Optimization method and system for generating teaching content through large model

Abstract

The invention provides an optimization method and system for generating teaching content through a large model, and relates to the technical field of education. Generating initial teaching content by using the multi-modal large model; constructing a teaching knowledge graph according to the authoritative education data, and calculating the integrating degree of the initial teaching content and the teaching knowledge graph based on a knowledge point hierarchical relationship and a course target in the teaching knowledge graph; comparing the integrating degree with a preset threshold value to obtain corrected teaching content; selecting a cue word framework from a preset cue word template library, and generating optimized teaching content in combination with the corrected teaching content; according to the teacher and student feedback data and the authoritative education data, the matching degree of the optimized teaching content and the course teaching outline and the student understanding difficulty score are confirmed, and the self-adaptive generation and iterative optimization of the teaching content are realized in combination with the preset standard, so that the integrating degree of the teaching content and the course target and the student understanding degree are improved.

Description

Technical Field

[0001] This application relates to the field of educational technology, and in particular to an optimized method and system for generating teaching content from large models. Background Technology

[0002] With the deepening of digital transformation in education, large models are being used more and more widely in the generation of teaching content. Especially when dealing with complex and abstract knowledge, it is urgent to dynamically adjust the content presentation format according to students' understanding to improve learning efficiency.

[0003] Currently, mainstream solutions assess students' comprehension of content generated by large models by analyzing their interactive behavior data during the learning process. When a comprehension bottleneck is detected, a 3D modeling engine is invoked to convert 2D teaching materials into 3D graphics, which are then pushed to students' tablets or AR glasses via wireless network for display. However, existing solutions have certain shortcomings. For example, while 3D graphics have a greater sense of space than 2D images, they lack realistic spatial depth and interactive immersion, requiring students to have strong spatial imagination to understand complex structures. Furthermore, the system only adjusts content based on surface interaction data, failing to deeply integrate the curriculum knowledge system and teaching objectives, resulting in a disconnect between the visualized content and the teaching logic, and hindering the continuous optimization of the content generation model. Summary of the Invention

[0004] The purpose of this application is to provide an optimized method and system for generating teaching content from large models, in order to solve the problems in existing technologies such as the lack of realistic spatial depth and interactive immersion; the disconnect between visual content and teaching logic, making it difficult to continuously feed back into the optimization of the content generation model.

[0005] To address the aforementioned technical problems, firstly, this application provides an optimized method for generating teaching content from large models, comprising:

[0006] Acquire authoritative educational data and teacher / student feedback data in teaching scenarios. The authoritative educational data includes course syllabus, course objectives, and hierarchical relationship data of knowledge points. The teacher / student feedback data includes teachers' error annotation data on historically generated teaching content and students' feedback data on the difficulty of understanding historically generated teaching content.

[0007] Using a multimodal large model, the authoritative educational data and the teacher and student feedback data are correlated and matched to generate initial teaching content, which includes test questions and question explanations.

[0008] Based on the authoritative educational data, a teaching knowledge graph is constructed. Based on the hierarchical relationship of knowledge points and course objectives in the teaching knowledge graph, the degree of fit between the initial teaching content and the teaching knowledge graph is calculated.

[0009] The fit is compared with a preset threshold to correct the initial teaching content and obtain the corrected teaching content.

[0010] Select a prompt word framework from the preset prompt word template library, combine it with the revised teaching content, and use a multimodal large model to generate optimized teaching content. The prompt word framework includes expression norms and optimization directions adapted to the teaching scenario.

[0011] Based on the teacher and student feedback data and the authoritative educational data, the matching degree between the optimized teaching content and the course syllabus and the student comprehension difficulty score are confirmed. Combined with the preset standards, the optimized teaching content is iteratively optimized.

[0012] Optionally, a multimodal large model is used to correlate and match the authoritative educational data and the teacher and student feedback data to generate initial teaching content. The initial teaching content includes test questions and question explanations, including:

[0013] Using a multimodal large model, the core knowledge points of the course syllabus are matched with the error types of the error annotation data to generate a knowledge point error table;

[0014] The competency development requirements of the course objectives are matched with the difficulty range of the comprehension difficulty feedback data to generate a competency difficulty table;

[0015] The knowledge point association paths in the knowledge point hierarchy relationship data are matched with the high-frequency comprehension difficulties in the comprehension difficulty feedback data to generate a path difficulty table;

[0016] Based on the knowledge point error table, the ability difficulty table, and the path difficulty table, test questions and corresponding question explanations are determined to generate initial teaching content. The test questions include basic verification questions, error avoidance questions, and comprehensive application questions.

[0017] Optionally, based on the knowledge point error table, the ability difficulty table, and the path difficulty table, test questions and corresponding question explanations are determined to generate initial teaching content. The test questions include basic verification questions, error avoidance questions, and comprehensive application questions, including:

[0018] Based on the knowledge point error table, the examination direction for each core knowledge point is determined. The examination direction includes the direction of basic concept verification, the direction of high-frequency error avoidance, and the direction of comprehensive application of multiple knowledge points.

[0019] Based on the aforementioned ability difficulty table and the aforementioned path difficulty table, basic verification questions are generated for the core knowledge points of the basic concept verification direction, error avoidance questions are generated for the core knowledge points of the high-frequency error avoidance direction, and comprehensive application questions are generated for the core knowledge points of the multi-knowledge point comprehensive application direction.

[0020] Based on the difficulty range corresponding to each core knowledge point in the difficulty table, the analysis depth corresponding to each test question is determined. Based on the high-frequency understanding difficulty nodes corresponding to the knowledge point association paths in the path difficulty table, the important difficulty explanation content corresponding to each test question is determined. Based on the error type corresponding to each core knowledge point in the knowledge point error table, the error prompt content corresponding to each question is determined. By integrating the analysis depth, the important difficulty explanation content, and the error prompt content, the analysis of the questions to be adjusted is obtained. Among them, each test question includes basic verification questions, error avoidance questions, and comprehensive application questions.

[0021] Calculate the matching degree between the analysis depth of the question to be adjusted and the difficulty range corresponding to each core knowledge point in the ability difficulty table. When the matching degree is lower than the preset matching threshold, adjust the analysis depth of the question to be adjusted to obtain the question analysis. Combine the questions with the test questions to generate the initial teaching content.

[0022] Optionally, based on the authoritative educational data, a teaching knowledge graph is constructed. Based on the hierarchical relationships of knowledge points and course objectives within the teaching knowledge graph, the degree of fit between the initial teaching content and the teaching knowledge graph is calculated, including:

[0023] Using the core knowledge points in the authoritative educational data as nodes, the knowledge point association paths in the knowledge point hierarchical relationship data as edges between nodes, and the ability cultivation requirements of the course objectives as attributes of the corresponding nodes, a teaching knowledge graph is constructed.

[0024] Extract the knowledge points and assessment objectives from the test questions of the initial teaching content, and extract the analytical knowledge points from the question analysis of the initial teaching content;

[0025] Based on the hierarchical relationship of knowledge points and course objectives in the teaching knowledge graph, the first degree of overlap between the examination knowledge points and the core knowledge points, the second degree of overlap between the examination objectives and the ability cultivation requirements, and the third degree of overlap between the actual path formed by the analytical knowledge points in the teaching knowledge graph and the path associated with the knowledge points are calculated.

[0026] Based on preset dimension weights, the first overlap, the second overlap, and the third overlap are weighted and summed to obtain the fit between the initial teaching content and the teaching knowledge graph.

[0027] Optionally, the fit is compared with a preset threshold to modify the initial teaching content, resulting in modified teaching content, including:

[0028] The fit is compared with a preset threshold. If the fit is greater than or equal to the preset threshold, the initial teaching content is used as the revised teaching content.

[0029] If the degree of fit is less than the preset threshold and the first degree of overlap is less than the first preset sub-threshold, then supplement the test questions and question analysis corresponding to the missing knowledge points in the teaching knowledge graph until the first degree of overlap is greater than or equal to the first preset sub-threshold, and obtain the corrected teaching content.

[0030] If the degree of fit is less than the preset threshold and the second degree of overlap is less than the second preset sub-threshold, then the target test questions in the initial teaching content that do not match the examination objectives and ability cultivation requirements are adjusted until the second degree of overlap is greater than or equal to the second preset sub-threshold, and the corrected teaching content is obtained.

[0031] If the degree of fit is less than a preset threshold, and the third degree of overlap is less than a third preset sub-threshold, then the target question analysis in the initial teaching content is corrected to prevent the actual path of question analysis and the knowledge point association path from matching, until the third degree of overlap is greater than or equal to the third preset sub-threshold, and the corrected teaching content is obtained.

[0032] Optionally, a prompt word framework is selected from a preset prompt word template library, and combined with the revised teaching content, an optimized teaching content is generated using a multimodal large model. The prompt word framework includes expression norms and optimization directions adapted to the teaching scenario, including:

[0033] The teaching scenarios of the revised teaching content are analyzed to obtain the core knowledge point types, test question difficulty levels, and question analysis length characteristics, so as to determine the teaching scenario type corresponding to the revised teaching content.

[0034] Extract the prompt word framework corresponding to the teaching scenario type from the preset prompt word template library;

[0035] Key content elements are extracted from the revised teaching content, and the key content elements are formatted according to the expression specifications of the prompt word framework to obtain formatted content elements.

[0036] The optimization directions of the formatted content elements and the prompt word framework are integrated into prompt words. Based on the prompt words, the modified teaching content is optimized using the multimodal large model to generate preliminary optimized teaching content.

[0037] If the preliminarily optimized teaching content cannot meet the key optimization indicators in the prompt word framework, the prompt words are modified until the generated preliminarily optimized teaching content meets the key optimization indicators in the prompt word framework, thus obtaining the optimized teaching content.

[0038] Optionally, key content elements are extracted from the revised teaching content, and the key content elements are formatted according to the expression specifications of the prompt word framework to obtain formatted content elements, including:

[0039] Extract the core knowledge point descriptions and test question stems corresponding to the test questions from the test questions of the revised teaching content, and extract the core logic of the question analysis corresponding to the test question stems from the question analysis of the revised teaching content.

[0040] Based on the core logic of the question analysis and combined with the ability development requirements, the angle from which the test questions examine the ability development requirements is determined, and the angle of examination is taken as the examination point corresponding to the ability development requirements.

[0041] Replace the non-standard terms in the description of the core knowledge points and the non-standard terms in the examination points with standard terms that conform to the expression specifications of the prompt word framework to obtain formatted knowledge point descriptions and formatted examination points;

[0042] The format of the test question stem and the core logic of the question analysis are adjusted to conform to the language style requirements in the expression specification of the prompt word framework, so as to obtain the formatted question stem and the formatted core logic.

[0043] The formatted knowledge point description, the formatted examination points, the formatted question stem, and the formatted core logic are integrated to obtain the formatted content elements.

[0044] Secondly, this application provides an optimization system for generating teaching content from large models, including:

[0045] The acquisition module is used to acquire authoritative educational data and teacher and student feedback data in the teaching scenario. The authoritative educational data includes course teaching outline, course objectives and knowledge point hierarchical relationship data. The teacher and student feedback data includes teachers' error annotation data on historically generated teaching content and students' feedback data on the difficulty of understanding historically generated teaching content.

[0046] The association module is used to use a multimodal large model to associate and match the authoritative educational data and the teacher and student feedback data to generate initial teaching content, which includes test questions and question explanations.

[0047] The module is used to construct a teaching knowledge graph based on the authoritative educational data, and to calculate the fit between the initial teaching content and the teaching knowledge graph based on the hierarchical relationship of knowledge points and course objectives in the teaching knowledge graph.

[0048] The correction module is used to compare the fit with a preset threshold to correct the initial teaching content and obtain the corrected teaching content.

[0049] The optimization module is used to select a prompt word framework from a preset prompt word template library, combine it with the revised teaching content, and use a multimodal large model to generate optimized teaching content. The prompt word framework includes expression norms and optimization directions adapted to the teaching scenario.

[0050] The confirmation module is used to confirm the matching degree between the optimized teaching content and the course syllabus and the student comprehension difficulty score based on the teacher and student feedback data and the authoritative educational data, and to iteratively optimize the optimized teaching content in combination with preset standards.

[0051] Thirdly, this application provides an electronic device, comprising:

[0052] Memory, used to store computer programs;

[0053] A processor is configured to execute the computer program to implement the steps of an optimized method for generating teaching content from a large model as described in the first aspect above.

[0054] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps of the optimized method for generating teaching content from a large model as described in the first aspect above.

[0055] This application provides an optimized method for generating teaching content using a large-scale model. This method acquires authoritative educational data and teacher / student feedback data within a teaching scenario. The authoritative educational data includes course syllabus, course objectives, and hierarchical relationship data of knowledge points. The teacher / student feedback data includes teachers' error annotations of previously generated teaching content and students' feedback data on the difficulty of understanding the previously generated teaching content. Using a multimodal large-scale model, the authoritative educational data and the teacher / student feedback data are correlated and matched to generate initial teaching content, which includes test questions and question explanations. Based on the authoritative educational data, a teaching knowledge graph is constructed, and based on the knowledge point hierarchy in the teaching knowledge graph... The system calculates the fit between the initial teaching content and the teaching knowledge graph based on hierarchical relationships and course objectives. The fit is compared with a preset threshold to revise the initial teaching content, resulting in revised teaching content. A prompt word framework is selected from a preset prompt word template library and, combined with the revised teaching content, an optimized teaching content is generated using a multimodal large model. The prompt word framework includes expression norms and optimization directions adapted to the teaching scenario. Based on teacher and student feedback data and authoritative educational data, the matching degree between the optimized teaching content and the course syllabus, as well as student comprehension difficulty scores, are confirmed. Iterative optimization of the optimized teaching content is then performed based on preset standards. By acquiring authoritative educational data and teacher and student feedback data in the teaching scenario, a systematic collection of core information required for teaching content generation is achieved, improving the intelligence level and teaching adaptability of content generation. This approach ensures the integrity and systematic nature of the teaching logic; it enables automatic identification and correction of knowledge deviations and target deviations in the generated content, enhancing the accuracy and compliance of the teaching content; it ensures that the output content is highly matched with the specific teaching context in terms of language style, cognitive difficulty, and teaching strategies, improving the clarity of expression and the guidance of teaching. Furthermore, this application uses a multimodal large-scale model to match core knowledge points in the course syllabus with the types of errors marked by teachers to generate a knowledge point error table, matches the ability requirements in the course objectives with the students' comprehension difficulty range to generate an ability difficulty table, and matches the correlation paths between knowledge points with students' frequently encountered comprehension difficulties to generate a path difficulty table; thus obtaining test questions and question explanations; and resolving the problem of content generation being disconnected from teaching objectives in existing solutions. Attached Figure Description

[0056] To more clearly illustrate the technical solutions of 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.

[0057] Figure 1A flowchart illustrating an optimization method for generating teaching content from a large model, provided in an embodiment of this application;

[0058] Figure 2 A schematic diagram illustrating a specific implementation of an optimization method for generating teaching content from a large model, provided in this application embodiment;

[0059] Figure 3 A schematic diagram of the structure of an optimization system for generating teaching content from a large model, provided in an embodiment of this application; Detailed Implementation

[0060] To address the problems of existing solutions that rely solely on students' surface-level interactive behavior data for adjustments, resulting in a disconnect between content and teaching objectives, a lack of immersive visualization, and difficulty in forming an effective optimization loop, this application proposes an optimized method and system for generating teaching content using a large-scale model. By systematically integrating authoritative data such as course outlines, knowledge point levels, and teaching objectives, as well as in-depth feedback information such as teacher error annotations and student comprehension difficulties, a structured teaching knowledge graph is constructed. This enables quantifiable evaluation of the generated content at the knowledge logic and training objective levels. Furthermore, by combining a multimodal large-scale model with a pre-set prompt word framework, the limitations of existing solutions are overcome, realizing a process of dynamic adaptation of teaching content from static presentation.

[0061] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0062] The core of this application is to provide an optimized method for generating teaching content from large models, and a flowchart of one specific implementation method is shown below. Figure 1 As shown, the method includes:

[0063] Step 101: Obtain authoritative educational data and teacher / student feedback data in the teaching scenario. The authoritative educational data includes course syllabus, course objectives, and hierarchical relationship data of knowledge points. The teacher / student feedback data includes teachers' error annotation data on historically generated teaching content and students' feedback data on the difficulty of understanding historically generated teaching content.

[0064] In this step, authoritative educational data in the teaching scenario refers to authoritative educational data used to support the generation of teaching content; teacher and student feedback data refers to feedback data given by teachers and students regarding past teaching content; course syllabus refers to a guiding document that specifies the teaching content, requirements, and arrangements for a particular course; course objectives refer to the learning goals that students are expected to achieve in a particular course; knowledge point hierarchy data refers to data recording the hierarchical or relational relationships between various knowledge points; historically generated teaching content refers to teaching content that has been generated before the implementation of this method; error annotation data refers to data formed by teachers annotating errors in historically generated teaching content; and comprehension difficulty feedback data refers to data formed by students providing feedback on the difficulty level of comprehension of historically generated teaching content.

[0065] In this embodiment of the application, authoritative educational data in the teaching scenario is obtained through official resources provided by the education management department for specific teaching scenarios. At the same time, error annotation data of teachers on historical generated teaching content is collected through teachers' daily teaching records, such as records of teachers annotating logical errors in test questions in historical generated teaching content. Feedback data on the difficulty of understanding historical generated teaching content is collected through student after-class feedback questionnaires or online feedback systems, such as whether students find the explanation of a certain knowledge point in historical generated teaching content difficult to understand, or whether a certain test question is too difficult. In this way, authoritative educational data and teacher and student feedback data in the teaching scenario are obtained.

[0066] Step 102: Using a multimodal large model, the authoritative educational data and the teacher and student feedback data are correlated and matched to generate initial teaching content, which includes test questions and question explanations.

[0067] In this step, the multimodal large model refers to an artificial intelligence model that can process multiple modal data such as text and images, and can perform correlation analysis and content generation based on the data; correlation matching refers to the operation of correspondingly associating two or more types of data according to a certain logical relationship; initial teaching content refers to the teaching-related content generated for the first time after correlation matching through the multimodal large model; test questions refer to questions used to examine students' mastery of knowledge points; and question analysis refers to the content of answering test questions, explaining the thought process, and expanding knowledge points.

[0068] Step 103: Based on the authoritative educational data, construct a teaching knowledge graph, and calculate the fit between the initial teaching content and the teaching knowledge graph based on the hierarchical relationship of knowledge points and course objectives in the teaching knowledge graph.

[0069] In this step, the teaching knowledge graph refers to a structured graph constructed with knowledge points as the core; the hierarchical relationship of knowledge points refers to the hierarchical, subordinate, or dependent relationships between knowledge points; the fit degree refers to the degree of matching between the initial teaching content and the teaching knowledge graph; the first overlap degree refers to the proportion of overlap between the knowledge points examined in the test questions of the initial teaching content and the core knowledge points in the teaching knowledge graph; the second overlap degree refers to the proportion of overlap between the examination objectives of the test questions of the initial teaching content and the ability development requirements of the core knowledge point nodes in the teaching knowledge graph; the third overlap degree refers to the proportion of overlap between the actual path formed by the analytical knowledge points of the questions in the initial teaching content in the teaching knowledge graph and the knowledge point association path; the preset dimension weights refer to the pre-set weight values ​​corresponding to the first, second, and third overlap degrees when calculating the fit degree.

[0070] Step 104: Compare the fit with a preset threshold to modify the initial teaching content and obtain the modified teaching content.

[0071] In this step, the preset threshold refers to a pre-set critical value used to determine whether the fit is up to standard; the revised teaching content refers to the teaching content that meets the requirements after the initial teaching content has been revised; the first preset sub-threshold refers to a pre-set critical value used to determine whether the first degree of overlap is up to standard; the missing knowledge point refers to the core knowledge point that exists in the teaching knowledge graph but is not covered in the initial teaching content; the second preset sub-threshold refers to a pre-set critical value used to determine whether the second degree of overlap is up to standard; the target test question refers to the test question in the initial teaching content where the examination objectives and ability cultivation requirements do not match; the third preset sub-threshold refers to a pre-set critical value used to determine whether the third degree of overlap is up to standard; and the target question analysis refers to the question analysis in the initial teaching content where the actual path of analyzing the knowledge point does not match the path of the knowledge point association.

[0072] Step 105: Select a prompt word framework from the preset prompt word template library, combine it with the revised teaching content, and use a multimodal large model to generate optimized teaching content. The prompt word framework includes expression norms and optimization directions adapted to the teaching scenario.

[0073] In this step, the preset prompt word template library refers to a pre-established database containing various prompt word frameworks; the prompt word framework refers to a structured prompt word template used to guide the multimodal large model to generate content that meets the requirements; the expression standard refers to a standard used to standardize the language expression of teaching content, adapted to specific teaching scenarios; the optimization direction refers to the specific direction for optimizing the revised teaching content; the optimized teaching content refers to the teaching content obtained after the revised teaching content has been optimized by the multimodal large model; the key content elements refer to the elements extracted from the revised teaching content that play a core role in generating the optimized teaching content; the formatted content elements refer to the content obtained after the key content elements have been formatted according to the expression standard; the prompt words refer to the text used to input the multimodal large model after integrating the formatted content elements and the optimization direction; the initially optimized teaching content refers to the optimized teaching content generated by the multimodal large model based on the prompt words for the first time; and the key optimization indicators refer to the indicators set in the prompt word framework used to judge whether the initially optimized teaching content meets the standards.

[0074] Step 106: Based on the teacher and student feedback data and the authoritative educational data, confirm the matching degree between the optimized teaching content and the course syllabus and the student comprehension difficulty score, and iteratively optimize the optimized teaching content in conjunction with the preset standards.

[0075] In this step, the matching degree refers to the degree of matching between the optimized teaching content and the course syllabus; the student comprehension difficulty score refers to the score given based on student feedback data on the difficulty of understanding the optimized teaching content; the preset standard refers to the pre-set standard used to judge whether the optimized teaching content needs iterative optimization; the knowledge point coverage matching ratio refers to the overlap ratio between the knowledge points involved in the optimized teaching content and the core knowledge points of the course syllabus; the goal alignment matching ratio refers to the consistency ratio between the examination direction of the optimized teaching content and the course goals of the course syllabus; the difficulty level frequency ratio refers to the proportion of the number of test questions of each difficulty level in the optimized teaching content to the total number of test questions; and the average number of points of confusion refers to the average number of student feedback points of confusion received for each test question and its corresponding explanation in the optimized teaching content.

[0076] In this embodiment, based on teacher and student feedback data regarding the matching accuracy of the optimized teaching content with the curriculum syllabus, student difficulty assessment data regarding the optimized teaching content, and core knowledge points and curriculum objectives from authoritative educational data, the knowledge point coverage matching ratio and objective fit matching ratio are calculated. Combined with preset matching degree calculation weights, the matching degree between the optimized teaching content and the curriculum syllabus is calculated using the formula: Matching Degree = Knowledge Point Coverage Matching Ratio × Preset Knowledge Point Coverage Weight + Objective Fit Matching Ratio × Preset Objective Fit Weight. Simultaneously, based on student difficulty assessment data, the frequency percentage of difficulty levels and the average number of points of confusion are calculated. Combined with preset scoring weights, the student difficulty assessment score is calculated: Student Difficulty Assessment Score = Difficulty Level Frequency Percentage × Preset Difficulty Weight + Average Number of Points of Confusion × Preset Confusion Weight. The matching degree and student comprehension difficulty score are compared with the preset standards. If both are met, there is no need to iterate and optimize the optimized teaching content. If one or both are not met, the optimized teaching content is adjusted, and the matching degree and student comprehension difficulty score are recalculated until the preset standards are met, thus completing the iterative optimization of the optimized teaching content.

[0077] This application's embodiments address the problems in the prior art where the generation of teaching content lacks authoritative data support, has low matching degree with the knowledge system, is difficult to adapt to students' comprehension difficulty, and is not precise in optimization. It realizes the precise and personalized generation of teaching content, and improves the quality and applicability of teaching content.

[0078] This application provides a specific embodiment, such as Figure 2 As shown, step 102 involves using a multimodal large model to correlate and match the authoritative educational data and the teacher-student feedback data to generate initial teaching content. This initial teaching content includes test questions and question explanations, specifically including the following steps:

[0079] Step 201: Using a multimodal large model, match the core knowledge points of the course syllabus with the error types of the error annotation data to generate a knowledge point error table.

[0080] In this step, the knowledge point error table refers to a structured table that records the historical error types corresponding to each core knowledge point; the core knowledge points of the course syllabus refer to the knowledge points specified in the course syllabus that students need to master; the error types of the error annotation data refer to the error categories related to knowledge points that teachers have annotated for the historically generated teaching content; and the multimodal big model refers to an artificial intelligence model that can process and associate multiple data types to achieve data matching and content generation.

[0081] In this embodiment, a multimodal large model is used to extract and classify the core knowledge points of the course syllabus, clarify the specific content and scope of each core knowledge point, identify and classify the error types in the error-annotated data, and distinguish the characteristics and applicable scenarios of different error types. Through the multimodal large model, an association mapping relationship between core knowledge points and error types is established. For each core knowledge point, all error types that have appeared in the past are selected to ensure that each core knowledge point can be matched with its associated error type. This correspondence between core knowledge points and error types is organized in tabular form to generate a knowledge point error table.

[0082] Step 202: Match the ability development requirements of the course objectives with the difficulty range of the comprehension difficulty feedback data to generate an ability difficulty table.

[0083] In this step, the competency development requirements refer to the specific competencies that need to be developed in students through course teaching as specified in the course objectives; the difficulty range of the difficulty feedback data refers to the range of different levels of difficulty divided according to students' feedback on the historically generated teaching content; and the competency difficulty table refers to a structured table that records the appropriate difficulty range corresponding to each competency development requirement.

[0084] In this embodiment, the competency development requirements of the curriculum objectives from authoritative educational data and the comprehension difficulty feedback data from teacher and student feedback data are input into a multimodal big data model. The multimodal big data model first breaks down the competency development requirements, clarifying the specific connotation and training objectives of each competency development requirement; it then analyzes the comprehension difficulty feedback data, dividing it into different difficulty ranges based on the students' comprehension levels, and statistically analyzing the percentage of student feedback within each difficulty range; subsequently, the multimodal big data model matches each competency development requirement with an optimal difficulty range based on the compatibility between the complexity of the competency development requirements and the difficulty of student feedback; finally, the relationship between the competency development requirements and the corresponding difficulty ranges is compiled into a table to generate a competency difficulty table, which will be used to determine the difficulty level of test items in subsequent tests.

[0085] Step 203: Match the knowledge point association paths in the knowledge point hierarchy relationship data with the high-frequency comprehension difficulties in the comprehension difficulty feedback data to generate a path difficulty table.

[0086] In this step, the knowledge point association path in the knowledge point hierarchy relationship data refers to the path in the knowledge point hierarchy relationship data that records the sequential dependence or logical relationship between knowledge points; the high-frequency comprehension difficulty feedback data refers to the knowledge points or knowledge point association nodes that appear more frequently than the preset frequency threshold in students' feedback on the historically generated teaching content and are marked by students as difficult to understand; the path difficulty table refers to the structured table that records the high-frequency comprehension difficulties existing in each knowledge point association path.

[0087] In this embodiment, hierarchical relationship data of knowledge points from authoritative educational data and comprehension difficulty feedback data from teacher and student feedback data are input into a multimodal big data model. The multimodal big data model first extracts all knowledge point association paths from the hierarchical relationship data, sorts out the order and logical relationship of knowledge points in each path, and identifies key nodes in the path; it then filters out high-frequency comprehension difficulties from the comprehension difficulty feedback data, counts the frequency of each difficulty, and locates the corresponding knowledge point or knowledge point association node for each difficulty; the multimodal big data model matches the knowledge point association paths with the high-frequency comprehension difficulties, and for each knowledge point association path, marks all comprehension difficulties with high student feedback frequency in that path, ensuring that the key difficulties of each path can be accurately located; finally, it organizes the relationship between knowledge point association paths and corresponding high-frequency comprehension difficulties into a table to generate a path difficulty table.

[0088] Step 204: Based on the knowledge point error table, the ability difficulty table, and the path difficulty table, determine the test questions and corresponding question explanations to generate initial teaching content. The test questions include basic verification questions, error avoidance questions, and comprehensive application questions.

[0089] In this step, the basic verification questions refer to test questions used to examine students' mastery of the basic concepts of core knowledge points; the error avoidance questions refer to test questions used to help students identify and avoid common mistakes in core knowledge points; the comprehensive application questions refer to test questions used to examine students' ability to comprehensively apply multiple core knowledge points to solve problems; the initial teaching content refers to the teaching content generated for the first time, which includes test questions and corresponding question explanations; and the question explanations refer to the content that provides solutions, explanations of thought processes, and explanations of difficulties for the test questions.

[0090] The embodiments of this application achieve a deep integration of initial teaching content with knowledge point characteristics, ability requirements, and student feedback, thereby improving the accuracy and practicality of the initial teaching content and laying a high-quality foundation for the revision and optimization of subsequent teaching content.

[0091] This application provides a specific embodiment. Step 204 involves determining test questions and corresponding question analyses based on the knowledge point error table, the ability difficulty table, and the path difficulty table to generate initial teaching content. The test questions include basic verification questions, error avoidance questions, and comprehensive application questions, specifically including the following steps:

[0092] Step 211: Based on the knowledge point error table, determine the examination direction for each core knowledge point. The examination direction includes the direction of basic concept verification, the direction of high-frequency error avoidance, and the direction of comprehensive application of multiple knowledge points.

[0093] In this step, the assessment focus is on the specific emphasis when designing test questions for core knowledge points; the basic concept verification focus is on verifying students' mastery of the basic definitions and properties of core knowledge points; the high-frequency error avoidance focus is on guiding students to identify and avoid common errors in core knowledge points; and the multi-knowledge point comprehensive application focus is on assessing students' ability to comprehensively apply multiple related core knowledge points to solve practical problems.

[0094] In this embodiment, based on the knowledge point error table, the error types and frequencies corresponding to each core knowledge point are analyzed. If the error type of a core knowledge point is mainly confusion of basic concepts, with few high-frequency errors and no errors related to multiple knowledge points, the examination direction of the core knowledge point is determined to be the direction of basic concept verification. If the high-frequency error type of a core knowledge point is clear, such as calculation error or logical error, and the error frequency is high, the examination direction of the core knowledge point is determined to be the direction of high-frequency error avoidance. If a core knowledge point needs to be combined with other related knowledge points to solve the problem, and there are errors in the application of multiple knowledge points in the historical errors, the examination direction of the core knowledge point is determined to be the direction of comprehensive application of multiple knowledge points. Each core knowledge point corresponds to a clear examination direction.

[0095] Step 212: Based on the ability difficulty table and the path difficulty table, generate basic verification questions for the core knowledge points of the basic concept verification direction, generate error avoidance questions for the core knowledge points of the high-frequency error avoidance direction, and generate comprehensive application questions for the core knowledge points of the multi-knowledge point comprehensive application direction.

[0096] In this embodiment, based on the ability difficulty table and the path difficulty table, corresponding test questions are generated for the core knowledge points of different examination directions. For core knowledge points in the basic concept verification direction, the basic verification questions, such as definition restatement and simple judgment, are generated by referring to the basic difficulty range corresponding to the core knowledge point in the ability difficulty table, ensuring that the difficulty of the questions matches the basic cognitive level. For core knowledge points in the high-frequency error avoidance direction, error avoidance questions containing typical error options or error scenarios are generated by combining the high-frequency error types in the knowledge point error table and referring to the corresponding difficulty range in the ability difficulty table, guiding students to identify errors. For core knowledge points in the multi-knowledge point comprehensive application direction, comprehensive application questions involving multiple related knowledge points are generated based on the knowledge point association paths in the path difficulty table and referring to the corresponding advanced or high-level difficulty range in the ability difficulty table, to examine comprehensive application ability. The generated basic verification questions, error avoidance questions, and comprehensive application questions together constitute the test question set.

[0097] Step 213: Based on the difficulty range corresponding to each core knowledge point in the difficulty table, determine the analysis depth corresponding to each test question; based on the high-frequency understanding difficulty nodes corresponding to the knowledge point association paths in the path difficulty table, determine the important difficulty explanation content corresponding to each test question; based on the error type corresponding to each core knowledge point in the knowledge point error table, determine the error prompt content corresponding to each question; integrate the analysis depth, the important difficulty explanation content, and the error prompt content to obtain the analysis of the questions to be adjusted, wherein each test question includes basic verification questions, error avoidance questions, and comprehensive application questions.

[0098] In this step, "analysis depth" refers to the level of detail in the explanation of the test questions and knowledge points in the question analysis; "high-frequency difficult points" refers to specific knowledge points or connections between knowledge points that students report as difficult to understand and that occur frequently; "important difficult point explanations" refers to the content that needs to be emphasized in the question analysis for high-frequency difficult points; "error prompts" refers to the content that needs to remind students to avoid common error types related to core knowledge points; and "question analysis to be adjusted" refers to the initial draft of the question analysis that integrates the analysis depth, important difficult point explanations, and error prompts, but whose matching of analysis depth and difficulty range has not yet been confirmed.

[0099] In this embodiment, for each test question, the depth of analysis is determined based on the difficulty range of the core knowledge point corresponding to the question in the difficulty table. Questions in the basic difficulty range correspond to basic analysis depth, questions in the intermediate difficulty range correspond to intermediate analysis depth, and questions in the advanced difficulty range correspond to advanced analysis depth. Based on the high-frequency difficult nodes in the path difficulty table related to the knowledge point of the question, the important difficult points to be explained in the analysis are determined. Based on the error type of the core knowledge point of the question in the knowledge point error table, the error prompts to be included in the analysis are determined. The determined analysis depth, important difficult points, and error prompts are integrated to generate a corresponding question analysis to be adjusted for each test question. All question analysis to be adjusted constitutes a set of questions analysis to be adjusted.

[0100] Step 214: Calculate the matching degree between the analysis depth of the question to be adjusted and the difficulty range corresponding to each core knowledge point in the ability difficulty table. When the matching degree is lower than the preset matching threshold, adjust the analysis depth of the question to be adjusted to obtain the question analysis. Combine the questions to generate the initial teaching content.

[0101] In this step, the preset matching threshold refers to a pre-set critical value used to determine whether the analysis depth of the question to be adjusted matches the difficulty range of the core knowledge points.

[0102] In this embodiment, for each question analysis to be adjusted, the matching degree between its analysis depth and the difficulty range of the corresponding core knowledge point is calculated. The calculation formula is: Matching degree = Number of dimensions in which the analysis depth of the question analysis to be adjusted meets the difficulty range requirements ÷ Total number of dimensions for analysis depth evaluation. The total dimensions for analysis depth evaluation include step detail dimension, completeness of thought dimension, and degree of expansion dimension. The calculated matching degree is compared with a preset matching threshold. If the matching degree is greater than or equal to the preset matching threshold, it means that the analysis depth is suitable for the difficulty range, and the question analysis to be adjusted is directly determined as the question analysis. If the matching degree is lower than the preset matching threshold, the analysis depth is adjusted, and the matching degree is recalculated until the preset matching threshold is met, thus obtaining the question analysis. All question analyses are combined with the corresponding test questions to generate initial teaching content.

[0103] This application's embodiments address the combined shortcomings of existing technologies, such as vague testing directions, mismatched difficulty between questions and explanations, and lack of targeted difficulty points and error prompts in explanations. It achieves precise matching between test questions and assessment needs, and between question explanations and learning needs, thereby improving the relevance and effectiveness of initial teaching content.

[0104] This application provides a specific embodiment. Step 103 involves constructing a teaching knowledge graph based on the authoritative educational data, and calculating the fit between the initial teaching content and the teaching knowledge graph based on the hierarchical relationship of knowledge points and course objectives in the teaching knowledge graph. This specifically includes the following steps:

[0105] Step 301: Using the core knowledge points in the authoritative educational data as nodes, the knowledge point association paths in the knowledge point hierarchical relationship data as edges between nodes, and the ability cultivation requirements of the course objectives as attributes of the corresponding nodes, construct a teaching knowledge graph.

[0106] In this step, core knowledge points refer to the key knowledge points that students need to master, as specified in the curriculum syllabus from authoritative educational data.

[0107] In this embodiment, core knowledge points in authoritative educational data are used as nodes, and knowledge point association paths in hierarchical relationship data are used as edges between nodes. That is, core knowledge point nodes with hierarchical or logical relationships are connected by edges. The ability training requirements of the course objectives are used as attributes of the corresponding core knowledge point nodes. That is, each core knowledge point node is associated with its matching ability training requirements. According to the above correspondence of nodes, edges, and attributes, a structured teaching knowledge graph is constructed.

[0108] Step 302: Extract the knowledge points and assessment objectives from the test questions of the initial teaching content, and extract the analytical knowledge points from the question analysis of the initial teaching content.

[0109] In this step, the knowledge points being examined refer to the knowledge points actually tested by the test questions in the initial teaching content; the examination objectives refer to the ability development directions that the test questions are expected to achieve; and the knowledge points being analyzed refer to the knowledge points involved in the analysis process that are identified from the analysis of the questions in the initial teaching content.

[0110] In this embodiment of the application, the specific knowledge points tested by each question in the test questions of the initial teaching content are identified as the test knowledge points, and the direction of the ability to be cultivated by each question is identified as the test target; the knowledge points explained or referenced in each analysis process in the question analysis of the initial teaching content are identified as the analysis knowledge points.

[0111] Step 303: Based on the hierarchical relationship of knowledge points and course objectives in the teaching knowledge graph, calculate the first degree of overlap between the examination knowledge points and the core knowledge points, the second degree of overlap between the examination objectives and the ability development requirements, and the third degree of overlap between the actual path formed by the analytical knowledge points in the teaching knowledge graph and the path associated with the knowledge points.

[0112] In this step, the first overlap refers to the proportion of overlap between the knowledge points being examined and the core knowledge points in the teaching knowledge graph; the second overlap refers to the proportion of overlap between the examination objectives and the ability development requirements of the core knowledge point nodes in the teaching knowledge graph; the actual path refers to the path formed by the logical connections of the analytical knowledge points in the teaching knowledge graph; and the third overlap refers to the proportion of overlap between the actual path and the knowledge point connection paths in the teaching knowledge graph.

[0113] In this embodiment, based on the hierarchical relationship of knowledge points and course objectives in the teaching knowledge graph, the overlap degree in three dimensions is calculated. The overlap between the examined knowledge points and the core knowledge points in the teaching knowledge graph is calculated using the following formula: First overlap degree = Number of overlaps between the examined knowledge points and core knowledge points ÷ Total number of core knowledge points in the teaching knowledge graph; Second overlap degree = Number of consistency between the examination objectives and the corresponding ability development requirements of the core knowledge point nodes in the teaching knowledge graph ÷ Total number of ability development requirements in the teaching knowledge graph; Third overlap degree = Number of consistency between the examination objectives and ability development requirements ÷ Total number of ability development requirements in the teaching knowledge graph; Finally, the actual paths formed by knowledge points in the teaching knowledge graph are analyzed, and the number of correspondences between these actual paths and the corresponding knowledge point association paths in the teaching knowledge graph is calculated using the following formula: Third overlap degree = Number of correspondences between the actual paths and knowledge point association paths ÷ Total number of knowledge point association paths in the teaching knowledge graph.

[0114] Step 304: Based on preset dimension weights, perform a weighted summation of the first overlap, the second overlap, and the third overlap to obtain the fit between the initial teaching content and the teaching knowledge graph.

[0115] In this step, the preset dimension weights refer to the pre-set weight values ​​assigned to the first, second, and third degrees of overlap when calculating the degree of fit.

[0116] In this embodiment of the application, the first overlap, the second overlap, and the third overlap are weighted and summed based on preset dimension weights to obtain the fit between the initial teaching content and the teaching knowledge graph. The calculation formula is: Fit = First overlap × Preset first overlap weight + Second overlap × Preset second overlap weight + Third overlap × Preset third overlap weight.

[0117] This application's embodiments address the combined shortcomings of existing technologies, such as the lack of quantitative assessment of the matching degree between initial teaching content and knowledge system, and the ambiguity of assessment results. It achieves accurate quantitative judgment of the matching degree of initial teaching content, providing a clear basis for subsequent revision of the initial teaching content.

[0118] This application provides a specific embodiment. Step 104 involves comparing the fit with a preset threshold to correct the initial teaching content, resulting in corrected teaching content. This specifically includes the following steps:

[0119] Step 401: Compare the fit with the preset threshold. If the fit is greater than or equal to the preset threshold, the initial teaching content is used as the revised teaching content.

[0120] In this embodiment, the calculated initial teaching content is compared with the teaching knowledge graph and a preset threshold. When the degree of fit is greater than or equal to the preset threshold, it indicates that the initial teaching content meets the requirements of the teaching knowledge graph in terms of knowledge point coverage, ability goal alignment, and knowledge path matching. No adjustment is needed to the initial teaching content, and it can be directly determined as the corrected teaching content. This corrected teaching content can serve as the basis for generating optimized teaching content. If the degree of fit is less than the preset threshold, it is necessary to further examine the relationship between the first overlap, the second overlap, and the third overlap and the corresponding preset sub-thresholds, and carry out targeted corrections.

[0121] Step 402: If the degree of fit is less than the preset threshold and the first degree of overlap is less than the first preset sub-threshold, then supplement the test questions and question explanations corresponding to the missing knowledge points in the teaching knowledge graph until the first degree of overlap is greater than or equal to the first preset sub-threshold, and obtain the corrected teaching content.

[0122] In this step, the first preset sub-threshold refers to the pre-set critical value used to determine whether the first overlap degree meets the standard; the missing knowledge point refers to the core knowledge point contained in the teaching knowledge graph but not involved in the initial teaching content.

[0123] In this embodiment, when the fit is less than a preset threshold and the first overlap is less than a first preset sub-threshold, it indicates that the knowledge points examined in the initial teaching content do not fully cover the core knowledge points in the teaching knowledge graph, and there are omissions. In this case, it is necessary to first sort out the core knowledge points in the teaching knowledge graph and filter out the missing knowledge points not covered in the initial teaching content; then, for each missing knowledge point, design corresponding test questions and question explanations; integrate the supplemented test questions and question explanations into the initial teaching content, recalculate the first overlap until the first overlap is greater than or equal to the first preset sub-threshold, at which point the integrated teaching content is determined as the corrected teaching content.

[0124] Step 403: If the degree of fit is less than the preset threshold and the second degree of overlap is less than the second preset sub-threshold, then adjust the target test questions in the initial teaching content that do not match the examination objectives and ability cultivation requirements until the second degree of overlap is greater than or equal to the second preset sub-threshold, and obtain the corrected teaching content.

[0125] In this step, the second preset sub-threshold refers to the pre-set critical value used to determine whether the second overlap degree meets the standard; the target test item refers to the test item in the initial teaching content that does not match the ability cultivation requirements of the core knowledge point nodes in the teaching knowledge graph.

[0126] In this embodiment, when the fit is less than a preset threshold and the second overlap is less than a second preset sub-threshold, it indicates that the assessment objectives of some test questions in the initial teaching content do not align with the ability development requirements of the teaching knowledge graph. In this case, the target test questions in the initial teaching content must first be selected, and determined by comparing the assessment objectives of each test question with the ability development requirements of the corresponding core knowledge points. Then, for each target test question, its assessment direction is adjusted to ensure that the adjusted assessment objectives are consistent with the ability development requirements. The adjusted target test questions replace the original questions in the initial teaching content, and the second overlap is recalculated until the second overlap is greater than or equal to the second preset sub-threshold. At this point, the replaced teaching content is determined as the revised teaching content.

[0127] Step 404: If the degree of fit is less than the preset threshold and the third degree of overlap is less than the third preset sub-threshold, then correct the target question analysis in the initial teaching content where the actual path of question analysis and the knowledge point association path do not match, until the third degree of overlap is greater than or equal to the third preset sub-threshold, and obtain the corrected teaching content.

[0128] In this step, the third preset sub-threshold refers to the pre-set critical value used to determine whether the third overlap degree meets the standard; the target question analysis refers to the question analysis where the actual path formed by the knowledge points in the teaching knowledge graph in the initial teaching content does not match the knowledge point association path.

[0129] In this embodiment, when the fit is less than a preset threshold and the third overlap is less than a third preset sub-threshold, it indicates that the knowledge paths of some questions in the initial teaching content do not match the knowledge point association paths of the teaching knowledge graph. In this case, it is necessary to first filter out the target question analyses in the initial teaching content, and determine this by comparing the actual path of each analyzed knowledge point with the corresponding knowledge point association path; then, for each target question analysis, correct the order of knowledge point explanations or association logic in its analysis to ensure that the corrected actual path is consistent with the knowledge point association path; replace the original analysis in the initial teaching content with the corrected target question analysis, and recalculate the third overlap until the third overlap is greater than or equal to the third preset sub-threshold. At this point, the replaced teaching content is determined as the corrected teaching content.

[0130] The embodiments of this application solve the combined defects of the prior art, which lack a clear direction and are not accurate in modifying the initial teaching content. They achieve dimensional and accurate modification based on quantitative indicators, ensuring that the modified teaching content is highly matched with the teaching knowledge graph, and improving the standardization and applicability of the teaching content.

[0131] This application provides a specific embodiment. Step 105 involves selecting a prompt word framework from a preset prompt word template library, combining it with the revised teaching content, and using a multimodal large model to generate optimized teaching content. The prompt word framework includes expression norms and optimization directions adapted to the teaching scenario, specifically including the following steps:

[0132] Step 501: Analyze the teaching scenarios of the revised teaching content to obtain the core knowledge point types, test question difficulty levels, and question analysis length characteristics, so as to determine the teaching scenario type corresponding to the revised teaching content.

[0133] In this step, the core knowledge point type refers to the subject classification or knowledge domain category to which the core knowledge points in the revised teaching content belong; the test question difficulty level refers to the level of difficulty of the test questions in the revised teaching content; the question explanation length characteristics refer to the characteristics of the question explanations in the revised teaching content, which are divided according to the length of the text or the level of detail of the explanation; and the teaching scenario type refers to the scenario category determined based on the core knowledge point type, the test question difficulty level, and the question explanation length characteristics, which is adapted to specific teaching needs.

[0134] In this embodiment of the application, the teaching scenario of the revised teaching content is analyzed to identify the subject classification or knowledge domain to which each core knowledge point belongs and determine the type of core knowledge point; the difficulty level of the test questions is determined according to the depth of the test questions and the cognitive level requirements of the students; the length of the question analysis is determined according to the length of the text and the level of detail of the explanation; and the teaching scenario type corresponding to the revised teaching content is determined based on the combination relationship of the above core knowledge point types, the difficulty level of the test questions and the length of the question analysis.

[0135] Step 502: Extract the prompt word framework corresponding to the teaching scenario type from the preset prompt word template library.

[0136] In this embodiment of the application, a preset prompt word template library stores multiple prompt word frames corresponding to different teaching scenario types. The prompt word frame that perfectly matches the teaching scenario type is searched and extracted from the preset prompt word template library. The prompt word frame includes expression norms and optimization directions adapted to the current teaching scenario.

[0137] Step 503: Extract key content elements from the revised teaching content, and convert the key content elements into formatted content elements according to the expression specifications of the prompt word framework.

[0138] In this step, key content elements refer to the information selected from the revised teaching content that plays a core supporting role in generating the optimized teaching content; formatted content elements refer to the standardized information formed after the key content elements are formatted according to the expression specifications of the prompt word framework.

[0139] In this embodiment, the descriptions of core knowledge points, the stems of test questions, the core logic of question analysis, and the examination points corresponding to the ability development requirements are selected from the revised teaching content as key content elements. Referring to the expression specifications of the prompt word framework, the key content elements are formatted by replacing non-standard terms in the descriptions of core knowledge points and the examination points corresponding to the ability development requirements with standard terms, and adjusting the stems of test questions and the core logic of question analysis to expressions that conform to the scenario. The above conversion yields formatted content elements, which will be used for subsequent integration to generate prompt words.

[0140] Step 504: Integrate the optimization directions of the formatted content elements and the prompt word framework into prompt words, and optimize the revised teaching content using the multimodal large model based on the prompt words to generate preliminary optimized teaching content.

[0141] In this embodiment, the formatted content elements and the optimization direction of the prompt word framework are integrated, and the text is combined according to the structure of formatted content elements + optimization direction execution instructions to form prompt words for input into the multimodal large model. The prompt words are input into the multimodal large model, and the multimodal large model optimizes the corrected teaching content based on the formatted content elements and optimization direction instructions in the prompt words to generate preliminary optimized teaching content. This preliminary optimized teaching content will be used for subsequent verification of whether it meets the key optimization indicators.

[0142] Step 505: If the preliminarily optimized teaching content cannot meet the key optimization indicators in the prompt word framework, then modify the prompt words until the generated preliminarily optimized teaching content meets the key optimization indicators in the prompt word framework, thus obtaining the optimized teaching content.

[0143] In this step, the key optimization indicators refer to the pre-set standards in the prompt word framework used to judge whether the teaching content after preliminary optimization meets the optimization requirements.

[0144] In this embodiment, the preliminarily optimized teaching content is examined against the key optimization indicators in the prompt word framework to determine whether its expression conforms to the expression specifications of the prompt word framework, whether the optimization direction instructions are fully executed, and whether key information such as core knowledge points are completely retained. If the preliminarily optimized teaching content meets all key optimization indicators, it is directly identified as optimized teaching content. If the preliminarily optimized teaching content does not meet some or all of the key optimization indicators, the prompt words are modified for the unmet indicators, and the modified prompt words are re-input into the multimodal large model to generate new preliminarily optimized teaching content. The examination and modification operations are repeated until the generated preliminarily optimized teaching content meets all key optimization indicators and is identified as optimized teaching content.

[0145] The embodiments of this application solve the complex defects in the prior art, such as lack of scenario adaptability, unclear optimization direction, and difficulty in achieving optimization results, thereby improving the scenario adaptability and teaching practicality of the optimized teaching content.

[0146] This application provides a specific embodiment. Step 503 involves extracting key content elements from the revised teaching content and converting the key content elements into formatted content elements according to the expression specifications of the prompt word framework. This specifically includes the following steps:

[0147] Step 511: Extract the core knowledge point descriptions and test question stems corresponding to the test questions from the test questions of the revised teaching content, and extract the core logic of the question analysis corresponding to the test question stems from the question analysis of the revised teaching content.

[0148] In this step, the core knowledge point description refers to the specific definition, characteristics, or scope of application of the core knowledge point corresponding to each test question in the revised teaching content; the test question stem refers to the text describing the test question in the revised teaching content; and the core logic of the question analysis refers to the key reasoning steps, causal relationships, or problem-solving framework in the question analysis process in the revised teaching content.

[0149] In this embodiment, for each test question in the revised teaching content, the core knowledge point description directly related to the test question is identified and extracted from the question stem description and examination focus of the test question; at the same time, the complete question statement of the test question is extracted as the test question stem; for each test question stem, the corresponding question analysis is found from the revised teaching content, and the key reasoning links and thought framework supporting the solution conclusion are sorted out and extracted from the analysis text as the core logic of the question analysis, ensuring that each test question corresponds to a unique core knowledge point description and test question stem, and that each test question stem corresponds to a unique core logic of the question analysis.

[0150] Step 512: Based on the core logic of the question analysis and combined with the ability development requirements, determine the angle from which the test questions examine the ability development requirements, and take the examination angle as the examination point corresponding to the ability development requirements.

[0151] In this step, the examination angle refers to the specific direction in which the test questions, through their specific stems and analytical logic, embody the requirements for ability development; the examination points corresponding to the requirements for ability development refer to the specific examination objects that directly correspond to the requirements for ability development after the examination angle has been clarified.

[0152] In this embodiment, the core logic of each question is analyzed, the problem-solving ability requirements reflected in the logic are analyzed, and the ability development requirements associated with the core knowledge points corresponding to the question in the teaching knowledge graph are combined to determine the specific direction of the ability development requirements of the test question, i.e., the examination angle. The examination angle is further clarified into quantifiable and verifiable specific content, which serves as the examination point corresponding to the ability development requirements, ensuring that each test question corresponds to a unique examination point corresponding to the ability development requirements.

[0153] Step 513: Replace the non-standard terms in the core knowledge point description and the non-standard terms in the examination points with standard terms that conform to the expression specifications of the prompt word framework, to obtain the formatted knowledge point description and the formatted examination points.

[0154] In this step, non-standard terms refer to terms in the core knowledge point descriptions or assessment points corresponding to ability development requirements that do not conform to the expression specifications of the prompt word framework; standard terms refer to standard terms that conform to the expression specifications of the prompt word framework, are industry-wide common terms, or are agreed upon in teaching scenarios; formatted knowledge point descriptions refer to the standardized text formed after replacing non-standard terms in the core knowledge point descriptions with standard terms; formatted assessment points refer to the standardized text formed after replacing non-standard terms in the assessment points corresponding to ability development requirements with standard terms.

[0155] In this embodiment, the terminology usage standard is extracted from the expression specification of the prompt word framework. The core knowledge point descriptions and the corresponding examination points for ability development requirements are examined one by one against the standard, and non-standard terms are screened out. The non-standard terms in the core knowledge point descriptions are replaced with the standard terms specified in the terminology usage standard to form a formatted knowledge point description. Similarly, the non-standard terms in the examination points corresponding to ability development requirements are replaced with standard terms to form formatted examination points, ensuring that both the formatted knowledge point descriptions and formatted examination points comply with the terminology usage standard.

[0156] Step 514: Adjust the format of the test question stem and the core logic of the question analysis to conform to the language style requirements in the expression specification of the prompt word framework, so as to obtain the formatted question stem and the formatted core logic.

[0157] In this step, formatting the question stem refers to adjusting the format of the test question stem to conform to the language style requirements of the prompt word framework expression specification; formatting the core logic refers to adjusting the format of the core logic of the question analysis to conform to the language style requirements.

[0158] In this embodiment, language style requirements are extracted from the expression specifications of the prompt word framework. The test question stems are then adjusted according to these requirements. If the language style requirement is conciseness and clarity, redundant expressions in the question stem are removed. If the requirement is logical clarity, the order of the conditions in the question stem is adjusted to form a formatted question stem. At the same time, the core logic of the question analysis is adjusted according to the same language style requirements. If the requirement is step-by-step, the core logic is broken down into orderly step-by-step expressions. If the requirement is logical organization, logical connectors are used to organize the reasoning relationships to form a formatted core logic.

[0159] Step 515: Integrate the formatted knowledge point description, the formatted examination points, the formatted question stem, and the formatted core logic to obtain the formatted content elements.

[0160] In this embodiment of the application, a set of elements corresponding to each test question are combined to ensure that the logical relationship between each element is coherent; the combined elements corresponding to all test questions are aggregated to form a set containing all standardized elements, which is the formatted content element.

[0161] This application's embodiments address the combined shortcomings of existing technologies, such as inaccurate extraction of key content, inconsistent element formats, and poor adaptability to the prompt word framework. It achieves standardized processing of key content elements, providing high-quality input for subsequent prompt word generation and multimodal large-scale model optimization, and ensuring the scenario adaptability and standardization of the optimized teaching content.

[0162] Figure 3 This is a schematic diagram illustrating a specific implementation of an optimization system for generating teaching content from a large model, as provided in this application. (Refer to...) Figure 3 The system may include:

[0163] The acquisition module 21 is used to acquire authoritative educational data and teacher and student feedback data in the teaching scenario. The authoritative educational data includes course teaching outlines, course objectives and knowledge point hierarchical relationship data. The teacher and student feedback data includes teachers' error annotation data on historically generated teaching content and students' feedback data on the difficulty of understanding historically generated teaching content.

[0164] The association module 22 is used to use a multimodal large model to associate and match the authoritative educational data and the teacher and student feedback data to generate initial teaching content, which includes test questions and question explanations.

[0165] The construction module 23 is used to construct a teaching knowledge graph based on the authoritative educational data, and to calculate the fit between the initial teaching content and the teaching knowledge graph based on the hierarchical relationship of knowledge points and course objectives in the teaching knowledge graph.

[0166] The correction module 24 is used to compare the fit with a preset threshold to correct the initial teaching content and obtain the corrected teaching content.

[0167] The optimization module 25 is used to select a prompt word framework from a preset prompt word template library, combine it with the revised teaching content, and use a multimodal large model to generate optimized teaching content. The prompt word framework includes expression norms and optimization directions adapted to the teaching scenario.

[0168] The confirmation module 26 is used to confirm the matching degree between the optimized teaching content and the course syllabus and the student comprehension difficulty score based on the teacher and student feedback data and the authoritative educational data, and to iteratively optimize the optimized teaching content in combination with preset standards.

[0169] An optimization system for generating teaching content using a large model, as described in this application, is used to implement the aforementioned optimization method for generating teaching content using a large model. Therefore, the specific implementation of the optimization system for generating teaching content using a large model can be found in the embodiment section of the optimization method for generating teaching content using a large model described above. The specific implementation can be referred to the description of the corresponding embodiments, and will not be repeated here.

[0170] This application also provides an electronic device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of the optimization method for generating teaching content from a large model as described above.

[0171] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the optimization method for generating teaching content from a large model as described above.

[0172] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory, random access memory, portable hard drives, magnetic disks, or optical disks.

[0173] Embodiments of the present invention also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above-described embodiments of the optimization method for generating teaching content from a large model.

[0174] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0175] The above provides a detailed description of the optimization method and system for generating teaching content from a large model, as provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. An optimized method for generating teaching content using a large model, characterized in that, include: Acquire authoritative educational data and teacher / student feedback data in teaching scenarios. The authoritative educational data includes course syllabus, course objectives, and hierarchical relationship data of knowledge points. The teacher / student feedback data includes teachers' error annotation data on historically generated teaching content and students' feedback data on the difficulty of understanding historically generated teaching content. Using a multimodal large model, the authoritative educational data and the teacher and student feedback data are correlated and matched to generate initial teaching content, which includes test questions and question explanations. Based on the authoritative educational data, a teaching knowledge graph is constructed. Based on the hierarchical relationship of knowledge points and course objectives in the teaching knowledge graph, the degree of fit between the initial teaching content and the teaching knowledge graph is calculated. The fit is compared with a preset threshold to correct the initial teaching content and obtain the corrected teaching content. Select a prompt word framework from the preset prompt word template library, combine it with the revised teaching content, and use a multimodal large model to generate optimized teaching content. The prompt word framework includes expression norms and optimization directions adapted to the teaching scenario. Based on the teacher and student feedback data and the authoritative educational data, the matching degree between the optimized teaching content and the course syllabus and the student comprehension difficulty score are confirmed. Combined with the preset standards, the optimized teaching content is iteratively optimized.

2. The method according to claim 1, characterized in that, Using a multimodal large model, the authoritative educational data and the teacher-student feedback data are correlated and matched to generate initial teaching content. This initial teaching content includes test questions and question explanations, including: Using a multimodal large model, the core knowledge points of the course syllabus are matched with the error types of the error annotation data to generate a knowledge point error table; The competency development requirements of the course objectives are matched with the difficulty range of the comprehension difficulty feedback data to generate a competency difficulty table; The knowledge point association paths in the knowledge point hierarchy relationship data are matched with the high-frequency comprehension difficulties in the comprehension difficulty feedback data to generate a path difficulty table; Based on the knowledge point error table, the ability difficulty table, and the path difficulty table, test questions and corresponding question explanations are determined to generate initial teaching content. The test questions include basic verification questions, error avoidance questions, and comprehensive application questions.

3. The method according to claim 2, characterized in that, Based on the knowledge point error table, the ability difficulty table, and the path difficulty table, test questions and corresponding question explanations are determined to generate initial teaching content. The test questions include basic verification questions, error avoidance questions, and comprehensive application questions, including: Based on the knowledge point error table, the examination direction for each core knowledge point is determined. The examination direction includes the direction of basic concept verification, the direction of high-frequency error avoidance, and the direction of comprehensive application of multiple knowledge points. Based on the aforementioned ability difficulty table and the aforementioned path difficulty table, basic verification questions are generated for the core knowledge points of the basic concept verification direction, error avoidance questions are generated for the core knowledge points of the high-frequency error avoidance direction, and comprehensive application questions are generated for the core knowledge points of the multi-knowledge point comprehensive application direction. Based on the difficulty range corresponding to each core knowledge point in the difficulty table, the analysis depth corresponding to each test question is determined. Based on the high-frequency understanding difficulty nodes corresponding to the knowledge point association paths in the path difficulty table, the important difficulty explanation content corresponding to each test question is determined. Based on the error type corresponding to each core knowledge point in the knowledge point error table, the error prompt content corresponding to each question is determined. By integrating the analysis depth, the important difficulty explanation content, and the error prompt content, the analysis of the questions to be adjusted is obtained. Among them, each test question includes basic verification questions, error avoidance questions, and comprehensive application questions. Calculate the matching degree between the analysis depth of the question to be adjusted and the difficulty range corresponding to each core knowledge point in the ability difficulty table. When the matching degree is lower than the preset matching threshold, adjust the analysis depth of the question to be adjusted to obtain the question analysis. Combine the questions with the test questions to generate the initial teaching content.

4. The method according to claim 1, characterized in that, Based on the authoritative educational data, a teaching knowledge graph is constructed. Based on the hierarchical relationships of knowledge points and course objectives within the teaching knowledge graph, the fit between the initial teaching content and the teaching knowledge graph is calculated, including: Using the core knowledge points in the authoritative educational data as nodes, the knowledge point association paths in the knowledge point hierarchical relationship data as edges between nodes, and the ability cultivation requirements of the course objectives as attributes of the corresponding nodes, a teaching knowledge graph is constructed. Extract the knowledge points and assessment objectives from the test questions of the initial teaching content, and extract the analytical knowledge points from the question analysis of the initial teaching content; Based on the hierarchical relationship of knowledge points and course objectives in the teaching knowledge graph, the first degree of overlap between the examination knowledge points and the core knowledge points, the second degree of overlap between the examination objectives and the ability cultivation requirements, and the third degree of overlap between the actual path formed by the analytical knowledge points in the teaching knowledge graph and the path associated with the knowledge points are calculated. Based on preset dimension weights, the first overlap, the second overlap, and the third overlap are weighted and summed to obtain the fit between the initial teaching content and the teaching knowledge graph.

5. The method according to claim 1, characterized in that, The fit is compared with a preset threshold to correct the initial teaching content, resulting in corrected teaching content, including: The fit is compared with a preset threshold. If the fit is greater than or equal to the preset threshold, the initial teaching content is used as the revised teaching content. If the degree of fit is less than the preset threshold and the first degree of overlap is less than the first preset sub-threshold, then supplement the test questions and question analysis corresponding to the missing knowledge points in the teaching knowledge graph until the first degree of overlap is greater than or equal to the first preset sub-threshold, and obtain the corrected teaching content. If the degree of fit is less than the preset threshold and the second degree of overlap is less than the second preset sub-threshold, then the target test questions in the initial teaching content that do not match the examination objectives and ability cultivation requirements are adjusted until the second degree of overlap is greater than or equal to the second preset sub-threshold, and the corrected teaching content is obtained. If the degree of fit is less than a preset threshold, and the third degree of overlap is less than a third preset sub-threshold, then the target question analysis in the initial teaching content is corrected to prevent the actual path of question analysis and the knowledge point association path from matching, until the third degree of overlap is greater than or equal to the third preset sub-threshold, and the corrected teaching content is obtained.

6. The method according to claim 1, characterized in that, Select a prompt word framework from a pre-set prompt word template library, combine it with the revised teaching content, and use a multimodal large model to generate optimized teaching content. The prompt word framework includes expression norms and optimization directions adapted to the teaching scenario, including: The teaching scenarios of the revised teaching content are analyzed to obtain the core knowledge point types, test question difficulty levels, and question analysis length characteristics, so as to determine the teaching scenario type corresponding to the revised teaching content. Extract the prompt word framework corresponding to the teaching scenario type from the preset prompt word template library; Key content elements are extracted from the revised teaching content, and the key content elements are formatted according to the expression specifications of the prompt word framework to obtain formatted content elements. The optimization directions of the formatted content elements and the prompt word framework are integrated into prompt words. Based on the prompt words, the modified teaching content is optimized using the multimodal large model to generate preliminary optimized teaching content. If the preliminarily optimized teaching content cannot meet the key optimization indicators in the prompt word framework, the prompt words are modified until the generated preliminarily optimized teaching content meets the key optimization indicators in the prompt word framework, thus obtaining the optimized teaching content.

7. The method according to claim 6, characterized in that, Key content elements are extracted from the revised teaching content, and formatted according to the expression specifications of the prompt word framework to obtain formatted content elements, including: Extract the core knowledge point descriptions and test question stems corresponding to the test questions from the test questions of the revised teaching content, and extract the core logic of the question analysis corresponding to the test question stems from the question analysis of the revised teaching content. Based on the core logic of the question analysis and combined with the ability development requirements, the angle from which the test questions examine the ability development requirements is determined, and the angle of examination is taken as the examination point corresponding to the ability development requirements. Replace the non-standard terms in the description of the core knowledge points and the non-standard terms in the examination points with standard terms that conform to the expression specifications of the prompt word framework to obtain formatted knowledge point descriptions and formatted examination points; The format of the test question stem and the core logic of the question analysis are adjusted to conform to the language style requirements in the expression specification of the prompt word framework, so as to obtain the formatted question stem and the formatted core logic. The formatted knowledge point description, the formatted examination points, the formatted question stem, and the formatted core logic are integrated to obtain the formatted content elements.

8. An optimization system for generating teaching content from a large model, characterized in that, include: The acquisition module is used to acquire authoritative educational data and teacher and student feedback data in the teaching scenario. The authoritative educational data includes course teaching outline, course objectives and knowledge point hierarchical relationship data. The teacher and student feedback data includes teachers' error annotation data on historically generated teaching content and students' feedback data on the difficulty of understanding historically generated teaching content. The association module is used to use a multimodal large model to associate and match the authoritative educational data and the teacher and student feedback data to generate initial teaching content, which includes test questions and question explanations. The module is used to construct a teaching knowledge graph based on the authoritative educational data, and to calculate the fit between the initial teaching content and the teaching knowledge graph based on the hierarchical relationship of knowledge points and course objectives in the teaching knowledge graph. The correction module is used to compare the fit with a preset threshold to correct the initial teaching content and obtain the corrected teaching content. The optimization module is used to select a prompt word framework from a preset prompt word template library, combine it with the revised teaching content, and use a multimodal large model to generate optimized teaching content. The prompt word framework includes expression norms and optimization directions adapted to the teaching scenario. The confirmation module is used to confirm the matching degree between the optimized teaching content and the course syllabus and the student comprehension difficulty score based on the teacher and student feedback data and the authoritative educational data, and to iteratively optimize the optimized teaching content in combination with preset standards.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of an optimized method for generating instructional content from a large model as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, enables an optimized method for generating teaching content from a large model as described in any one of claims 1 to 7.

Images