning needs, thereby improving the learning quality.

Claims

1. An intelligent adjustment method for student learning, characterized in that: The following operations are included: S1. During the process of students learning the content of the current chapter, based on the students' learning behavior at the current stage, the current stage attention is obtained; and it is determined whether the current stage attention exceeds the attention threshold; If the current stage attention is not greater than the first attention threshold, the current stage content is displayed again on the learning desktop; If the current stage of attention is between the first attention threshold and the second attention threshold, after displaying a warning reminder on the learning desktop, the next stage of content is displayed on the learning desktop; If the current stage's attention is not less than the second attention threshold, the next stage's content is displayed on the learning desktop; The first attention threshold is less than the second attention threshold; After all the contents of the last stage of the current chapter are displayed, the learning attention of all stages and the weight of the corresponding stage content are weighted and summed to obtain the comprehensive attention of the chapter; If the chapter comprehensive attention is not greater than the comprehensive attention threshold, the learning content in the current chapter content whose stage content weight is greater than the content weight threshold is displayed on the learning desktop as the re-learning content. After all the content in the re-learning content is displayed, execute S2; If the chapter comprehensive attention is greater than the comprehensive attention threshold, execute S2; S2, obtaining the answering time and correct rate of each test question in the process of the student completing the test question bank of the current chapter, and performing the answer judgment operation; In the answer judgment operation, if the answer time of the current test question is less than the answer time threshold, and the accuracy is not less than the accuracy threshold, the next test question will be displayed on the learning desktop; if the accuracy of the current test question is less than the accuracy threshold, the learning content corresponding to the current test question will be displayed on the learning desktop; After all learning contents corresponding to the current test question are displayed, the current test question and the consolidation questions whose content similarity with the current test question is greater than the similarity threshold are displayed on the learning desktop, and the answer judgment operation is performed; the consolidation questions are in the consolidation question bank of the current chapter; if the answering time of the current test question is not less than the answering time threshold, but the accuracy is not less than the accuracy threshold, the consolidation questions whose content similarity with the current test question is greater than the similarity threshold are displayed on the learning desktop, and the answer judgment operation is performed; After all questions are answered, the absorption level of the current chapter is obtained based on the total answering time and the comprehensive correct rate of the first answer of the current chapter test question bank; S3. Based on the content of the current chapter, the absorption level of the current chapter and the student’s personal information, a graph structure is constructed; the graph structure is processed by graph convolution to obtain comprehensive learning features; The comprehensive learning characteristics, the individual characteristics of the students, and the characteristics of the learning method to be recommended are processed through probability mapping to obtain the recommended value; Based on the recommended learning methods whose recommendation values ​​are greater than the recommendation threshold, the form of the learning content of the next chapter is updated.

2. The intelligent adjustment method for student learning according to claim 1, characterized in that: In S1, the operation of obtaining the current learning attention is specifically: Obtain the student learning behavior video of the current stage, and identify the student head turning events, student sight deviation events, student expression abnormal events, and student body posture abnormal events from the student learning behavior video to form a set of abnormal learning behaviors in the current stage; The difference between the corresponding score of each event in the current stage of abnormal learning behavior set and the learning interface jump event score obtained based on the number of learning interface jumps and the jump interface type and the total score after weighted summation is the current learning attention.

3. The intelligent adjustment method for student learning according to claim 1, characterized in that: In S2, the method for obtaining content similarity is: The current test question and the consolidation question are separately segmented to obtain the current test question word set and the consolidation question word set; the ratio of the number of intersection elements of the current test question word set and the consolidation question word set to the number of union elements of the current test question word set and the consolidation question word set is used as the content similarity between the current test question and the consolidation question.

4. The intelligent adjustment method for student learning according to claim 1, characterized in that: In S3, based on the recommended learning method whose recommendation value is greater than the recommendation threshold, the operation of updating the form of the learning content of the next chapter is specifically as follows: If the learning method to be recommended with a recommendation value greater than the recommendation threshold is a picture display learning method, then in the learning content format of the next chapter, the picture display duration is greater than the first display duration; If the learning method to be recommended with a recommendation value greater than the recommendation threshold is an audio-visual learning method, then in the learning content format of the next chapter, the video display duration is greater than the first display duration; If the recommended learning method with a recommendation value greater than the recommendation threshold is a question-solving learning method, then in the learning content format of the next chapter, the total duration of question display and answer display is greater than the first display duration.

5. The intelligent adjustment method for student learning according to claim 1, characterized in that: Also includes: After obtaining the absorption degree of the next chapter, determining whether the difference between the absorption degree of the next chapter and the absorption degree of the current chapter is greater than the absorption degree difference threshold; If not, the content format of the next chapter will not be updated; If it is greater, the content of the next chapter and the absorption degree of the next chapter are added to the current graph structure to obtain an updated graph structure for performing graph convolution processing operations.

6. The intelligent adjustment method for student learning according to claim 5, characterized in that: In the updated graph structure, the nodes are chapters and students, and the edges are the degree of chapter absorption and the degree of association between chapters; The degree of association between chapters is obtained by converting the contents of two chapters into vector form to obtain the first chapter vector and the second chapter vector; the cosine similarity between the first chapter vector and the second chapter vector is used as the degree of association between chapters.

7. The intelligent adjustment method for student learning according to claim 1, characterized in that: In S3, the student's personal characteristics are obtained based on the student's age, gender, grade and hobby information.

8. An intelligent adjustment system for student learning, used to implement the intelligent adjustment method for student learning according to claim 1, characterized in that: include: The learning stage adjustment module is used to obtain the current stage attention based on the student's learning behavior during the student's learning of the current chapter content; Determine whether the current stage of attention exceeds the attention threshold; If the current stage attention is not greater than the first attention threshold, the current stage content is displayed again on the learning desktop; If the current stage of attention is between the first attention threshold and the second attention threshold, after displaying a warning reminder on the learning desktop, the next stage of content is displayed on the learning desktop; If the current stage's attention is not less than the second attention threshold, the next stage's content is displayed on the learning desktop; The first attention threshold is less than the second attention threshold; after all contents of the last stage of the current chapter are displayed, the learning attention of all stages and the weight of the corresponding stage content are weighted and summed to obtain the comprehensive attention of the chapter; If the chapter comprehensive attention is not greater than the comprehensive attention threshold, the learning content in the current chapter content whose stage content weight is greater than the content weight threshold is displayed on the learning desktop as re-learning content. After all the content in the re-learning content is displayed, the test stage adjustment module is executed; If the chapter comprehensive attention is greater than the comprehensive attention threshold, the test phase adjustment module is executed; The test phase adjustment module is used to obtain the answering time and correct rate of each test question in the process of completing the test question bank of the current chapter, and perform the answer judgment operation; in the answer judgment operation, if the answering time of the current test question is less than the answering time threshold, and the correct rate is not less than the correct rate threshold, the next test question will be displayed on the learning desktop; if the correct rate of the current test question is less than the correct rate threshold, the learning content corresponding to the current test question will be displayed on the learning desktop; After all the learning contents corresponding to the current test question are displayed, the current test question and the consolidation questions whose content similarity with the current test question is greater than the similarity threshold are displayed on the learning desktop, and the answer judgment operation is performed; the consolidation questions are in the consolidation question bank of the current chapter; if the answering time of the current test question is not less than the answering time threshold, but the accuracy is not less than the accuracy threshold, the consolidation questions whose content similarity with the current test question is greater than the similarity threshold are displayed on the learning desktop, and the answer judgment operation is performed; after all the questions are answered, the absorption degree of the current chapter is obtained based on the total answering time and the comprehensive accuracy of the first answer of the test question bank of the current chapter; The learning content form update module is used to construct a graph structure based on the current chapter content, the current chapter absorption level and the student's personal information; the graph structure is processed by graph convolution to obtain comprehensive learning features; The comprehensive learning characteristics, the individual characteristics of the students, and the characteristics of the learning method to be recommended are processed through probability mapping to obtain the recommended value; Based on the recommended learning methods whose recommendation values ​​are greater than the recommendation threshold, the form of the learning content of the next chapter is updated.

9. An intelligent adjustment device for student learning, characterized in that: It comprises a processor and a memory, wherein when the processor executes the computer program stored in the memory, the intelligent adjustment method for student learning as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: Used to store a computer program, wherein when the computer program is executed by a processor, the intelligent adjustment method for student learning as described in any one of claims 1-7 is implemented.