High-order question chain generation method and system based on standardized knowledge graph construction

Abstract

The present invention discloses a high-order question chain generation method and system based on standardized knowledge graph construction, which relates to the technical field of question chain generation, and includes the following steps: dividing the chemistry learning process into several phased learning chapters according to the logical progression of knowledge points, each phased learning chapter covers representative knowledge themes or core concepts; in the learning process of each phased learning chapter, collecting the learning data of each student, and analyzing the collected learning data to understand the learning outcomes of the students. The present invention divides the learning chapters and uses machine learning models to predict the students' learning situation, divides the students into efficient, normal and inefficient settlements, evaluates the overall learning situation based on settlement analysis, and if a chapter has low learning efficiency, it is marked as an "abnormal chapter in the question chain", and the system reviews the students' performance, re-sorts and adjusts the question chain to ensure that the core knowledge points are fully covered, optimizes the learning path and improves the learning effect.

Description

Technical Field

[0001] The present invention relates to the technical field of question chain generation, and in particular to a high-order question chain generation method and system based on a standardized knowledge graph. Background Art

[0002] A high-order question chain generation system based on a standardized knowledge graph refers to the use of standardized and structured domain knowledge to generate a series of complex questions that are progressive in level by level through automatic or semi-automatic means. A knowledge graph is a graphical structure used to represent entities and their relationships, with entities as nodes and relationships as edges. The generation of high-order question chains relies on the associations between these nodes, combining natural language processing and logical reasoning technology to design a more in-depth and challenging sequence of questions. The generation process usually includes problem decomposition, logical hierarchy analysis, and context association, which helps users to conduct in-depth learning and exploration by gradually guiding from basic questions to more complex questions. For example, in the knowledge graph of chemical reactions, the primary question may be "What is the product of this chemical reaction", while the high-order question gradually extends to more complex content such as reaction conditions, energy changes, and catalyst mechanisms. Through standardized knowledge graphs, the generated question chains are ensured to be consistent, logical, and have high-quality reasoning capabilities.

[0003] The prior art has the following deficiencies:

[0004] In the process of learning chemistry, the high-order question generation system built based on the standardized knowledge graph may ignore key knowledge points. When this phenomenon occurs, it is usually difficult to detect in time, and the generated question chain will not cover all relevant knowledge points. If certain basic or core concepts (such as chemical bond energy, reaction mechanism, etc.) are omitted, students may find it difficult to accurately understand these key contents when they subsequently learn complex reactions or chemical phenomena. For example, ignoring the detailed mechanism of acid-base neutralization reactions may cause students to encounter difficulties in dealing with equilibrium constant calculations or reaction thermodynamics. This lack of concepts will gradually accumulate, eventually forming serious knowledge blind spots in exams and experiments, leading to incorrect reasoning or even completely wrong conclusions, seriously affecting academic performance, and hindering students' future academic progress and development.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0006] The purpose of the present invention is to provide a high-order question chain generation method and system based on a standardized knowledge graph. By dividing the learning process into phased learning chapters, the pre-trained machine learning model is used to predict and evaluate the students' learning situation, and the students are divided into efficient, normal and inefficient learner communities. According to the results of the community analysis, the overall learning situation of the students is comprehensively evaluated. When the overall learning efficiency of a chapter is low, it is marked as an "abnormal question chain chapter", indicating that the chapter may have key knowledge points missing or insufficient explanation. After discovering the problem, the system automatically reviews the student's performance, re-sorts the key knowledge points and adjusts the question chain to ensure that all core knowledge points are fully covered, thereby optimizing the learning path and improving the student's performance in subsequent learning, so as to solve the problems in the above-mentioned background technology.

[0007] In order to achieve the above object, the present invention provides the following technical solution: a method for generating a high-order question chain based on a standardized knowledge graph, comprising the following steps:

[0008] Divide the chemistry learning process into several phased learning chapters according to the logical progression of knowledge points. Each phased learning chapter covers representative knowledge topics or core concepts.

[0009] During the learning process of each stage of learning chapter, each student's learning data is collected, and the collected learning data is analyzed to understand the student's learning outcomes. After collecting the student's learning data, the pre-trained machine learning model is used to predict and evaluate each student's learning situation;

[0010] Based on the prediction results of the machine learning model, students are grouped according to their learning situation and divided into three clusters: efficient learner cluster, normal learner cluster, and inefficient learner cluster.

[0011] Conduct statistical analysis on individual students in each cluster, and make a comprehensive assessment of the overall learning situation of students based on the proportion of inefficient learner clusters in efficient learner clusters and normal learner clusters;

[0012] Based on the results of comprehensive assessment, the students' overall learning situation is divided into normal learning groups and inefficient learning groups, and the phased learning chapters corresponding to the inefficient learning groups are marked as "abnormal chapters in the question chain" to preliminarily identify the omission of key knowledge points when generating high-level questions;

[0013] Re-organize the key knowledge points of the problem chapter, automatically review the students' performance in this stage of learning, and adjust the question chain to ensure that all core knowledge points are fully covered and explained.

[0014] Preferably, the criterion for dividing the phased learning chapters in the chemistry learning process may be based on the logical structure of knowledge points, the difficulty level or the interdependence of knowledge points.

[0015] Preferably, in the learning process of each staged learning chapter, learning data of each student is collected, wherein the learning data includes the interruption frequency of the student's reasoning chain in the problem-solving process and the time when the student's understanding of a certain knowledge point does not progress in the learning process. After analyzing the interruption frequency of the student's reasoning chain in the problem-solving process and the time when the student's understanding of a certain knowledge point does not progress in the learning process, a reasoning chain break reference value and an understanding stagnation reference value are generated respectively. The frequency of the interruption of the reasoning chain due to the lack of necessary knowledge or reasoning steps in the problem-solving process of the student is quantified by the reasoning chain break reference value, and the stagnation time of the student's understanding progress on a certain knowledge point is quantified by the understanding stagnation reference value;

[0016] After obtaining the reasoning chain break reference value and understanding stagnation reference value generated by analyzing the learning data of each student, the reasoning chain break reference value and the understanding stagnation reference value are input into a pre-trained machine learning model to generate a learning efficiency prediction value, and the learning situation of each student is predicted based on the learning efficiency prediction value.

[0017] Preferably, the learning efficiency prediction value generated by predicting the learning situation of students in the learning process of each staged learning chapter is compared and analyzed with the pre-set learning efficiency reference range, and the students are grouped according to their learning situation. The specific division steps are as follows:

[0018] Students whose predicted learning efficiency values ​​are within the learning efficiency reference range are divided into normal learner clusters, students whose predicted learning efficiency values ​​are greater than the maximum value of the learning efficiency reference range are divided into efficient learner clusters, and students whose predicted learning efficiency values ​​are less than the minimum value of the learning efficiency reference range are divided into inefficient learner clusters.

[0019] Preferably, in the learning process of each staged learning chapter, the specific steps of generating a reference value of reasoning chain breakage after analyzing the interruption frequency of the reasoning chain in the problem-solving process of students are as follows:

[0020] Monitor each student's problem-solving process in real time, record the number of times the student's reasoning chain is interrupted for each question, and mark the interruption point each time the reasoning chain is interrupted ,in Indicates that the reasoning of student i on question j is interrupted. The interruption of the reasoning chain usually occurs when the student cannot smoothly transition to the next logical step or is stuck in thinking. The expression for the statistical number of interruptions of the reasoning chain is: , is the number of interruptions of student i during the jth question, Is a logical function of whether the kth reasoning step is completed. If the reasoning chain is interrupted, it is False, otherwise it is True. False is false, indicating a logical "error" or "invalidity", and True is true, indicating a logical "correctness" or "validity". represents the total number of reasoning steps required to answer the jth question, is an indicator function, which is recorded as 1 when the reasoning chain is interrupted and 0 otherwise;

[0021] After identifying the number of interruptions in the reasoning chain for each question, a weight is assigned to each interruption. The weight calculation expression is: , where is the weight of the interruption of the kth reasoning step of student i in the jth question, is the coefficient for adjusting weight growth, controlling the weight change range at different reasoning steps, k is the reasoning step where the interruption occurs, is the expected position of the reasoning step when the j-th question is interrupted, indicating the preset reasoning step position of the core part of the knowledge chain;

[0022] According to the number of interruptions and weights of the reasoning chain in each question, the reference value of the student's reasoning chain interruption in the entire chapter is calculated. The reference value of the reasoning chain interruption reflects the severity of the student's reasoning chain interruption in the entire learning process. It combines the frequency and weight of the interruption of each question and quantifies the incoherence of the student in the logical reasoning process. The calculation expression is: , where is the reference value of the reasoning chain break of student i, M represents the total number of questions in the staged learning chapter, is the complexity coefficient of the j-th question, which is used to adjust the impact of complex questions on the break of the reasoning chain.

[0023] Preferably, in the learning process of each staged learning chapter, the specific steps of generating a reference value of stagnation of understanding after analyzing the time when students do not make progress in understanding a certain knowledge point in the learning process are as follows:

[0024] For each student i, the time when his / her understanding of a certain knowledge point does not progress during the learning process is recorded, and the continuous stagnation time of the student when learning a specific knowledge point is defined as , where u represents the number of the knowledge point;

[0025] Introducing knowledge point complexity factor , used to adjust the impact of dead time on different knowledge points, is the complexity factor of the u-th knowledge point. The complexity factor of the knowledge point is set according to the level of the knowledge point in the knowledge graph and the difficulty level in the student learning path. The specific expression is: ,in, is the hierarchical depth of the u-th knowledge point, indicating the position of the knowledge point in the knowledge graph, is the criticality of the u-th knowledge point in the learning path, indicating its importance in subsequent learning. is the general mastery rate of the u-th knowledge point, which is used to measure students’ mastery of the knowledge point;

[0026] By combining the stagnation time and the complexity factor of the knowledge point of each student i, the reference value of understanding stagnation is calculated. The calculation expression is: ,in, represents the reference value of student i’s understanding stagnation, is the stagnation time of student i on knowledge point u, w represents the total number of knowledge points involved in the learning chapter of this stage, is the depth of student i’s understanding of knowledge point u.

[0027] Preferably, statistical analysis is performed on the individual students in each cluster, and the proportion of inefficient learner clusters in efficient learner clusters is marked as ,in, represents the number of students in the cluster of inefficient learners, represents the number of students in the efficient learner cluster, and the proportion of the inefficient learner cluster in the normal learner cluster is marked as ,in, represents the number of students in the normal learner cluster, and the proportion of the inefficient learner cluster in the efficient learner cluster is marked as The proportion of inefficient learner clusters in the normal learner clusters is marked as Perform comprehensive analysis to generate overall efficiency coefficients , through the overall efficiency coefficient Comprehensively evaluate the overall learning situation of students and the overall efficiency coefficient The calculation expression is: , where and are the proportion of inefficient learner clusters in efficient learner clusters The proportion of inefficient learner clusters in the normal learner clusters The weight of .

[0028] Preferably, the overall efficiency coefficient generated by comprehensive analysis of the overall learning situation of the students is compared with a preset overall efficiency coefficient reference threshold, and the overall learning situation of the students is divided into a normal learning group and an inefficient learning group. The specific division steps are as follows:

[0029] If the overall efficiency coefficient is greater than or equal to the overall efficiency coefficient reference threshold, the overall learning situation of the students is divided into an inefficient learning group, and the staged learning chapters corresponding to the inefficient learning group are marked as "abnormal chapters in the question chain" to preliminarily identify the omission of key knowledge points when generating high-level questions;

[0030] If the overall efficiency coefficient is less than the overall efficiency coefficient reference threshold, the students' overall learning situation will be divided into the normal learning group, and the staged learning chapters corresponding to the normal learning group will not be marked.

[0031] Preferably, the key knowledge points in the problem chapter are systematically sorted out to automatically review the students' learning performance in the staged learning chapter. The specific steps are:

[0032] First, the knowledge points used in generating the question chain for the phased learning chapter are compared in detail with the knowledge point data set in the entire chapter. Through data analysis, it is checked whether the knowledge points are fully covered in the question chain.

[0033] Next, identify the knowledge points that were ignored or not explained when generating the question chain, and find out the weak links;

[0034] Finally, make targeted adjustments to the question chain to ensure that all core knowledge points are fully covered and each key link is explained and guided.

[0035] A high-level question chain generation system based on a standardized knowledge graph, including a learning content division module, a learning data collection and prediction module, a student grouping module, a settlement statistics analysis module, a question chapter identification module, and a question chain adjustment module;

[0036] The learning content division module divides the chemistry learning process into several phased learning chapters according to the logical progression of knowledge points. Each phased learning chapter covers representative knowledge topics or core concepts.

[0037] The learning data collection and prediction module collects each student's learning data during the learning process of each stage of learning chapter, analyzes the collected learning data to understand the student's learning outcomes, and uses the pre-trained machine learning model to predict and evaluate each student's learning situation after collecting the student's learning data;

[0038] The student grouping module groups students according to their learning situations based on the prediction results of the machine learning model. The modules are divided into three clusters: efficient learner cluster, normal learner cluster, and inefficient learner cluster.

[0039] The cluster statistical analysis module conducts statistical analysis on individual students in each cluster and makes a comprehensive assessment of the overall learning situation of students based on the proportion of inefficient learner clusters in efficient learner clusters and normal learner clusters.

[0040] The problem chapter identification module divides students' overall learning status into normal learning groups and inefficient learning groups based on the results of comprehensive evaluation, and marks the phased learning chapters corresponding to the inefficient learning groups as "abnormal chapters in the problem chain", preliminarily identifying the omission of key knowledge points when generating high-level questions;

[0041] The question chain adjustment module re-sorts the key knowledge points of the question chapter, automatically reviews students' performance in the stage-by-stage learning chapter, and adjusts the question chain to ensure that all core knowledge points are fully covered and explained.

[0042] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0043] The present invention divides the learning process into phased learning chapters, predicts and evaluates the learning situation of students for the phased learning chapters, then divides the students into different settlements according to the learning effect, and then comprehensively evaluates the overall learning situation of students according to all settlements. When the overall learning efficiency corresponding to the phased learning chapter is low, the phased learning chapter corresponding to the inefficient learning group is marked as an "abnormal problem chain chapter", indicating that the overall learning efficiency of students in the phased learning chapter is very low, which is likely due to the fact that the key link knowledge points of the problem chain are ignored or not explained during the learning process. When the problem is found, the key knowledge points of the problem chapter are re-sorted, the performance of students in the phased learning chapter is automatically reviewed, and the problem chain is adjusted to ensure that all core knowledge points are fully covered and explained, and to ensure that students fully understand and master all core knowledge points in the learning process, avoid knowledge blind spots and low learning efficiency caused by omission or non-explanation of key knowledge points, and timely discover and correct the key links in the problem chain and optimize the learning path, so as to effectively improve the performance of students in subsequent learning and ensure that students can smoothly cope with more complex knowledge and problems. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0045] Figure 1 A method flow chart of a high-order question chain generation method based on a standardized knowledge graph constructed in the present invention;

[0046] Figure 2 This is a module schematic diagram of a high-order question chain generation system constructed based on a standardized knowledge graph in the present invention. DETAILED DESCRIPTION

[0047] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0048] The present invention provides Figure 1 A high-order question chain generation method based on a standardized knowledge graph is shown, comprising the following steps:

[0049] Divide the chemistry learning process into several phased learning chapters according to the logical progression of knowledge points. Each phased learning chapter covers representative knowledge topics or core concepts.

[0050] Representative knowledge topics or core concepts, such as acid-base reactions, redox reactions, reaction thermodynamics, etc. This division method can not only help students gradually master knowledge, but also make subsequent evaluations more refined. Through phased learning, the learning effect of students at different stages can be tested according to the progress of each chapter, so as to avoid missing important knowledge points. After each stage, the system can summarize and evaluate the stage, which helps to find problems and provide feedback in time.

[0051] The standard for dividing the learning chapters in the chemistry learning process can be based on the logical structure of knowledge points, the difficulty level or the interdependence of knowledge points; first, the logical structure of knowledge points means that learning should be carried out from simple to complex, first mastering the basic concepts, and then gradually transitioning to complex theories and applications. Secondly, the difficulty level means that the learning content should be progressively advanced according to the degree of difficulty, ensuring that students can challenge more difficult content after being familiar with the basic knowledge. Finally, the interdependence of knowledge points means that a knowledge point is often the basis for subsequent knowledge, so more in-depth or complex knowledge should be introduced after students understand the previous concepts.

[0052] During the learning process of each stage of learning chapter, each student's learning data is collected, and the collected learning data is analyzed to understand the student's learning outcomes. After collecting the student's learning data, the pre-trained machine learning model is used to predict and evaluate each student's learning situation;

[0053] During the learning process of each stage of learning chapter, the learning data of each student is collected, wherein the learning data includes the frequency of interruption of the student's reasoning chain in the process of solving problems and the time when the student's understanding of a certain knowledge point does not progress in the learning process. After analyzing the frequency of interruption of the student's reasoning chain in the process of solving problems and the time when the student's understanding of a certain knowledge point does not progress in the learning process, a reasoning chain break reference value and an understanding stagnation reference value are generated respectively. The reasoning chain break reference value is used to quantify the frequency of interruption of the reasoning chain due to lack of necessary knowledge or reasoning steps in the process of solving problems, reflecting the logical coherence problems in the student's reasoning process. The understanding stagnation reference value is used to quantify the stagnation time of the student's understanding progress on a certain knowledge point, reflecting the learning bottleneck or jamming encountered by the student on a specific knowledge point.

[0054] After obtaining the reasoning chain break reference value and understanding stagnation reference value generated by analyzing the learning data of each student, the reasoning chain break reference value and the understanding stagnation reference value are input into a pre-trained machine learning model to generate a learning efficiency prediction value, and the learning situation of each student is predicted based on the learning efficiency prediction value.

[0055] A pre-trained machine learning model refers to a model that has been trained and optimized with a large amount of historical data before it is put into use. Specifically, the model is repeatedly trained and adjusted using the learning data of previous students, such as the reference value of reasoning chain breakage and the reference value of understanding stagnation, as well as the corresponding learning effects as output. Through this process, the model can learn from the data the relationship between different input features and learning effects, and optimize its internal parameters so that it can accurately predict learning efficiency when faced with new student data. The training process usually includes steps such as data preprocessing, model selection, parameter tuning, and verification to ensure that the model has good generalization capabilities and can adapt to the learning performance of different students, thereby generating personalized learning efficiency prediction values ​​for each student, helping to promptly identify learning problems and optimize teaching strategies.

[0056] The high frequency of interruptions in the reasoning chain during the problem-solving process usually indicates that the students' learning efficiency is poor. The interruption of the reasoning chain means that the students have encountered logical obstacles many times when solving problems and are unable to smoothly connect the various knowledge points to complete the deduction. This may be due to insufficient grasp of basic knowledge points, unclear problem-solving ideas, or the key knowledge points in the problem chain have been ignored. If the key knowledge points in the problem chain are omitted or not explained, students may lack the necessary knowledge support. Therefore, the high frequency of interruptions in the reasoning chain is often one of the signs of unsatisfactory learning effects and low efficiency of students.

[0057] In the process of learning each stage of learning chapter, the specific steps of analyzing the interruption frequency of students' reasoning chain in the problem-solving process and generating the reference value of reasoning chain break are as follows:

[0058] Monitor each student's problem-solving process in real time, record the number of times the student's reasoning chain is interrupted for each question, and mark the interruption point each time the reasoning chain is interrupted ,in Indicates that student i's reasoning on question j is interrupted. Reasoning chain interruptions usually occur when students cannot smoothly transition to the next logical step or get stuck in thinking. For example, a student needs 5 reasoning steps to solve the problem, but is interrupted at the 3rd step, which is recorded as an interruption point. The expression for the number of interruptions in the reasoning chain is: , is the number of interruptions of student i during the jth question, Is a logical function of whether the kth reasoning step is completed. If the reasoning chain is interrupted, it is False, otherwise it is True. False is false, indicating a logical "error" or "invalidity", and True is true, indicating a logical "correctness" or "validity". represents the total number of reasoning steps required to answer the jth question, is an indicator function, which is recorded as 1 when the reasoning chain is interrupted and 0 otherwise;

[0059] After identifying the number of interruptions in the reasoning chain for each question, a weight is assigned to each interruption. If the interruption occurs early in the reasoning chain or involves key steps, the weight is higher because these interruptions have a greater impact on the problem-solving process. The weight is calculated based on the importance of the reasoning steps. For example, when the interruption occurs at a key knowledge point or the core of the problem chain, the weight is larger. The calculation expression of the weight is: , where is the weight of the interruption of the kth reasoning step of student i in the jth question, is the coefficient for adjusting weight growth, controlling the weight change range at different reasoning steps, k is the reasoning step where the interruption occurs, is the expected position of the reasoning step when the jth question is interrupted, indicating the preset reasoning step position of the core part of the knowledge chain (determined by the knowledge graph), the index part The role of is to control the weight of the interruption of the reasoning step With reasoning step k relative to the expected position Specifically, the exponential function determines the distance of the interrupted reasoning step from the expected position. The farther the distance is, the smaller the change in weight is. When the reasoning step k is close to the core step When the exponential part approaches 0, the weight The index value will be close to 1, indicating that the interruption has a greater impact on the problem solving. On the contrary, if the reasoning step is far away from the core step, the index value will increase and the weight value will approach 0, indicating that the interruption of this step has a smaller impact on the problem solving. Therefore, the index part effectively adjusts the importance of interruptions at different steps, so that the reasoning steps of the core knowledge points get a greater weight.

[0060] Determine the first (The expected position of the reasoning steps when the j-th question is interrupted) is obtained by analyzing the core knowledge points involved in the question and the knowledge structure related to it. In the knowledge graph, knowledge points exist in the form of nodes, and the connections between nodes represent the dependencies between knowledge points. By analyzing the relationship between the knowledge points tested in the question and other related knowledge points, the key steps in the reasoning chain can be determined. For example, a question may involve knowledge of chemical reaction equilibrium. The knowledge graph will determine the core position of these knowledge points in the reasoning chain based on the correlation between knowledge points such as reaction mechanism and chemical bond energy. That is, the key reasoning steps that students must master when solving the problem are located. The weight and importance of this step to solving the problem are inferred through the dependency relationship of the knowledge graph, and then it is set as the expected position.

[0061] According to the number of interruptions and weights of the reasoning chain in each question, the reference value of the student's reasoning chain interruption in the entire chapter is calculated. The reference value of the reasoning chain interruption reflects the severity of the student's reasoning chain interruption in the entire learning process. It combines the frequency and weight of the interruption of each question and quantifies the incoherence of the student in the logical reasoning process. The calculation expression is: , where is the reference value of the reasoning chain break of student i, M represents the total number of questions in the staged learning chapter, is the complexity coefficient of the jth question, which is used to adjust the impact of complex questions on the break of the reasoning chain;

[0062] Complexity coefficient It can be obtained through the knowledge graph. The knowledge graph shows each knowledge point and its interrelationship in a structured manner. The complexity coefficient can be determined based on the number of knowledge points involved in the question and the dependencies between the knowledge points. If a question involves multiple knowledge points and the dependencies between these knowledge points are complex (for example, multiple levels of logical deduction or cross-disciplinary associations), the complexity of the question is high and the complexity coefficient is high. Through the knowledge graph, we can analyze the knowledge points involved in the question and their positions in the entire graph, so as to more accurately evaluate the complexity of the question.

[0063] In the learning process of each staged learning chapter, the larger the performance value of the reference value of the reasoning chain break generated by analyzing the frequency of interruption of the reasoning chain in the problem-solving process of students, the lower the learning efficiency of students is usually. The reference value of the reasoning chain break reflects the frequency and severity of the interruption of the reasoning chain in the problem-solving process of students. When the reference value of the reasoning chain break is high, it means that students frequently encounter obstacles in the derivation process and cannot carry out logical reasoning smoothly. This is usually caused by insufficient understanding of relevant knowledge points or poor grasp of key knowledge points (the reason for insufficient understanding of relevant knowledge points may be that the knowledge points of the key links in the problem chain are ignored and not mentioned). Frequent interruptions of the reasoning chain not only prolong the problem-solving time, but also increase the thinking burden of students and reduce learning efficiency. Therefore, the larger the reference value of the reasoning chain break, the lower the learning efficiency of students is often, indicating that they have major learning problems in this learning chapter.

[0064] If students do not make progress in understanding a certain knowledge point for a long time during the learning process, it usually indicates that the students have encountered obstacles in this knowledge point, reflecting poor learning efficiency. This may be caused by multiple reasons. First, students do not have a good grasp of the relevant basic knowledge points and cannot correctly apply these basic concepts to solve more complex problems. Secondly, unclear problem-solving ideas may also cause students to be unable to successfully complete problem reasoning and encounter thinking blocks. Finally, if the key knowledge points in the problem chain are omitted or not explained, students may lack the necessary knowledge support, resulting in stagnation of understanding. This situation not only prolongs the students' problem-solving time, but may also lead to a chain reaction in the mastery of subsequent knowledge, further affecting the learning effect.

[0065] In the process of learning each stage of learning chapter, the specific steps for analyzing the time when students' understanding of a certain knowledge point does not progress during the learning process and generating a reference value for understanding stagnation are as follows:

[0066] For each student i, the time when his / her understanding of a certain knowledge point does not progress during the learning process is recorded, and the continuous stagnation time of the student when learning a specific knowledge point is defined as , where u represents the number of the knowledge point;

[0067] For example, if a student spends too much time on a topic but fails to complete the correct answer or understanding, their dead time This parameter is used to quantify the time students spend stuck on a specific knowledge point.

[0068] Different knowledge points have different levels of complexity, so it is necessary to introduce a knowledge point complexity factor , used to adjust the impact of dead time on different knowledge points, is the complexity factor of the u-th knowledge point. For example, a knowledge point with higher complexity may allow for a longer stagnation time. The complexity factor of the knowledge point is set according to the level of the knowledge point in the knowledge graph and the difficulty level in the student's learning path. The specific expression is: ,in, is the hierarchical depth of the u-th knowledge point, indicating the position of the knowledge point in the knowledge graph. The higher the level, the more complex it is. is the criticality of the u-th knowledge point in the learning path, indicating its importance in subsequent learning. is the general mastery rate of the u-th knowledge point, which is used to measure students’ mastery of the knowledge point. The lower it is, the more difficult the knowledge point is;

[0069] The depth of knowledge points, the importance of knowledge points in the learning path, and the general mastery rate of knowledge points can all be determined through knowledge graphs. The following is a specific determination method:

[0070] 1. The depth of knowledge points:

[0071] The hierarchical depth in the knowledge graph represents the position of a certain knowledge point in the entire knowledge system. Knowledge graphs are usually presented in the form of nodes and edges, where nodes represent knowledge points and edges represent the relationship between them. More basic knowledge points are usually located at the bottom of the knowledge graph, while advanced concepts that rely on these basic knowledge points are located at higher levels of the graph. The hierarchical depth can be determined by calculating the distance from the knowledge point to the root node of the graph. For example, the concept of acid-base reaction may be at the basic level, while the calculation of acid-base equilibrium constant is at a higher level because the latter depends on the former. Therefore, the hierarchical depth of a knowledge point is an important indicator that reflects the complexity and prerequisites of the knowledge point.

[0072] 2. The importance of knowledge points in the learning path:

[0073] The criticality of a knowledge point reflects its importance in the learning path, indicating the influence of the knowledge point on subsequent learning or related knowledge points. The criticality of a knowledge point in the learning path can be evaluated through the out-degree (the number of other nodes pointed to by a node) in the knowledge graph. For example, if a knowledge point has many associated subsequent knowledge points, it indicates that it is a core node in the learning path and students must master it in order to smoothly understand other concepts. By analyzing the out-degree and association of each knowledge point in the knowledge graph, it can be determined which knowledge points are key knowledge in subsequent learning.

[0074] 3. General mastery rate of knowledge points:

[0075] The general mastery rate of a knowledge point can be determined by analyzing the student answer data of the knowledge point in the knowledge graph. The specific method is: for each knowledge point u, the system counts the correct answer rate of all students related to the knowledge point in the question chain. Assume that a total of N students have been exposed to the knowledge point u, where If students correctly answer questions related to this knowledge point, then the general mastery rate of this knowledge point is It can be calculated by the formula: , this ratio value reflects the overall mastery of the knowledge point by the students. In this way, the system can directly determine the difficulty of mastering each knowledge point based on the students' answering performance, thereby guiding teaching adjustments.

[0076] By combining the stagnation time and the complexity factor of the knowledge point of each student i, the reference value of understanding stagnation is calculated. The calculation expression is: ,in, represents the reference value of student i’s understanding stagnation, is the stagnation time of student i on knowledge point u, w represents the total number of knowledge points involved in the learning chapter of this stage, is the depth of student i’s understanding of knowledge point u;

[0077] The depth of understanding of a knowledge point can be determined by the frequency and quality of students’ interactions with the knowledge point in the knowledge graph. The specific method is: for each knowledge point u, the number of exercises, study time, and correct answer rate of students on the knowledge point are counted. Assuming that student i completes The average correct rate was , and the learning time spent is , the student's understanding depth of this knowledge point can be determined by the following formula , the calculation expression is: ; This formula takes into account the number of times students practice, accuracy and time, and reflects the students' understanding of the knowledge point. The higher the depth of understanding, the more comprehensive and in-depth the students' mastery of the knowledge point.

[0078] In the learning process of each staged learning chapter, the performance value of the understanding stagnation reference value generated by analyzing the time when students do not make progress in understanding a certain knowledge point during the learning process is larger, indicating that the student's learning efficiency is lower, which is usually caused by insufficient understanding of the relevant knowledge points or weak grasp of key knowledge points. When students spend too much time on a certain knowledge point but fail to make progress, it indicates that they have problems understanding this knowledge point. This may be because students do not have enough basic knowledge to support their current learning, or some key knowledge points in the problem chain are ignored and not explained, resulting in students being unable to smoothly understand the subsequent content. The omission or weak links of such knowledge points will directly affect students' learning effects, which will ultimately be reflected in a higher understanding stagnation reference value.

[0079] The machine learning model is not specifically limited here, and can achieve the reference value of breaking the reasoning chain of student i and the reference value of student i's understanding stagnation Perform comprehensive analysis to generate learning efficiency predictions All machine learning models are available, among which, represents the predicted value of learning efficiency of student i. In order to realize the technical solution of the present invention, the present invention provides a specific implementation method. The calculation formula for generating the predicted value of learning efficiency of student i is: , where , are the reference values ​​of reasoning chain break of student i and the reference value of student i's understanding stagnation The preset scaling factor of , Both are greater than 0.

[0080] It can be seen from the learning efficiency prediction value that in the learning process of each stage learning chapter, the greater the performance value of the reasoning chain break reference value generated after analyzing the frequency of interruption of the student's reasoning chain in the problem-solving process, the greater the performance value of the understanding stagnation reference value generated after analyzing the time when the student's understanding of a certain knowledge point does not progress in the learning process. In other words, the greater the learning efficiency prediction value generated by predicting the student's learning situation in the learning process of each stage learning chapter, the lower the student's learning efficiency, and vice versa.

[0081] Based on the prediction results of the machine learning model, students are grouped according to their learning situation and divided into three clusters: efficient learner cluster, normal learner cluster, and inefficient learner cluster.

[0082] The learning efficiency prediction value generated by predicting the students' learning situation in the learning process of each staged learning chapter is compared and analyzed with the pre-set learning efficiency reference range, and the students are grouped according to their learning situation. The specific division steps are as follows:

[0083] Students whose predicted learning efficiency values ​​are within the learning efficiency reference range are divided into normal learner clusters, students whose predicted learning efficiency values ​​are greater than the maximum value of the learning efficiency reference range are divided into efficient learner clusters, and students whose predicted learning efficiency values ​​are less than the minimum value of the learning efficiency reference range are divided into inefficient learner clusters.

[0084] Efficient learners tend to perform well in all tests and have a deep understanding of the knowledge points; normal learners perform averagely; while inefficient learners may have problems such as a weak grasp of the knowledge points and lagging behind in learning progress.

[0085] The learning efficiency reference range is a standardized numerical range used to measure and evaluate students' learning efficiency, which is determined based on historical data or predefined learning performance standards. This range defines the upper and lower limits of normal learning efficiency, and is usually a reasonable range obtained by analyzing the predicted values ​​of learning efficiency of a large number of students. Students within this range are considered to have normal learning efficiency, neither too high nor too low. Students who exceed the upper limit of this range are considered to be efficient learners, indicating that their learning progress and knowledge mastery are faster; students below the lower limit are classified as inefficient learners, indicating that they may have encountered difficulties or stagnation in the learning process. Therefore, the learning efficiency reference range provides a basis for distinguishing different learning efficiency groups.

[0086] Conduct statistical analysis on individual students in each cluster, and make a comprehensive assessment of the overall learning situation of students based on the proportion of inefficient learner clusters in efficient learner clusters and normal learner clusters;

[0087] Statistical analysis was performed on the individual students in each cluster, and the proportion of inefficient learner clusters in efficient learner clusters was marked as ,in, represents the number of students in the cluster of inefficient learners, represents the number of students in the efficient learner cluster, and the proportion of the inefficient learner cluster in the normal learner cluster is marked as ,in, represents the number of students in the normal learner cluster, and the proportion of the inefficient learner cluster in the efficient learner cluster is marked as The proportion of inefficient learner clusters in the normal learner clusters is marked as Perform comprehensive analysis to generate overall efficiency coefficients , through the overall efficiency coefficient Comprehensively evaluate the overall learning situation of students and the overall efficiency coefficient The calculation expression is: , where and are the proportion of inefficient learner clusters in efficient learner clusters The proportion of inefficient learner clusters in the normal learner clusters The weight of .

[0088] It can be seen from the overall efficiency coefficient that the larger the performance value of the overall efficiency coefficient generated by the comprehensive evaluation of the overall learning situation of students, the higher the proportion of inefficient learners in the efficient learner cluster and the normal learner cluster, suggesting that the overall learning progress of students is lagging behind. This situation is likely due to the fact that the key knowledge points in the problem chain are ignored or not explained during the learning process, causing students to encounter obstacles in high-order problems, which in turn affects their understanding and mastery of the entire knowledge system. Specifically, when the important basic knowledge points in the knowledge chain are not fully covered or explained, students will not have a solid grasp of these knowledge points, resulting in a jam in the learning progress, especially when dealing with more complex reasoning and comprehensive problems, which is manifested as a significant decrease in learning efficiency. In addition, the omission of key knowledge points may also lead to knowledge gaps in students' subsequent learning process, affecting the absorption and application of new knowledge, and ultimately making it difficult for students to establish an effective reasoning chain when dealing with comprehensive problems.

[0089] Based on the results of comprehensive assessment, the students' overall learning situation is divided into normal learning groups and inefficient learning groups, and the phased learning chapters corresponding to the inefficient learning groups are marked as "abnormal chapters in the question chain" to preliminarily identify the omission of key knowledge points when generating high-level questions;

[0090] The overall efficiency coefficient generated by comprehensive analysis of the students' overall learning situation is compared with the pre-set overall efficiency coefficient reference threshold, and the students' overall learning situation is divided into normal learning groups and inefficient learning groups. The specific division steps are as follows:

[0091] If the overall efficiency coefficient is greater than or equal to the overall efficiency coefficient reference threshold, the overall learning situation of the students is divided into an inefficient learning group, and the staged learning chapters corresponding to the inefficient learning group are marked as "abnormal chapters in the question chain" to preliminarily identify the omission of key knowledge points when generating high-level questions;

[0092] If the overall efficiency coefficient is less than the overall efficiency coefficient reference threshold, the students' overall learning situation will be divided into the normal learning group, and the staged learning chapters corresponding to the normal learning group will not be marked.

[0093] If the evaluation shows that the overall learning situation of students is an inefficient learning group, that is, the overall learning situation is poor, for example, many students perform poorly on a certain knowledge point, the system can mark the staged learning chapter as an "abnormal problem chain chapter", which means that there may be missing knowledge points or understanding deviations in the learning chapter. Timely judgment of the overall learning situation of students helps to discover potential problems in learning in advance and avoid cumulative errors in knowledge points in subsequent learning.

[0094] Re-organize the key knowledge points of the problem chapter, automatically review the students' performance in the chapter at that stage, and adjust the question chain to ensure that all core knowledge points are fully covered and explained;

[0095] To ensure that students have a comprehensive grasp of all core knowledge points, it is necessary to systematically sort out the key knowledge points in the problem chapters, automatically review students' learning performance in this phase of learning chapters, and pay special attention to those knowledge points that are prone to errors. The specific steps are:

[0096] First, the knowledge points used in generating the question chain for this phased learning chapter are compared in detail with the knowledge point data set in the entire chapter. Through data analysis, check whether the knowledge points are fully covered in the question chain. Next, identify the knowledge points that may be ignored or not explained when generating the question chain, and find out the weak links. Finally, based on these identification results, make targeted adjustments to the question chain to ensure that all core knowledge points are fully covered and each key link is explained and guided, so as to optimize students' learning paths, avoid knowledge blind spots, and improve overall learning effects.

[0097] The present invention divides the learning process into phased learning chapters, inputs the students' learning data into a pre-trained machine learning model for the phased learning chapters, predicts and evaluates their learning situation, and then divides the students into efficient learner clusters, normal learner clusters and inefficient learner clusters according to the learning effects, performs statistical analysis on these clusters, and comprehensively evaluates the students' overall learning situation based on the proportion of inefficient learner clusters in efficient learner clusters and normal learner clusters. When the overall learning efficiency corresponding to the phased learning chapters is low, the phased learning chapters corresponding to the inefficient learning group are marked as "abnormal problem chain chapters", indicating that the overall learning efficiency of students in this phased learning chapter is very low. If the score is low, it is likely that the key knowledge points in the problem chain are ignored or not explained during the learning process. When this problem is found, the key knowledge points of the problem chapter will be re-sorted, and the students' performance in this stage of learning will be automatically reviewed. The problem chain will be adjusted to ensure that all core knowledge points are fully covered and explained, and to ensure that students have a comprehensive understanding and mastery of all core knowledge points in the learning process, avoiding knowledge blind spots and low learning efficiency caused by omissions or failure to explain key knowledge points. By timely discovering and correcting the key links in the problem chain and optimizing the learning path, it can effectively improve students' performance in subsequent learning and ensure that students can smoothly cope with more complex knowledge and problems.

[0098] The present invention provides Figure 2 A high-order question chain generation system based on a standardized knowledge graph is shown, comprising a learning content division module, a learning data collection and prediction module, a student grouping module, a settlement statistics analysis module, a question chapter identification module, and a question chain adjustment module;

[0099] The learning content division module divides the chemistry learning process into several phased learning chapters according to the logical progression of knowledge points. Each phased learning chapter covers representative knowledge topics or core concepts.

[0100] The learning data collection and prediction module collects each student's learning data during the learning process of each stage of learning chapter, analyzes the collected learning data to understand the student's learning outcomes, and uses the pre-trained machine learning model to predict and evaluate each student's learning situation after collecting the student's learning data;

[0101] The student grouping module groups students according to their learning situations based on the prediction results of the machine learning model. The modules are divided into three clusters: efficient learner cluster, normal learner cluster, and inefficient learner cluster.

[0102] The cluster statistical analysis module conducts statistical analysis on individual students in each cluster and makes a comprehensive assessment of the overall learning situation of students based on the proportion of inefficient learner clusters in efficient learner clusters and normal learner clusters.

[0103] The problem chapter identification module divides students' overall learning status into normal learning groups and inefficient learning groups based on the results of comprehensive evaluation, and marks the phased learning chapters corresponding to the inefficient learning groups as "abnormal chapters in the problem chain", preliminarily identifying the omission of key knowledge points when generating high-level questions;

[0104] The question chain adjustment module re-sorts the key knowledge points of the question chapter, automatically reviews students' performance in the chapter at that stage, and adjusts the question chain to ensure that all core knowledge points are fully covered and explained;

[0105] An embodiment of the present invention provides a method for generating a high-order question chain based on a standardized knowledge graph, which is realized by the above-mentioned high-order question chain generation system based on a standardized knowledge graph. The specific method and process of a high-order question chain generation system based on a standardized knowledge graph are detailed in the embodiment of the above-mentioned method for generating a high-order question chain based on a standardized knowledge graph, which will not be repeated here.

[0106] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0107] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0108] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A high-order question chain generation method based on standardized knowledge graph, characterized in that: The following steps are involved: Divide the chemistry learning process into several phased learning chapters according to the logical progression of knowledge points. Each phased learning chapter covers representative knowledge topics or core concepts. During the learning process of each stage of learning chapter, each student's learning data is collected, and the collected learning data is analyzed to understand the student's learning outcomes. After collecting the student's learning data, the pre-trained machine learning model is used to predict and evaluate each student's learning situation; During the learning process of each staged learning chapter, the learning data of each student is collected, wherein the learning data includes the interruption frequency of the student's reasoning chain in the problem-solving process and the time when the student's understanding of a certain knowledge point does not progress in the learning process. After analyzing the interruption frequency of the student's reasoning chain in the problem-solving process and the time when the student's understanding of a certain knowledge point does not progress in the learning process, a reference value for the interruption of the reasoning chain and a reference value for the stagnation of understanding are generated respectively. The reference value for the interruption of the reasoning chain is used to quantify the frequency of the interruption of the student's reasoning chain in the problem-solving process, and the reference value for the stagnation of understanding is used to quantify the stagnation time of the student's understanding progress on a certain knowledge point; After obtaining the inference chain break reference value and the understanding stagnation reference value generated by analyzing the learning data of each student, the inference chain break reference value and the understanding stagnation reference value are input into the pre-trained machine learning model to generate a learning efficiency prediction value, and the learning situation of each student is predicted by the learning efficiency prediction value; Based on the prediction results of the machine learning model, students are grouped according to their learning situation and divided into three clusters: efficient learner cluster, normal learner cluster, and inefficient learner cluster. Conduct statistical analysis on individual students in each cluster, and make a comprehensive assessment of the overall learning situation of students based on the proportion of inefficient learner clusters in efficient learner clusters and normal learner clusters; Based on the results of comprehensive evaluation, the students' overall learning situation is divided into normal learning group and inefficient learning group, and the phased learning chapters corresponding to the inefficient learning group are marked as "abnormal chapters in the question chain" to preliminarily identify the omission of key knowledge points when generating high-level questions; Re-organize the key knowledge points of the problem chapter, automatically review the students' performance in this stage of learning, and adjust the question chain to ensure that all core knowledge points are fully covered and explained.

2. According to claim 1, a high-order question chain generation method based on standardized knowledge graph construction is characterized in that: The standard for dividing the learning chapters into stages in the chemistry learning process can be based on the logical structure of knowledge points, the difficulty level or the interdependence of knowledge points.

3. According to claim 1, a high-order question chain generation method based on standardized knowledge graph construction is characterized in that: The learning efficiency prediction value generated by predicting the students' learning situation in the learning process of each staged learning chapter is compared and analyzed with the pre-set learning efficiency reference range, and the students are grouped according to their learning situation. The specific division steps are as follows: Students whose predicted learning efficiency values ​​are within the learning efficiency reference range are divided into normal learner clusters, students whose predicted learning efficiency values ​​are greater than the maximum value of the learning efficiency reference range are divided into efficient learner clusters, and students whose predicted learning efficiency values ​​are less than the minimum value of the learning efficiency reference range are divided into inefficient learner clusters.

4. According to claim 1, a high-order question chain generation method based on standardized knowledge graph construction is characterized in that: In the process of learning each stage of learning chapter, the specific steps of analyzing the interruption frequency of students' reasoning chain in the problem-solving process and generating the reference value of reasoning chain break are as follows: Monitor each student's problem-solving process in real time, record the number of times the student's reasoning chain is interrupted for each question, and mark the interruption point each time the reasoning chain is interrupted ,in It means that the reasoning of student i on question j is interrupted. The statistical expression of the number of interruptions in the reasoning chain is: , is the number of interruptions of student i during the jth question, Is a logical function of whether the kth reasoning step is completed. If the reasoning chain is interrupted, it is False, otherwise it is True. False is false, indicating logical "error" or "invalidation", and True is true, indicating logical "correctness" or "validation". represents the total number of reasoning steps required to answer the jth question, is an indicator function, which is recorded as 1 when the reasoning chain is interrupted and 0 otherwise; After identifying the number of interruptions in the reasoning chain for each question, a weight is assigned to each interruption. The weight calculation expression is: , where is the weight of the interruption of the kth reasoning step of student i in the jth question, is the coefficient for adjusting weight growth, controlling the weight change range at different reasoning steps, k is the reasoning step where the interruption occurs, is the expected position of the reasoning step when the j-th question is interrupted, indicating the preset reasoning step position of the core part of the knowledge chain; According to the number of interruptions and weights of the reasoning chain in each question, the reference value of the student's reasoning chain interruption in the entire chapter is calculated. The reference value of the reasoning chain interruption reflects the severity of the student's reasoning chain interruption in the entire learning process. It combines the frequency and weight of the interruption of each question and quantifies the incoherence of the student in the logical reasoning process. The calculation expression is: , where is the reference value of the reasoning chain break of student i, M represents the total number of questions in the staged learning chapter, is the complexity coefficient of the j-th question, which is used to adjust the impact of complex questions on the break of the reasoning chain.

5. According to claim 4, a method for generating a high-order question chain based on a standardized knowledge graph is characterized in that: In the process of learning each stage of learning chapter, the specific steps for analyzing the time when students' understanding of a certain knowledge point does not progress during the learning process and generating a reference value for understanding stagnation are as follows: For each student i, record the time when his understanding of a certain knowledge point does not progress during the learning process, and define the continuous stagnation time of the student when learning the knowledge point as , where u represents the number of the knowledge point; Introducing knowledge point complexity factor , used to adjust the impact of dead time on different knowledge points, is the complexity factor of the u-th knowledge point. The complexity factor of the knowledge point is set according to the level of the knowledge point in the knowledge graph and the difficulty level in the student learning path. The specific expression is: ,in, is the hierarchical depth of the u-th knowledge point, indicating the position of the knowledge point in the knowledge graph, is the criticality of the u-th knowledge point in the learning path, indicating its importance in subsequent learning. is the general mastery rate of the u-th knowledge point, which is used to measure students’ mastery of the knowledge point; By combining the stagnation time and the complexity factor of the knowledge point of each student i, the reference value of understanding stagnation is calculated. The calculation expression is: ,in, represents the reference value of student i’s understanding stagnation, is the stagnation time of student i on knowledge point u, w represents the total number of knowledge points involved in the learning chapter of this stage, is the depth of student i’s understanding of knowledge point u.

6. According to claim 1, a method for generating a high-order question chain based on a standardized knowledge graph is characterized in that: Statistical analysis was performed on the individual students in each cluster, and the proportion of inefficient learner clusters in efficient learner clusters was marked as ,in, represents the number of students in the cluster of inefficient learners, represents the number of students in the efficient learner cluster, and the proportion of the inefficient learner cluster in the normal learner cluster is marked as ,in, represents the number of students in the normal learner cluster, and the proportion of the inefficient learner cluster in the efficient learner cluster is marked as The proportion of inefficient learner clusters in the normal learner clusters is marked as Perform comprehensive analysis to generate overall efficiency coefficients , through the overall efficiency coefficient Comprehensively evaluate the overall learning situation of students and the overall efficiency coefficient The calculation expression is: , where and are the proportion of inefficient learner clusters in efficient learner clusters The proportion of inefficient learner clusters in the normal learner clusters The weight of .

7. According to claim 6, a method for generating a high-order question chain based on a standardized knowledge graph is characterized in that: The overall efficiency coefficient generated by comprehensive analysis of the students' overall learning situation is compared with the pre-set overall efficiency coefficient reference threshold, and the students' overall learning situation is divided into normal learning groups and inefficient learning groups. The specific division steps are as follows: If the overall efficiency coefficient is greater than or equal to the overall efficiency coefficient reference threshold, the overall learning situation of the students is divided into an inefficient learning group, and the staged learning chapters corresponding to the inefficient learning group are marked as "abnormal chapters in the question chain" to preliminarily identify the omission of key knowledge points when generating high-level questions; If the overall efficiency coefficient is less than the overall efficiency coefficient reference threshold, the students' overall learning situation will be divided into the normal learning group, and the staged learning chapters corresponding to the normal learning group will not be marked.

8. According to claim 6, a method for generating a high-order question chain based on a standardized knowledge graph is characterized in that: Systematically sort out the key knowledge points in the problem chapters and automatically review the students' learning performance in the chapters at this stage. The specific steps are: First, the knowledge points used in generating the question chain for the phased learning chapter are compared in detail with the knowledge point data set in the entire chapter. Through data analysis, it is checked whether the knowledge points are fully covered in the question chain. Next, identify the knowledge points that were ignored or not explained when generating the question chain, and find out the weak links; Finally, make targeted adjustments to the question chain to ensure that all core knowledge points are fully covered and each key link is explained and guided.

9. A high-order question chain generation system based on a standardized knowledge graph, used to implement the high-order question chain generation method based on a standardized knowledge graph as described in any one of claims 1 to 8, characterized in that: It includes learning content division module, learning data collection and prediction module, student grouping module, settlement statistics analysis module, problem chapter identification module and problem chain adjustment module; The learning content division module divides the chemistry learning process into several phased learning chapters according to the logical progression of knowledge points. Each phased learning chapter covers representative knowledge topics or core concepts. The learning data collection and prediction module collects each student's learning data during the learning process of each stage of learning chapter, analyzes the collected learning data to understand the student's learning outcomes, and uses the pre-trained machine learning model to predict and evaluate each student's learning situation after collecting the student's learning data; The student grouping module groups students according to their learning situations based on the prediction results of the machine learning model. The modules are divided into three clusters: efficient learner cluster, normal learner cluster, and inefficient learner cluster. The cluster statistical analysis module conducts statistical analysis on individual students in each cluster and makes a comprehensive assessment of the overall learning situation of students based on the proportion of inefficient learner clusters in efficient learner clusters and normal learner clusters. The problem chapter identification module divides students' overall learning status into normal learning groups and inefficient learning groups based on the results of comprehensive evaluation, and marks the phased learning chapters corresponding to the inefficient learning groups as "abnormal problem chain chapters", preliminarily identifying the omission of key knowledge points when generating high-level questions; The question chain adjustment module re-sorts the key knowledge points of the question chapter, automatically reviews students' performance in the stage-by-stage learning chapter, and adjusts the question chain to ensure that all core knowledge points are fully covered and explained.

Images