AI-based teaching material generation method and system, computer device, and medium

Abstract

The application relates to an AI-based teaching material generation method and system, computer equipment and medium, and relates to the technical field of teaching, comprising the following steps: acquiring learning behavior performance information of students, comparing according to a preset knowledge point mastery standard, generating a knowledge point mastery degree and a weak table; screening basic materials from a teaching material library based on the behavior information to form a candidate material set; adjusting the difficulty parameters of the candidate materials in combination with the mastery degree and the weak table to obtain difficulty-adapted materials; sorting the difficulty-adapted materials in a knowledge point association order by using a preset AI model, generating an ordered material sequence, and marking the sequence and supplementing the content according to the weak table to obtain target teaching materials. The technical problem that most traditional technologies still remain at the level of simple question pushing or video recommendation and fail to systematically combine the knowledge point mastery degree, weak links and material difficulty is solved.

Description

Technical Field

[0001] This invention relates to the field of teaching technology, and in particular to AI-based teaching material generation methods, systems, computer equipment, and media. Background Technology

[0002] Current methods for generating teaching materials generally suffer from a "one-size-fits-all" approach, lacking in-depth analysis and utilization of students' real-time learning data. Although some systems have attempted to introduce adaptive learning mechanisms, most remain at the level of simple question pushes or video recommendations, failing to systematically couple the level of knowledge mastery, weak areas, and the difficulty and structure of the materials. Furthermore, the content in existing material libraries often lacks fine-grained knowledge point annotation, resulting in a crude selection and reorganization process that struggles to form logically coherent and focused learning paths, thus limiting the effectiveness of teaching interventions and the improvement of learning efficiency. Summary of the Invention

[0003] The purpose of this invention is to at least partially solve one of the technical problems existing in the prior art.

[0004] To achieve the above objectives, this invention provides an AI-based method for generating teaching materials, comprising the following steps: The learning behavior performance information of students is obtained, and the learning behavior performance information is compared with the preset knowledge point mastery standards to obtain a knowledge point mastery level and knowledge point weakness table. Based on the learning behavior performance information, the basic materials in the teaching material library are screened to obtain a candidate material set; Based on the mastery level of the knowledge points and the weak knowledge points table, the difficulty parameters of the candidate material set are adjusted to obtain difficulty-suitable materials. Based on a preset AI model, the difficulty-adapted materials are sorted according to the order of knowledge point association to obtain an ordered material sequence. Then, based on the knowledge point weakness table, key content is marked and added to the ordered material sequence to obtain the target teaching materials.

[0005] Furthermore, based on preset knowledge point mastery standards, the learning behavior performance information is compared to obtain a knowledge point mastery level and knowledge point weakness table, including: The learning behavior performance information is analyzed for knowledge point coverage to obtain behavior coverage knowledge points. The mastery level of each knowledge point in the behavior coverage knowledge points is judged by a preset knowledge point mastery standard to obtain the knowledge point mastery level. Based on the mastery of the knowledge points, weak knowledge points are selected from the knowledge points covered by the behavior to obtain a set of weak knowledge points, and a knowledge point weakness table is generated based on the set of weak knowledge points.

[0006] Furthermore, based on the learning behavior performance information, the basic materials in the teaching material library are screened to obtain a candidate material set, including: The learning behavior performance information is statistically analyzed to obtain a learning behavior statistics table, and a knowledge point requirement table is generated based on the learning behavior statistics table. The knowledge point requirement table is analyzed for knowledge point correlation to obtain a knowledge point correlation table. Based on the knowledge point correlation table, the matching degree of materials in the teaching material library is calculated to obtain a material matching degree table. Based on the material matching degree table, the basic materials in the teaching material library are classified into different levels to obtain a material level set. The material level set is then filtered according to the filtering rules to obtain a candidate material set.

[0007] Furthermore, based on the mastery level of the knowledge points and the candidate material set in the knowledge point weakness table, the difficulty parameters of the materials are adjusted to obtain difficulty-appropriate materials, including: Difficulty parameters are extracted from each material in the candidate material set to obtain an initial material parameter table. The parameter types in the initial material parameter table are then classified to obtain parameter classification results. Based on the mastery level of the knowledge points, a difficulty adjustment coefficient is set for each knowledge point in the candidate material set, a parameter adjustment coefficient table is generated, and the coefficients of the parameter adjustment coefficient table are corrected based on the knowledge point weakness table to obtain a corrected coefficient table. Based on the parameter classification results, the corrected coefficient table is associated and matched with the initial parameter table of the material to obtain a parameter matching table. The initial parameters in the parameter matching table are multiplied with the corresponding correction coefficients to obtain the adjusted parameter table. Based on the adjusted parameter table, the content of the materials in the candidate material set is modified to obtain parameter-adapted materials, and the knowledge point matching verification of the parameter-adapted materials is performed to obtain difficulty-adapted materials.

[0008] Furthermore, based on a preset AI model, the difficulty-adapted materials are sorted according to the order of knowledge point associations to obtain an ordered material sequence, including: Knowledge point tags are extracted from the difficulty-adapted materials to obtain material knowledge point tags, and the material knowledge point tags are matched with the preset teaching syllabus knowledge point database to obtain tag matching results; A knowledge point logical association graph is constructed based on the complete tags in the tag matching results, and the association strength is assigned to the knowledge point logical association graph according to the association order of knowledge points to obtain an association strength table; Based on the association strength table, the complete tags of the difficulty-adapted materials are sorted in association order to obtain an initial material sorting table. The logical coherence of the initial material sorting table is then verified by a preset AI model to obtain an ordered material sequence.

[0009] Furthermore, based on the knowledge point weakness table, key content is marked and added to the ordered material sequence to obtain target teaching materials, including: Frequency statistics are performed on the weak knowledge points in the weak knowledge point weakness table to obtain a weak point frequency table. Based on the weak point frequency table, the corresponding materials in the ordered material sequence are marked to obtain the key marked materials. Content gaps are identified in the key marked materials to obtain the content gaps, and supplementary materials are retrieved from the teaching material library based on the knowledge point weakness table to obtain supplementary materials for the gaps; The missing supplementary materials are inserted into the corresponding positions of the key marked materials to obtain an integrated material sequence. The integrated material sequence is then formatted and standardized to obtain the target teaching materials.

[0010] Furthermore, based on the vulnerability frequency table, corresponding materials in the ordered material sequence are marked to obtain key marked materials, including: Frequency clustering analysis is performed on the frequency table of weaknesses to obtain a key level matrix, and a knowledge point weight network is constructed based on the key level matrix. The ordered material sequence is semantically segmented to obtain a set of semantic units, and the set of semantic units is matched with keywords based on the knowledge point weight network to obtain a matching position table. Based on the matching location table, visual feature allocation is performed to obtain a marker feature set, and multi-dimensional style mapping is performed on the marker feature set to obtain a marker style table; The markup style sheet is applied to the corresponding materials in the ordered material sequence to mark them, resulting in a markup effect image. The markup effect image is then visually optimized to obtain the key marked materials.

[0011] This invention also provides an AI-based teaching material generation system, comprising: The acquisition module is used to acquire students' learning behavior performance information, compare the learning behavior performance information based on the preset knowledge point mastery standards, and obtain a knowledge point mastery level and knowledge point weakness table. The filtering module is used to filter the basic materials in the teaching material library based on the learning behavior performance information to obtain a candidate material set; The adjustment module is used to adjust the difficulty parameters of the materials based on the mastery level of the knowledge points and the candidate material set in the knowledge point weakness table, so as to obtain materials with appropriate difficulty. The sorting module is used to sort the difficulty-adapted materials according to the order of knowledge point association based on a preset AI model to obtain an ordered material sequence, and to mark and add key content to the ordered material sequence based on the knowledge point weakness table to obtain target teaching materials.

[0012] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of any of the methods described above.

[0013] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of any of the methods described above.

[0014] This invention provides an AI-based method for generating teaching materials, comprising the following steps: acquiring students' learning behavior performance information; comparing the learning behavior performance information with a preset knowledge point mastery standard to obtain a knowledge point mastery level and a knowledge point weakness table; filtering basic materials in a teaching material library based on the learning behavior performance information to obtain a candidate material set; adjusting the material difficulty parameters of the candidate material set based on the knowledge point mastery level and knowledge point weakness table to obtain difficulty-suitable materials; sorting the difficulty-suitable materials according to the knowledge point association order based on a preset AI model to obtain an ordered material sequence; and marking and adding key content to the ordered material sequence based on the knowledge point weakness table to obtain target teaching materials. This method solves the technical problem that most traditional technologies still remain at the level of simple question push or video recommendation, failing to systematically combine knowledge point mastery level, weak links, and material difficulty. It achieves the technical effect of initially screening basic materials in the teaching material library based on learning behavior information, effectively filtering out content that is irrelevant to the student's current learning status or redundant, narrowing the candidate range, improving material matching efficiency, and ensuring that the candidate material set covers the required knowledge points and matches the student's cognitive level. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of the steps of an AI-based teaching material generation method in one embodiment of the present invention; Figure 2This is a schematic diagram of the steps of an AI-based teaching material generation system in one embodiment of the present invention; Figure 3 This is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0017] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0018] The embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. The step numbers in the following embodiments are set only for ease of explanation, and there is no limitation on the order between the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0019] The following describes in detail, with reference to the accompanying drawings, an AI-based method for generating teaching materials according to an embodiment of the present invention. First, the AI-based method for generating teaching materials according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

[0020] Figure 1 This invention provides an AI-based method for generating teaching materials, comprising the following steps: Step S1: Obtain students' learning behavior performance information, compare the learning behavior performance information with the preset knowledge point mastery standards, and obtain a knowledge point mastery level and knowledge point weakness table.

[0021] Specifically, this step first acquires information on students' learning behavior, including but not limited to quantifiable data such as their answer records, answer times, error types, number of repeated practice sessions, and video viewing completion rates in exercises, quizzes, classroom interactions, or online learning platforms. Then, this learning behavior information is compared item by item with preset knowledge point mastery standards. These preset standards are pre-defined mastery thresholds or rule systems for each knowledge point; for example, a knowledge point might require an accuracy rate of at least 80% and an answer time within a reasonable range to be considered "mastered." Through this comparison, the system can determine the level of mastery for each knowledge point. The system identifies the student's mastery level based on the points, generating a quantitative result that includes the degree of mastery for each knowledge point (e.g., "Mastered", "Partially Mastered", "Not Mastered"). This result is then further summarized into a knowledge point weakness table, which clearly identifies which knowledge points the student has significant deficiencies in. For example, in a middle school mathematics teaching scenario, if a student's accuracy rate for solving quadratic equations is only 50% in multiple exercises and they frequently skip related explanation videos, the system will determine that this knowledge point is "Not Mastered" based on preset standards, add it to the knowledge point weakness table, and record their mastery level as low. This provides a precise basis for the selection and adjustment of subsequent teaching materials.

[0022] Step S2: Based on the learning behavior performance information, the basic materials in the teaching material library are screened to obtain a candidate material set.

[0023] Specifically, in this step, the basic materials in the teaching material library are screened based on the learning behavior performance information. The specific implementation process is as follows: First, all basic materials with labeled knowledge point tags are retrieved from the teaching material library. These basic materials include structured resources such as courseware, video explanations, example solutions, and practice questions. Then, the student's current learning stage, mastered knowledge points, and unmastered knowledge points reflected in the learning behavior performance information are used as screening conditions and matched with the knowledge point tags associated with the basic materials. Materials related to the student's weak knowledge points or knowledge points that need to be consolidated are retained, while content that only covers knowledge points that have been mastered or is beyond the current learning scope is excluded. This results in a set of candidate materials highly relevant to the student's individual learning status. For example, in a middle school mathematics teaching scenario, if a student's learning behavior shows that they have made multiple mistakes in "solving quadratic equations" and have not watched all related videos, the system will select all basic materials related to the knowledge point of "solving quadratic equations" from the teaching material library, such as standard solution videos, typical error analysis, and step-by-step guided exercises, while skipping content such as "properties of quadratic function graphs" that have not yet been included in the learning plan or have already been mastered. The final set of candidate materials focuses on the knowledge dimensions that the student urgently needs to strengthen, providing an input basis for subsequent adjustment of difficulty parameters.

[0024] Step S3: Based on the mastery level of the knowledge points and the candidate material set in the knowledge point weakness table, adjust the material difficulty parameters to obtain difficulty-suitable materials.

[0025] Specifically, in this step, the difficulty parameters of the candidate material set are adjusted based on the knowledge point mastery level and the knowledge point weakness table. The specific implementation process is as follows: First, the preset difficulty parameters associated with each basic material in the candidate material set are extracted. These parameters are usually expressed in numerical form, such as levels 1 to 5, corresponding to different cognitive levels from basic concepts to comprehensive applications. Then, combined with the specific mastery status of each knowledge point in the knowledge point mastery level (such as "not mastered", "partially mastered", "mastered") and the list of weak knowledge points marked in the knowledge point weakness table, the original difficulty parameters of the materials are dynamically corrected. For materials belonging to weak knowledge points, if their original difficulty is higher than the student's current mastery level, their difficulty parameters are lowered. For example, the original difficulty of the comprehensive application material with a difficulty level of 4 is lowered. The questions are adjusted to basic example problems of difficulty level 2, or the difficulty is reduced by simplifying the language, adding hints, or breaking down the problem. Conversely, if a knowledge point is in the candidate set but the student has a high level of mastery, its difficulty can be maintained or slightly increased to promote consolidation. Ultimately, the materials adjusted in the above way form difficulty-appropriate materials to ensure that the cognitive load matches the student's actual ability. For example, in the context of junior high school mathematics teaching, if a student's mastery of the weak knowledge point of "solving quadratic equations in one variable" is "not mastered", the system replaces or transforms the high-difficulty questions in the candidate set that originally contained complex parameters or required multiple steps of reasoning into basic difficulty materials that only contain standard forms, with step-by-step analysis and graphical assistance, so as to obtain difficulty-appropriate materials that truly match the student's current learning needs.

[0026] Step S4: Based on the preset AI model, the difficulty-adapted materials are sorted according to the order of knowledge point association to obtain an ordered material sequence. Based on the knowledge point weakness table, key content is marked and added to the ordered material sequence to obtain the target teaching materials.

[0027] Specifically, in this step, a pre-set AI model is first invoked. This model has been trained using an educational knowledge graph or curriculum standards, embedding logical dependencies and cognitive progression paths between knowledge points. Then, the difficulty-adapted materials are input into the AI ​​model. The model automatically sorts the materials based on the context of each material's knowledge point within the knowledge system, prioritization constraints, and cognitive complexity levels. For example, it might first introduce basic concepts, then transition to typical examples, and finally introduce variation exercises, thus outputting an ordered sequence of materials that conforms to teaching logic and learning patterns. Next, the system reads the knowledge point weakness table, identifies the marked weak knowledge points, and locates all material positions related to these weak knowledge points in the ordered material sequence. The system performs key content annotation operations at these locations, such as adding highlight marks and inserting "Key Points to Master" prompts. Simultaneously, it supplements auxiliary content according to preset rules, including error analysis, analogies, multiple solution demonstrations, or micro-lesson links. Finally, the integrated materials with annotations and added content constitute the target teaching materials. For example, in a middle school mathematics teaching scenario, if the knowledge point weakness table indicates that students have a weak grasp of "solving quadratic equations in one variable," the system will not only place relevant materials for this knowledge point in a reasonable front position in the ordered material sequence, but also automatically add "Common Mistake Warning" annotations next to the corresponding videos or examples, and insert a supplementary analytical animation specifically addressing discriminant misuse issues, thus forming targeted and clearly structured target teaching materials.

[0028] In a specific embodiment, the learning behavior performance information is compared based on a preset knowledge point mastery standard to obtain a knowledge point mastery level and knowledge point weakness table, including: The learning behavior performance information is analyzed for knowledge point coverage to obtain behavior coverage knowledge points. The mastery level of each knowledge point in the behavior coverage knowledge points is judged by a preset knowledge point mastery standard to obtain the knowledge point mastery level. Based on the mastery of the knowledge points, weak knowledge points are selected from the knowledge points covered by the behavior to obtain a set of weak knowledge points, and a knowledge point weakness table is generated based on the set of weak knowledge points.

[0029] Specifically, in this step, the learning behavior performance information is first analyzed for knowledge point coverage. This involves parsing various behavioral data generated by students during the learning process, including answer records, practice submissions, video viewing logs, and interactive Q&A text. This behavioral data is then mapped and matched with knowledge point tags in the teaching knowledge system to identify which knowledge points have been touched or involved by the student's behavior, ultimately forming a behavioral coverage knowledge point structure. For example, in a middle school mathematics teaching scenario, if a student completes a practice paper containing questions such as "solving quadratic equations using the completing the square method," "finding roots using the quadratic formula," and "applying the discriminant," and watches related explanatory videos, the system will analyze their answers to the questions... The system assigns a video ID and a number to a pre-defined knowledge point tag library to determine the specific knowledge points covered, including items such as "solution of quadratic equations" and "meaning and application of discriminant". Then, it assesses the level of mastery of each knowledge point within the covered knowledge points using pre-defined mastery standards. These standards are typically expressed as quantitative rules, such as "accuracy ≥ 80% and ≤ 2 attempts are considered 'mastered', accuracy between 50% and 80% is considered 'partially mastered', and accuracy below 50% or giving up after multiple errors is considered 'not mastered'". The system then compares each covered knowledge point against these standards, taking into account the student's actual answer accuracy, response time, retry frequency, and other indicators. The system outputs the corresponding knowledge point mastery level results. Next, based on the knowledge point mastery level, it filters weak knowledge points within the behavior-covered knowledge points. Specifically, it extracts knowledge points with a mastery level of "not mastered" or "partially mastered" from all behavior-covered knowledge points, forming a set of weak knowledge points. During this process, if a knowledge point is covered but its mastery level does not reach a preset threshold, it is included in the set, while knowledge points that are "mastered" are excluded. Finally, a knowledge point weakness table is generated based on the set of weak knowledge points. This table lists all weak knowledge points in a structured format, along with their corresponding mastery level, relevant behavioral data summaries (such as average accuracy and typical error types), etc., as information for subsequent analysis. The key input for generating teaching materials; continuing with junior high school mathematics as an example, if a student gets the "Application of Discriminant" knowledge point correct only once out of three practice sessions, and skips the explanation video twice, the system determines that their mastery level is "not mastered" and adds them to the set of weak knowledge points. Finally, the weak knowledge point table clearly marks "Application of Discriminant: Not mastered, accuracy rate 33%, common error is ignoring the discriminant sign judgment"; the entire process is strictly executed according to the original data and preset rules, without introducing subjective judgment, ensuring that the mastery level of knowledge points and the generation of the weak knowledge point table are traceable, consistent and operable, thus providing an accurate and reliable basis for the subsequent personalized generation of AI-based teaching materials.

[0030] In a specific embodiment, based on the level of mastery of the knowledge points, weak knowledge points are screened for the knowledge points covered by the behavior to obtain a set of weak knowledge points, including: The mastery level of the knowledge points is segmented and graded to obtain a knowledge point mastery level sequence, and a threshold analysis is performed on the knowledge point mastery level sequence to obtain a level distribution map; Based on the hierarchical distribution map, the correlation degree of behavior coverage knowledge points is calculated to obtain the knowledge point correlation matrix, and hierarchical clustering is performed on the knowledge point correlation matrix to obtain the knowledge point clustering tree. Weak branches are identified in the knowledge point clustering tree to obtain weak knowledge chains, and knowledge points are added to the weak knowledge chains to obtain complete weak sequences. Based on the complete weak knowledge sequence, the knowledge points are prioritized to obtain a priority knowledge point table. The priority knowledge point table is then linked and integrated to obtain a set of weak knowledge points.

[0031] Specifically, in this step, the mastery level of the knowledge points is first segmented and graded. This involves dividing the mastery level value (such as accuracy, response stability, and other quantitative indicators) for each knowledge point into several discrete levels according to a preset level range. For example, it could be divided into three levels: "Mastered (≥80%)", "Partially Mastered (50%–79%)", and "Not Mastered (<50%)", thus forming a knowledge point mastery level sequence. This sequence is arranged in order of knowledge points, recording the mastery level of each knowledge point covered by each behavior. Subsequently, a threshold analysis is performed on this knowledge point mastery level sequence, i.e., a weakness judgment threshold is set (e.g., both "Partially Mastered" and "Not Mastered" are considered potential weaknesses), and the frequency and distribution density of each level are statistically analyzed. A level distribution map is generated, which presents the proportion and concentration areas of different mastery levels within the overall knowledge points in a visual or structured data format. Next, based on the level distribution map, the correlation degree of the behavioral coverage knowledge points is calculated. Here, correlation degree refers to the logical dependency strength or co-occurrence frequency between knowledge points. The system calculates the correlation weight between any two knowledge points at the weak level based on the course knowledge graph or historical learning path data, constructing a knowledge point correlation matrix. Each element in the matrix represents the correlation strength of the corresponding knowledge point pair. Then, hierarchical clustering is performed on this knowledge point correlation matrix, using an algorithm such as agglomerative hierarchical clustering, with correlation strength as the distance metric, gradually merging highly correlated knowledge point groups, ultimately generating a tree reflecting the clustering of knowledge points. A knowledge point clustering tree with a composite structure is constructed. Based on this, weak branches are identified within the knowledge point clustering tree. This involves traversing each subtree of the tree; if a branch contains multiple knowledge points at the "not mastered" or "partially mastered" level and the overall average mastery is below a set threshold, then the branch is identified as a weak branch, and all knowledge points within it are extracted to form a weak knowledge chain. Subsequently, the weak knowledge chain is supplemented by knowledge point completion. This involves checking for any missing knowledge points that are actually prerequisites for the weak chain but are not covered by the student, based on the prior relationships in the knowledge graph. If any are found, they are added to the chain to obtain a complete weak sequence, ensuring that teaching interventions cover knowledge gaps. Finally, based on the complete weak sequence, knowledge points are prioritized and ranked. The priority rules may include: prioritizing knowledge points over subsequent knowledge points, prioritizing those with lower mastery levels, and prioritizing those with higher relevance levels, thereby generating a priority knowledge point table. Finally, the priority knowledge point table is integrated by associating and combining knowledge points that are semantically similar, structurally adjacent, or frequently taught together, merging or marking their linkage relationships to eliminate redundancy and form a set of weak knowledge points with a clear structure and logical coherence. For example, in a junior high school mathematics teaching scenario, if a student's behavior covers knowledge points including "standard form of a quadratic equation," "completing the square," "formula method," and "application of discriminant," and their mastery levels are "mastered," "partially mastered," "not mastered," and "not mastered," respectively, after segmentation and threshold analysis, the latter three items enter the weak candidate list.Correlation analysis revealed a high correlation between the "formula method" and "discriminant application," forming a weak branch after clustering. Further analysis revealed that while the "square root property" was not covered, it is a necessary prerequisite for the "completing the square method," and was therefore added to the weak chain. Prioritization placed the "square root property" first, followed by "discriminant application" due to its lowest level of mastery. Finally, a set of four weak knowledge points was integrated to guide the generation of key points in subsequent teaching materials.

[0032] In a specific embodiment, the basic materials in the teaching material library are screened based on the learning behavior performance information to obtain a candidate material set, including: The learning behavior performance information is statistically analyzed to obtain a learning behavior statistics table, and a knowledge point requirement table is generated based on the learning behavior statistics table. The knowledge point requirement table is analyzed for knowledge point correlation to obtain a knowledge point correlation table. Based on the knowledge point correlation table, the matching degree of materials in the teaching material library is calculated to obtain a material matching degree table. Based on the material matching degree table, the basic materials in the teaching material library are classified into different levels to obtain a material level set. The material level set is then filtered according to the filtering rules to obtain a candidate material set.

[0033] Specifically, in this step, the learning behavior performance information is first statistically analyzed. This involves collecting raw behavioral data generated by students during the learning process, including answer records, exercise completion status, video viewing time and progress, number of times incorrect questions were re-answered, and interactive questions. This data is then quantified and summarized according to preset dimensions, such as the average accuracy rate, answering time, abandonment rate, and frequency of repeated practice for each knowledge point, thus forming a structured learning behavior statistics table. Next, a knowledge point needs table is generated based on the learning behavior statistics table. This process maps behavioral characteristics to teaching needs by setting rules. For example, if the accuracy rate for a knowledge point is below 60% and the number of re-answers is greater than 2, then that knowledge point is considered to have a "reinforcement need." If the video viewing completion rate is below 50% but related exercises were attempted, it is marked as a "comprehension need." Finally, the knowledge point needs table lists all knowledge points requiring intervention and their corresponding need types and intensities in an itemized format. The process begins with a first level; then, a knowledge point correlation analysis is performed on the knowledge point requirement table. This involves calculating the logical dependencies or teaching synergy frequencies between each knowledge point and other knowledge points based on the course knowledge graph or historical co-occurrence data. For example, "discriminative application" highly depends on mastering the "quadratic equation root-solving formula." These relationships are quantified as correlation weights, generating a knowledge point correlation table where each row records the weight value of a required knowledge point and its associated knowledge points. Based on this, the matching degree of materials in the teaching material library is calculated using the knowledge point correlation table. Specifically, each basic material in the teaching material library (such as micro-lecture videos, example explanations, interactive exercises, etc.) is traversed, and its labeled knowledge point tags and coverage depth are extracted. These are then cross-compared with the knowledge point requirement table and the knowledge point correlation table. If a material covers a high-demand knowledge point or its strongly associated prerequisite knowledge point, a higher matching score is assigned. The matching degree calculation can use a weighted summation model, such as demand intensity × 0.6. + Association weight × 0.4, finally forming a material matching degree table, which records the matching degree value corresponding to each basic material; next, based on the material matching degree table, the basic materials in the teaching material library are classified into levels, that is, all basic materials are divided into several levels according to the matching degree value, such as "high matching (≥0.8)", "medium matching (0.5–0.79)" and "low matching (<0.8)".5) Simultaneously, the materials are categorized in multiple dimensions based on type (video, text, exercises) and cognitive level (memorization, understanding, application) to form a graded set of materials. This set is organized in a structured manner for easy subsequent screening. Then, the graded set of materials is filtered according to pre-set screening rules based on the teaching strategy, such as "only retaining highly and moderately matched materials," "prioritizing exercises containing error analysis," "excluding materials that do not cover any required knowledge points," and "retaining a maximum of 3 representative materials for the same knowledge point." The system executes the filtering operation according to each rule, eliminating materials that do not meet the conditions, and the remaining materials constitute the candidate material set. For example, in a middle school mathematics teaching scenario, if a student's learning behavior shows that their accuracy rate in exercises related to "solving quadratic equations" is only 45%, and they repeatedly skip the formula derivation video, then the learning behavior statistics table generated after behavioral feature statistics will mark this knowledge point as having a high error rate and low viewing completion rate. The knowledge point requirements table then categorized it as a dual requirement of "understanding + reinforcement." Knowledge point correlation analysis revealed a strong correlation between "root-finding formula," "discriminant application," and "completing the square method," with correlation weights of 0.85 and 0.7, respectively. During the matching degree calculation phase, a video explaining "root-finding using the formula method with a comprehensive example involving the discriminant" achieved a matching degree of 0.88 because it covered both the main requirement and the strongly related points, while an older video only explaining the "completing the square method" had a matching degree of only 0.42 because it did not address the current core error. After classification, the former was categorized as a high-matching video, and the latter as a low-matching video. After filtering using rules (such as "excluding materials with a matching degree <0.5" and "retaining 1 video + 2 exercises for each knowledge point"), the final candidate material set included 1 high-matching explanation video, 2 variation exercises addressing misuse of the discriminant, and 1 set of basic reinforcement questions with step-by-step hints. All of these focused on students' current real needs and were structurally sound and complementary, providing high-quality input for subsequent difficulty adjustments and sequence organization.

[0034] In a specific embodiment, the difficulty parameter of the materials is adjusted based on the candidate material set in the knowledge point mastery and knowledge point weakness table to obtain difficulty-adapted materials, including: Difficulty parameters are extracted from each material in the candidate material set to obtain an initial material parameter table. The parameter types in the initial material parameter table are then classified to obtain parameter classification results. Based on the mastery level of the knowledge points, a difficulty adjustment coefficient is set for each knowledge point in the candidate material set, a parameter adjustment coefficient table is generated, and the coefficients of the parameter adjustment coefficient table are corrected based on the knowledge point weakness table to obtain a corrected coefficient table. Based on the parameter classification results, the corrected coefficient table is associated and matched with the initial parameter table of the material to obtain a parameter matching table. The initial parameters in the parameter matching table are multiplied with the corresponding correction coefficients to obtain the adjusted parameter table. Based on the adjusted parameter table, the content of the materials in the candidate material set is modified to obtain parameter-adapted materials, and the knowledge point matching verification of the parameter-adapted materials is performed to obtain difficulty-adapted materials.

[0035] Specifically, in this step, the difficulty parameters of each material in the candidate material set are first extracted. This involves iterating through all basic materials in the candidate material set (such as videos, examples, and exercises) and reading the preset difficulty parameter fields from their metadata. These parameters typically include quantitative or semi-quantitative indicators such as cognitive level (e.g., memorization, comprehension, application, analysis), problem complexity (e.g., number of steps, number of variables), language abstraction, and reliance on background knowledge, thus forming an initial parameter table for the material. This table records the original difficulty parameter values ​​for each material, with each material as a row. Next, the parameter types in the initial parameter table are categorized based on their nature, into structural parameters (e.g., number of steps, number of knowledge points), cognitive parameters (e.g., Bloom's Taxonomy), and expressive parameters (e.g., terminology density, sentence length), etc. The system categorizes knowledge points to obtain parameter classification results, which are then used to implement subsequent differentiated adjustment strategies. Next, based on the level of mastery of these knowledge points, a difficulty adjustment coefficient is set for each knowledge point in the candidate material set. This process is achieved through mapping rules: for example, if a knowledge point is "mastered," the corresponding coefficient is set to 1.0 (maintaining the original difficulty); if it is "partially mastered," the coefficient is set to 0.7 (moderately reduced); if it is "not mastered," the coefficient is set to 0.5 (significantly reduced difficulty). This generates a parameter adjustment coefficient table, where each row corresponds to a knowledge point and its initial adjustment coefficient. Then, based on the knowledge point weakness table, the parameter adjustment coefficient table is corrected. That is, it checks whether the weak knowledge points marked in the weakness table appear in the adjustment coefficient table. If they do, their coefficients are further reduced (e.g., multiplied by 0 from the original coefficient).9) To enhance adaptability to weak points, a revised coefficient table is obtained. Next, based on the parameter classification results, the revised coefficient table is associated and matched with the initial parameter table of the materials. Specifically, for each material, the corresponding correction coefficient is found according to the knowledge points it covers, and the parameter classification results are used to determine which parameter types should be affected by the coefficient (e.g., structural and cognitive parameters are subject to coefficient adjustment, while non-difficulty parameters such as copyright information are not involved), thereby establishing a mapping relationship between parameters and coefficients and forming a parameter matching table. Then, the initial parameters in the parameter matching table are multiplied by the corresponding correction coefficients, that is, for each adjustable parameter, the formula "new parameter value = original parameter value ×" is executed. The calculation of the "correction coefficient" involves rounding down or using a threshold mapping method if the parameter is a discrete level (e.g., cognitive level 1–5). This process generates an adjusted parameter table that fully records the difficulty parameters of each piece of material after adjustment. Based on this, the materials in the candidate material set are modified according to the adjusted parameter table. This process is executed by the content generation engine. For example, if the number of steps in a practice question is reduced from 3 to 2, intermediate derivation steps are removed; if the cognitive level is reduced from "application" to "understanding," open-ended questions are changed to multiple-choice questions with hints; if the language abstraction parameter is reduced, technical terms are replaced with colloquial expressions or diagrams are added, thus obtaining parameter-adapted materials. Finally, the parameter-adapted materials are verified by knowledge point matching, i.e., the modified materials are validated. Whether it still accurately covers the knowledge points originally labeled and does not deviate from the teaching objectives due to simplification is verified through methods including keyword comparison, logical structure check, or AI semantic consistency assessment. If the verification passes, the material is confirmed as difficulty-appropriate; if it fails, it is reverted to the original material or a manual review mechanism is triggered. For example, in a junior high school mathematics teaching scenario, the candidate material set contains a comprehensive application problem about "solving a quadratic equation in one variable." Its initial parameters are: cognitive level 4 (application), number of steps 4, and high terminology density. Since this knowledge point is marked as "not mastered" in the knowledge point weakness table, its corrected coefficient is 0.45. After parameter matching and product calculation, the adjusted parameters become cognitive level 2 (understanding), number of steps 2, and low terminology density. Based on this, the system modifies the original problem to: "Given the equation x²." The question, "5x + 6 = 0, please use the quadratic formula to write the solution," was accompanied by a formula card and numerical substitution hints. During the verification phase, it was confirmed that the revised question still accurately tested the knowledge point of "finding roots using the formula method" and did not introduce irrelevant content. Therefore, this material was officially included in the difficulty-appropriate materials to ensure that it not only matches the students' current ability level but also maintains the consistency of teaching objectives.

[0036] In a specific embodiment, the difficulty-adapted materials are sorted according to the knowledge point association order based on a preset AI model to obtain an ordered material sequence, including: Knowledge point tags are extracted from the difficulty-adapted materials to obtain material knowledge point tags, and the material knowledge point tags are matched with the preset teaching syllabus knowledge point database to obtain tag matching results; A knowledge point logical association graph is constructed based on the complete tags in the tag matching results, and the association strength is assigned to the knowledge point logical association graph according to the association order of knowledge points to obtain an association strength table; Based on the association strength table, the complete tags of the difficulty-adapted materials are sorted in association order to obtain an initial material sorting table. The logical coherence of the initial material sorting table is then verified by a preset AI model to obtain an ordered material sequence.

[0037] Specifically, in this step, the knowledge point tags are first extracted from the difficulty-adapted materials. This involves reading the metadata fields of each piece of difficulty-adapted material item by item, parsing out the labeled knowledge point tags. These tags typically exist in the form of standard codes or natural language phrases, such as "standard form of a quadratic equation," "application of the quadratic formula," and "discriminant sign determination," thus forming a set of knowledge point tags for the materials. This ensures that each piece of material precisely corresponds to the teaching content it carries. Next, the knowledge point tags are matched against a pre-set teaching syllabus knowledge point database. This database is a structured knowledge directory built based on national curriculum standards or school-based teaching systems, containing all legal knowledge points and their unique identifiers, hierarchical relationships, and teaching sequences. The matching process uses string similarity algorithms (such as edit distance or semantic embedding comparison) or precise ID comparison. If a material tag can find a completely identical or semantically equivalent entry in the knowledge point database, it is considered a successful match. The system records the corresponding standard knowledge point ID and its corresponding chapter position, ultimately obtaining the tag matching result, which includes the successfully matched complete tags and their location information in the teaching syllabus. Subsequently, a knowledge point logical association graph is constructed based on the complete tags in the tag matching result. This graph represents the teaching dependencies between knowledge points in a directed graph form, with nodes representing successfully matched knowledge points and edges representing "prerequisite-successor" relationships. For example, "standard form of a quadratic equation" points to "derivation of the quadratic formula," and then to "discriminant application." The graph is constructed according to the preset sequence in the teaching syllabus or the logical path in the knowledge graph. Based on this, the association strength of the knowledge point logical association graph is assigned according to the association order of the knowledge points, that is, a numerical association strength is assigned to each directed edge. This strength can be based on historical learning data (such as the increase in students' accuracy on subsequent knowledge points after mastering prior knowledge points) or expert rules (such as setting the strength from core concepts to application problems to 0).9) The association strength table is calculated based on the curriculum standard weights, with each row recording a pair of knowledge points and their association direction and strength value. Next, the complete tags of the difficulty-appropriate materials are sorted according to their association order based on the association strength table. Specifically, all difficulty-appropriate materials are mapped to the knowledge point logical association graph according to their covered complete tags. Then, a topological sorting algorithm combined with an association strength weighting strategy is used to prioritize materials corresponding to knowledge points with an in-degree of 0 (no pre-dependencies) and high association strength. For cases with multiple paths, the path with the highest cumulative strength or the earlier syllabus number is selected, thus generating an initial material sorting table. This table initially determines the linear presentation order of the materials. Finally, the logical coherence of the initial material sorting table is verified using a preset AI model. This AI model is typically a sequence rationality evaluator trained based on Transformer or graph neural networks, with the input being the initial sorting table. The material sequence and its knowledge point tag sequence in the table are learned by the model based on the coherence patterns learned from a large number of high-quality teaching sequence samples. The model judges whether there are problems such as knowledge jumps, logical gaps or sudden changes in cognitive load in the current sorting. If the verification score is lower than the threshold, a local rearrangement is triggered (such as swapping adjacent materials or inserting transitional content) until the coherence requirement is met. The final output sequence is the ordered material sequence. For example, in the junior high school mathematics teaching scenario, the difficulty-adapted materials include three sets of content: A is a practice problem of "identifying the standard form of a quadratic equation", B is a micro-lesson of "substituting the root formula into calculation", and C is an example problem of "determining the number of roots with the discriminant". After tag extraction and matching, the three correspond to knowledge points K1, K2 and K3 in the teaching syllabus, and the syllabus stipulates that K1→K2→K3 is the standard order. After constructing the logical association graph, the association strength of K1→K2 is 0.85 and K2→K3 is 0.9. The initial sorting table is generated accordingly [A, During AI model validation, it was found that there was a lack of explicit introduction of the "discriminant definition" between B and C. Although C itself had been simplified, the logical connection still seemed abrupt. Therefore, the model suggested inserting a brief explanation of the source of the discriminant after B (this explanation already existed in the difficulty-appropriate materials but was not selected in the initial sorting). The system adjusted accordingly and re-validated, and the final ordered material sequence was [A, B, newly added discriminant explanation, C], ensuring smooth knowledge progression, reasonable cognitive steps, and perfectly matching the learning path needs of students in their current weak state.

[0038] In a specific embodiment, the logical coherence of the initial material sorting table is verified using a preset AI model to obtain an ordered material sequence. The specific steps are as follows: The initial sorting table of the materials is input into a preset AI model, which includes: a knowledge-enhanced pre-trained language model, a bidirectional gated recurrent unit network, and a graph attention network. The initial sorting table of materials is input into the knowledge-enhanced pre-trained language model for contextual semantic encoding to obtain a serialized semantic vector matrix. The knowledge-enhanced pre-trained language model has a 12-layer structure, each of which includes 12 self-attention heads and a 768-dimensional hidden layer. By processing the text content of each material in the initial sorting table, a 768-dimensional contextual semantic vector corresponding to each material is generated, and the contextual semantic vectors are sequentially combined to form the serialized semantic vector matrix. The serialized semantic vector matrix is ​​input into the bidirectional gated recurrent unit network for temporal dependency modeling to obtain a bidirectional temporal feature tensor. The bidirectional gated recurrent unit network has a two-layer structure, with each layer containing 256 hidden units. By performing forward and backward propagation calculations on each 768-dimensional context semantic vector in the serialized semantic vector matrix, the forward and backward hidden states of each time step are concatenated to generate a 512-dimensional bidirectional temporal feature vector. The bidirectional temporal feature vectors of all time steps are combined to form the bidirectional temporal feature tensor. The bidirectional temporal feature tensor is input into the graph attention network for global logical structure analysis to obtain an ordered sequence of materials. The graph attention network has a 3-layer structure, with 8 attention heads in each layer, and uses a LeakyReLU activation function with a negative slope of 0.2. By using each 512-dimensional bidirectional temporal feature vector in the bidirectional temporal feature tensor as a graph node, the normalized attention coefficient between any two nodes is calculated to generate an N×N logical association attention matrix, where N is the number of materials. Based on the logical association attention matrix, a beam search algorithm is used, with a beam width of 5. By iteratively calculating and expanding the candidate sequence with the highest cumulative probability, a globally logically optimal path is finally determined, and the arrangement of materials corresponding to this path is taken as the ordered sequence of materials.

[0039] In a specific embodiment, key content is marked and added to the ordered material sequence based on the knowledge point weakness table to obtain target teaching materials, including: Frequency statistics are performed on the weak knowledge points in the weak knowledge point weakness table to obtain a weak point frequency table. Based on the weak point frequency table, the corresponding materials in the ordered material sequence are marked to obtain the key marked materials. Content gaps are identified in the key marked materials to obtain the content gaps, and supplementary materials are retrieved from the teaching material library based on the knowledge point weakness table to obtain supplementary materials for the gaps; The missing supplementary materials are inserted into the corresponding positions of the key marked materials to obtain an integrated material sequence. The integrated material sequence is then formatted and standardized to obtain the target teaching materials.

[0040] Specifically, in this step, the frequency of weak knowledge points in the weak knowledge point list is first statistically analyzed. This involves iterating through all the weak knowledge points listed in the list and counting the number of times each knowledge point was repeatedly missed, skipped, or had a low completion rate in the historical learning behavior of a student group or individual. For a single student, the frequency of failures related to that knowledge point in different exercises or tests is counted, thus generating a weak point frequency table. This table uses knowledge points as indexes and records their corresponding frequency values. Next, based on the weak point frequency table, the corresponding materials in the ordered material sequence are labeled. This involves checking the knowledge point label covered by each material in the ordered material sequence item by item. If the label matches any weak knowledge point in the weak point frequency table... For knowledge point matching, different levels of emphasis are assigned based on frequency (e.g., high-frequency weak points are marked as "core focus," and mid-frequency points are marked as "needs attention"). This tagged information is embedded into the material metadata or front-end display layer to form key-marked materials. Subsequently, content gap identification is performed on these key-marked materials. This process compares the actual content structure of the key-marked materials with the teaching requirements of the weak knowledge points they cover. For example, it checks whether typical error analysis is included, whether step-by-step guidance is provided, and whether concept clarification is missing. If a key-marked material is found to cover "discriminative application" but fails to explain the common misunderstanding that "Δ<0 has no real roots," then a "concept clarification gap" is identified, thus generating a content gap in the material. Based on this list, supplementary materials are retrieved from the teaching material library according to the knowledge point weakness table. Specifically, the weak knowledge points and their corresponding content gap types are used as search criteria to search for materials with corresponding remedial functions in the teaching material library. For example, searching for "graphical explanation micro-lessons of discriminant Δ<0" or "life analogy cases in situations without real roots" is performed. After sorting the search results by matching degree, the optimal item is selected as the gap supplementary material. Next, the gap supplementary material is inserted into the corresponding position of the key marked material. The insertion rules are determined according to the teaching logic: if the gap is due to unclear preceding concepts, the supplementary material is placed before the original material; if it is due to missing subsequent applications, it is placed after; if it is due to omitted intermediate steps, it is embedded within the original material. The process involves defining specific nodes, such as inserting a "Δ symbol meaning hint card" after the "calculate Δ" step in a problem-solving exercise, thus forming an integrated material sequence. Finally, the integrated material sequence is formatted and standardized. This process includes style standardization (such as font, font size, and color scheme), interactive control alignment (such as button position and prompt box style), media resource adaptation (such as unified video resolution and image compression), and navigation structure consistency (such as chapter numbering and return-to-parent link). This ensures that all original and newly added supplementary materials are seamlessly integrated in terms of visual presentation and user experience. The final output of the target teaching materials retains the logical structure of the ordered material sequence, strengthens precise intervention in weak areas, and possesses good usability and consistency.For example, in a middle school math teaching scenario, a knowledge gap table showed that students repeatedly made mistakes on "discriminant application," and a frequency table of weaknesses indicated that these mistakes occurred 5 times. Therefore, the system marked the example problem about the discriminant in the ordered material sequence as a "core focus." Content gap identification revealed that the example problem only provided the standard solution and did not explain "why Δ<0 has no real solution." The system then retrieved a 30-second dynamic geometry demonstration video from the teaching material library, showing the case where a parabola does not intersect the x-axis, and inserted it as supplementary material before the example problem analysis. During the layout stage, the integrated sequence automatically adjusted the video window size to match the original example problem frame, and the title was unified as "Key Understanding: The Relationship Between Discriminant and Graphs." The resulting target teaching material was not only logically coherent but also provided visual remedies for students' recurring cognitive blind spots, effectively improving the effectiveness of teaching intervention.

[0041] In a specific embodiment, corresponding materials in the ordered material sequence are marked based on the weakness frequency table to obtain key marked materials, including: Frequency clustering analysis is performed on the frequency table of weaknesses to obtain a key level matrix, and a knowledge point weight network is constructed based on the key level matrix. The ordered material sequence is semantically segmented to obtain a set of semantic units, and the set of semantic units is matched with keywords based on the knowledge point weight network to obtain a matching position table. Based on the matching location table, visual feature allocation is performed to obtain a marker feature set, and multi-dimensional style mapping is performed on the marker feature set to obtain a marker style table; The markup style sheet is applied to the corresponding materials in the ordered material sequence to mark them, resulting in a markup effect image. The markup effect image is then visually optimized to obtain the key marked materials.

[0042] Specifically, in this step, the frequency clustering analysis of the weakness point frequency table is first performed. The specific operation involves taking each weak knowledge point in the table and its corresponding error or failure frequency as input data, and using algorithms such as K-means or hierarchical clustering to divide them into several clusters based on frequency, for example, into three categories: "high frequency (≥4 times)," "medium frequency (2–3 times)," and "low frequency (1 time)." Each cluster corresponds to a key level, thereby generating a key level matrix. This matrix uses knowledge points as rows and level categories as columns, using binary or probabilistic forms to identify the key level to which each knowledge point belongs. Next, a knowledge point weight network is constructed based on the key level matrix. This network is represented by a graph structure, where nodes represent weak knowledge points, and edges represent teaching associations or co-occurrence relationships between knowledge points. The weight of each node is mapped from its corresponding key level (e.g., high frequency corresponds to a weight of 0.9, medium frequency 0.6, and low frequency 0).3) The edge weights are inherited from the previously constructed knowledge point association strength, ultimately forming a weighted network that integrates frequency severity and knowledge structure. Subsequently, the ordered material sequence undergoes semantic segmentation, that is, each material in the sequence (such as text paragraphs, video scripts, and question descriptions) is segmented according to semantic integrity. For example, an example question is segmented into semantic units such as "question stem," "solution approach," "formula application," and "conclusion verification," or an explanation video is segmented into multiple semantic segments according to timestamps and content themes, thus obtaining a set of semantic units. Based on this, keyword matching is performed on the set of semantic units based on the knowledge point weight network. Specifically, keywords or named entities are extracted from each semantic unit, and... The semantic unit is matched precisely or semantically with the node labels in the knowledge point weight network (e.g., using BERT embedding to calculate similarity). If a match is successful, the semantic unit ID, corresponding knowledge point, and match confidence are recorded, ultimately forming a matching position table that clearly indicates which semantic units carry which weak knowledge points. Next, based on the matching position table, visual features are assigned, that is, corresponding visual cue features are assigned according to the weight level of the matched knowledge points. For example, high-frequency weak points are assigned "red highlight + flashing icon", mid-frequency weak points are assigned "orange border + emphasis symbol", and low-frequency weak points are assigned "light yellow background", thereby generating a tag feature set. This set defines the visual feedback format that should be presented for different levels of weak content. Then, the tag features are... The system performs multi-dimensional style mapping on the feature set, transforming abstract visual features into specific front-end style parameters, including color codes (e.g., #FF0000), animation types (e.g., pulse), font bolding, icon resource paths, and tooltip text templates. These parameters are then organized into a structured markup style sheet to ensure consistent performance across different devices or media at the same level. Next, the markup style sheet is applied to corresponding materials in an ordered sequence of materials for marking. The system iterates through each semantic unit in the sequence, searches for its corresponding markup style based on a matching position table, and injects the corresponding CSS classes, animation scripts, or UI components during rendering to generate a marked effect image with dynamic emphasis prompts. Finally, a visual layer is applied to the marked effect image. Level optimization involves automatically detecting issues such as overlapping elements, color conflicts, and information overload on a page or interface. It then adjusts the transparency, display duration, spatial layout, or interactive triggering methods of the markers (e.g., expanding detailed explanations only when the mouse hovers over them) to avoid interfering with the reading flow of the main teaching content. Simultaneously, it ensures that key information is sufficiently prominent. The final output of the key marker materials accurately reflects the severity of weak knowledge points and conforms to cognitive load theory and human-computer interaction standards in terms of visual communication. For example, in a middle school mathematics teaching scenario, an ordered sequence of materials includes an example problem about "discriminative application," whose semantic units are segmented into "Problem stem: Determine the root of the equation x² + 2x + 5 = 0" and "Solution: First calculate Δ = b²." "4ac" "Conclusion: Δ<0, no real roots"; The frequency table of weak points shows that the error frequency of "discriminative sign judgment" is 5 times, which is clustered into the high-frequency level and the weight is set to 0.9; Keyword matching identifies that the semantic unit "Δ<0, no real roots" is highly matched with this knowledge point; Based on this, the system assigns a red highlight background and a "common error warning" icon, and defines its CSS class as .highlight-error in the mark style table; After application, the conclusion sentence is displayed on the student's end with a red background and an exclamation mark icon; The visual hierarchy optimization module detects that the mark is too close to the exercise button below, automatically increases the spacing by 10 pixels and limits the highlighted area to no more than the line height, and finally forms a clear, prominent and non-interfering key mark material, effectively guiding students to focus on the core links where they repeatedly make mistakes.

[0043] The above describes the AI-based teaching material generation method in the embodiments of the present invention. The following describes the AI-based teaching material generation system in the embodiments of the present invention. Please refer to [link / reference]. Figure 2 One embodiment of the AI-based teaching material generation system of the present invention includes: The acquisition module 21 is used to acquire students' learning behavior performance information, compare the learning behavior performance information based on the preset knowledge point mastery standards, and obtain a knowledge point mastery level and knowledge point weakness table. The filtering module 22 is used to filter the basic materials in the teaching material library based on the learning behavior performance information to obtain a candidate material set; Adjustment module 23 is used to adjust the material difficulty parameters based on the knowledge point mastery level and the candidate material set in the knowledge point weakness table to obtain difficulty-adapted materials; The sorting module 24 is used to sort the difficulty-adapted materials according to the knowledge point association order based on the preset AI model to obtain an ordered material sequence, and to mark and add key content to the ordered material sequence based on the knowledge point weakness table to obtain the target teaching materials.

[0044] In this embodiment, the specific implementation of each module in the above system embodiment is described in the above method embodiment, and will not be repeated here.

[0045] Reference Figure 3 This invention also provides a computer device whose internal structure can be as follows: Figure 3As shown, the computer device includes a processor, memory, display screen, input device, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores the data corresponding to this embodiment. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements the above-described method.

[0046] Those skilled in the art will understand that Figure 3 The structures shown are merely block diagrams of some structures related to the present invention and do not constitute a limitation on the computer devices on which the present invention is applied.

[0047] An embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method. It is understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0048] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the present invention and embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0049] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0050] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for generating teaching materials based on AI, characterized in that, Includes the following steps: The learning behavior performance information of students is obtained, and the learning behavior performance information is compared with the preset knowledge point mastery standards to obtain a knowledge point mastery level and knowledge point weakness table. Based on the learning behavior performance information, the basic materials in the teaching material library are screened to obtain a candidate material set; Based on the mastery level of the knowledge points and the weak knowledge points table, the difficulty parameters of the candidate material set are adjusted to obtain difficulty-suitable materials. Based on a preset AI model, the difficulty-adapted materials are sorted according to the order of knowledge point association to obtain an ordered material sequence. Then, based on the knowledge point weakness table, key content is marked and added to the ordered material sequence to obtain the target teaching materials.

2. The AI-based teaching material generation method according to claim 1, characterized in that, The learning behavior performance information is compared based on preset knowledge point mastery standards to obtain a knowledge point mastery level and knowledge point weakness table, including: The learning behavior performance information is analyzed for knowledge point coverage to obtain behavior coverage knowledge points. The mastery level of each knowledge point in the behavior coverage knowledge points is judged by a preset knowledge point mastery standard to obtain the knowledge point mastery level. Based on the mastery of the knowledge points, weak knowledge points are selected from the knowledge points covered by the behavior to obtain a set of weak knowledge points, and a knowledge point weakness table is generated based on the set of weak knowledge points.

3. The AI-based teaching material generation method according to claim 1, characterized in that, Based on the learning behavior performance information, the basic materials in the teaching material library are screened to obtain a candidate material set, including: The learning behavior performance information is statistically analyzed to obtain a learning behavior statistics table, and a knowledge point requirement table is generated based on the learning behavior statistics table. A knowledge point correlation analysis is performed on the knowledge point requirement table to obtain a knowledge point correlation table. Based on the knowledge point correlation table, the matching degree of the basic materials in the teaching material library is calculated to obtain a material matching degree table. Based on the material matching degree table, the basic materials in the teaching material library are classified into different levels to obtain a material level set. The material level set is then filtered according to the filtering rules to obtain a candidate material set.

4. The AI-based teaching material generation method according to claim 1, characterized in that, Based on the knowledge point mastery level and the candidate material set in the knowledge point weakness table, the material difficulty parameters are adjusted to obtain difficulty-adapted materials, including: Difficulty parameters are extracted from each material in the candidate material set to obtain an initial material parameter table. The parameter types in the initial material parameter table are then classified to obtain parameter classification results. Based on the mastery level of the knowledge points, a difficulty adjustment coefficient is set for each knowledge point in the candidate material set, a parameter adjustment coefficient table is generated, and the coefficients of the parameter adjustment coefficient table are corrected based on the knowledge point weakness table to obtain a corrected coefficient table. Based on the parameter classification results, the corrected coefficient table is associated and matched with the initial parameter table of the material to obtain a parameter matching table. The initial parameters in the parameter matching table are multiplied with the corresponding correction coefficients to obtain the adjusted parameter table. Based on the adjusted parameter table, the content of the materials in the candidate material set is modified to obtain parameter-adapted materials, and the knowledge point matching verification of the parameter-adapted materials is performed to obtain difficulty-adapted materials.

5. The AI-based teaching material generation method according to claim 1, characterized in that, Based on a preset AI model, the difficulty-adapted materials are sorted according to the order of knowledge point associations to obtain an ordered material sequence, including: Knowledge point tags are extracted from the difficulty-adapted materials to obtain material knowledge point tags, and the material knowledge point tags are matched with the preset teaching syllabus knowledge point database to obtain tag matching results; A knowledge point logical association graph is constructed based on the complete tags in the tag matching results, and the association strength is assigned to the knowledge point logical association graph according to the association order of knowledge points to obtain an association strength table; Based on the association strength table, the complete tags of the difficulty-adapted materials are sorted in association order to obtain an initial material sorting table. The logical coherence of the initial material sorting table is then verified by a preset AI model to obtain an ordered material sequence.

6. The AI-based teaching material generation method according to claim 1, characterized in that, Based on the aforementioned knowledge gap list, key content is marked and added to the ordered material sequence to obtain target teaching materials, including: Frequency statistics are performed on the weak knowledge points in the weak knowledge point weakness table to obtain a weak point frequency table. Based on the weak point frequency table, the corresponding materials in the ordered material sequence are marked to obtain the key marked materials. Content gaps are identified in the key marked materials to obtain the content gaps, and supplementary materials are retrieved from the teaching material library based on the knowledge point weakness table to obtain supplementary materials for the gaps; The missing supplementary materials are inserted into the corresponding positions of the key marked materials to obtain an integrated material sequence. The integrated material sequence is then formatted and standardized to obtain the target teaching materials.

7. The AI-based teaching material generation method according to claim 6, characterized in that, Based on the vulnerability frequency table, corresponding materials in the ordered material sequence are marked to obtain key marked materials, including: Frequency clustering analysis is performed on the weakness frequency table to obtain the key level matrix, and a knowledge point weight network is constructed based on the key level matrix. The ordered material sequence is semantically segmented to obtain a set of semantic units, and the set of semantic units is matched with keywords based on the knowledge point weight network to obtain a matching position table. Based on the matching location table, visual feature allocation is performed to obtain a marker feature set, and multi-dimensional style mapping is performed on the marker feature set to obtain a marker style table; The markup style sheet is applied to the corresponding materials in the ordered material sequence to mark the corresponding materials, resulting in a markup effect image. The markup effect image is then visually optimized to obtain the key marked materials.

8. An AI-based teaching material generation system, characterized in that, The method for generating AI-based teaching materials according to any one of claims 1 to 7 includes: The acquisition module is used to acquire students' learning behavior performance information, compare the learning behavior performance information based on the preset knowledge point mastery standards, and obtain a knowledge point mastery level and knowledge point weakness table. The filtering module is used to filter the basic materials in the teaching material library based on the learning behavior performance information to obtain a candidate material set; The adjustment module is used to adjust the difficulty parameters of the materials based on the mastery level of the knowledge points and the candidate material set in the knowledge point weakness table, so as to obtain materials with appropriate difficulty. The sorting module is used to sort the difficulty-adapted materials according to the order of knowledge point association based on a preset AI model to obtain an ordered material sequence, and to mark and add key content to the ordered material sequence based on the knowledge point weakness table to obtain target teaching materials.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images