CN-122019804-A - Optimization method and system for generating teaching content by large model

CN122019804ACN 122019804 ACN122019804 ACN 122019804ACN-122019804-A

Abstract

The application provides an optimization method and system for generating teaching contents by a large model, which relate to the technical field of education, and the application acquires authoritative education data and teacher feedback data in a teaching scene; the method comprises the steps of generating initial teaching contents by using a multi-mode large model, constructing a teaching knowledge graph according to authoritative education data, calculating the fit degree of the initial teaching contents and the teaching knowledge graph based on the hierarchical relation of knowledge points in the teaching knowledge graph and a course target, comparing the fit degree with a preset threshold value to obtain corrected teaching contents, selecting a prompting word frame from a preset prompting word template library, combining the corrected teaching contents to generate optimized teaching contents, confirming the fit degree of the optimized teaching contents and the course teaching outline and the student understanding difficulty score according to teacher feedback data and authoritative education data, combining the preset standard to achieve self-adaptive generation and iterative optimization of the teaching contents, and improving the fit degree of the teaching contents and the course target and the student understanding degree.

Inventors

DU JIANBIN
GUO PANPAN
ZHANG JIANCONG
FENG XIYE
JIAN MUWEI
LIU JIANCHAO
ZHANG LILI
DU LIN
Nie Panpan
HOU NINGNING

Assignees

齐鲁师范学院

Dates

Publication Date: 20260512
Application Date: 20260202

Claims (10)

1. An optimization method for generating teaching contents by using a large model is characterized by comprising the following steps: the method comprises the steps of acquiring authoritative education data and teacher-student feedback data in a teaching scene, wherein the authoritative education data comprise course teaching outline, course targets and knowledge point hierarchical relation data, and the teacher-student feedback data comprise error marking data of a teacher on history generation teaching content and understanding difficulty feedback data of students on history generation teaching content; performing association matching on the authoritative education data and the teacher-student feedback data by using a multi-mode large model to generate initial teaching contents, wherein the initial teaching contents comprise test questions and question analysis; Constructing a teaching knowledge graph according to the authoritative education data, and calculating the degree of agreement between the initial teaching content and the teaching knowledge graph based on the knowledge point hierarchical relationship and the course target in the teaching knowledge graph; comparing the fit degree with a preset threshold value to correct the initial teaching content, and obtaining corrected teaching content; Selecting a prompting word frame from a preset prompting word template library, combining the corrected teaching content, and generating optimized teaching content by utilizing a multi-mode large model, wherein the prompting word frame comprises a description specification and an optimization direction which are matched with a teaching scene; And confirming the matching degree of the optimized teaching content and the course teaching outline and the student understanding difficulty score according to the teacher-student feedback data and the authoritative education data, and carrying out iterative optimization of the optimized teaching content by combining with a preset standard.
2. The method of claim 1, wherein the authoritative educational data and the teacher-student feedback data are associated and matched by using a multi-modal big model to generate initial teaching content, the initial teaching content comprising test questions and question parsing, comprising: matching the core knowledge points of the course teaching outline with the error types of the error labeling data by using a multi-mode big model to generate a knowledge point error table; Matching the capability culture requirement of the course target with the difficulty interval of the understanding difficulty feedback data to generate a capability difficulty table; matching the knowledge point association path of the knowledge point hierarchical relation data with the high-frequency understanding difficulty of the understanding difficulty feedback data to generate a path difficulty table; Based on the knowledge point error table, the capability difficulty table and the path difficulty table, test questions and corresponding question analyses are determined to generate initial teaching contents, wherein the test questions comprise basic verification questions, error avoidance questions and comprehensive application questions.
3. The method of claim 2, wherein determining test questions and corresponding question resolution based on the knowledge point error table, the capability difficulty table, and the path difficulty table to generate initial teaching content, the test questions including a base verification question, an error avoidance question, and a comprehensive application question comprises: determining the checking direction of each core knowledge point based on the knowledge point error table, wherein the checking direction comprises a basic concept verification direction, a high-frequency error avoidance direction and a multi-knowledge point comprehensive application direction; Based on the capability difficulty table and the path difficulty table, generating basic verification questions for core knowledge points in which an examination direction is a basic concept verification direction, generating error avoidance questions for core knowledge points in which the examination direction is a high-frequency error avoidance direction, and generating comprehensive application questions for core knowledge points in which the examination direction is a multi-knowledge point comprehensive application direction; Determining analysis depth corresponding to each test question based on a difficulty interval corresponding to each core knowledge point in the capability difficulty table, determining important difficulty explanation content corresponding to each test question based on a high-frequency understanding difficulty node corresponding to a knowledge point association path in the path difficulty table, determining error prompt content corresponding to each question based on an error type corresponding to each core knowledge point in the knowledge point error table, and integrating the analysis depth, the important difficulty explanation content and the error prompt content to obtain question analysis to be adjusted, wherein each test question comprises a basic verification question, an error avoidance question and a comprehensive application question; And calculating the analysis depth of the question analysis to be adjusted and the matching degree of the difficulty interval corresponding to each core knowledge point in the capability difficulty table, and adjusting the analysis depth of the question analysis to be adjusted when the matching degree is lower than a preset matching threshold value to obtain the question analysis, and generating initial teaching contents by combining all test questions.
4. The method of claim 1, wherein constructing a teaching knowledge graph from the authoritative educational data, and calculating the fitness of the initial teaching content and the teaching knowledge graph based on knowledge point hierarchical relationships and course objectives in the teaching knowledge graph, comprises: taking a core knowledge point in the authoritative education data as a node, taking a knowledge point association path of the knowledge point hierarchical relation data as an edge between the nodes, and constructing a teaching knowledge map by taking the capability culture requirement of a course target as an attribute of a corresponding node; Extracting examination knowledge points and examination targets from examination questions of the initial teaching content, and extracting analysis knowledge points from question analysis of the initial teaching content; Calculating a first overlapping degree of the examined knowledge points and the core knowledge points, a second overlapping degree of the examined objects and the capability culture requirements and a third overlapping degree of a path related to the knowledge points, wherein the actual path is formed by the analyzed knowledge points in the teaching knowledge map based on a knowledge point hierarchical relation and a course object in the teaching knowledge map; and carrying out weighted summation on the first contact ratio, the second contact ratio and the third contact ratio based on preset dimension weights to obtain the contact ratio of the initial teaching content and the teaching knowledge graph.
5. The method of claim 1, wherein comparing the fitness to a preset threshold to modify the initial teaching content to obtain modified teaching content comprises: comparing the fit degree with a preset threshold value, and taking the initial teaching content as the corrected teaching content if the fit degree is larger than or equal to the preset threshold value; If the degree of coincidence is smaller than a preset threshold value and the first degree of coincidence is smaller than a first preset sub-threshold value, supplementing test questions and question analysis corresponding to the missing knowledge points in the teaching knowledge graph until the first degree of coincidence is larger than or equal to the first preset sub-threshold value, and obtaining corrected teaching contents; If the fit degree is smaller than a preset threshold value and the second fit degree is smaller than a second preset sub-threshold value, adjusting a target test question of which the examination target is not matched with the capability culture requirement in the initial teaching content until the second fit degree is larger than or equal to the second preset sub-threshold value, and obtaining corrected teaching content; And if the fitting degree is smaller than the preset threshold value and the third fitting degree is smaller than a third preset sub-threshold value, correcting target topic analysis of which the actual topic analysis path and the knowledge point association path are not matched in the initial teaching content until the third fitting degree is larger than or equal to the third preset sub-threshold value, and obtaining corrected teaching content.
6. The method of claim 1, wherein selecting a cue word frame from a preset cue word template library, and generating optimized teaching content by using a multi-modal large model in combination with the corrected teaching content, wherein the cue word frame includes a description specification and an optimization direction adapted to a teaching scene, and the method comprises: Analyzing the teaching scene of the corrected teaching content to obtain a core knowledge point type, a test question difficulty level and a question analysis spread characteristic so as to determine the teaching scene type corresponding to the corrected teaching content; Extracting a prompting word frame corresponding to the teaching scene type from a preset prompting word template library; Extracting key content elements from the corrected teaching content, and performing format conversion on the key content elements according to the expression specification of the prompt word frame to obtain formatted content elements; Integrating the formatted content elements and the optimization direction of the prompt word frame into prompt words, so as to optimize the corrected teaching content by utilizing the multi-mode large model according to the prompt words, and generating preliminary optimized teaching content; And if the primarily optimized teaching content can not meet the key optimization index in the prompt word frame, modifying the prompt word until the generated primarily optimized teaching content meets the key optimization index in the prompt word frame, and obtaining the optimized teaching content.
7. The method of claim 6, wherein extracting key content elements from the modified tutorial content and converting the key content elements to format according to the presentation specification of the alert frame to obtain formatted content elements comprises: Extracting core knowledge point descriptions and test question stems corresponding to the test questions from the test questions of the corrected teaching contents, and extracting question analysis core logic corresponding to the test question stems from the question analysis of the corrected teaching contents; Based on the topic analysis core logic and the capability culture requirement, determining an examination angle of the test topic on the capability culture requirement, and taking the examination angle as an examination point corresponding to the capability culture requirement; Respectively replacing non-standard terms in the core knowledge point description and non-standard terms in the examination points with terms in expression specifications conforming to the prompt word frame to use standard specification terms to obtain formatted knowledge point description and formatted examination points; the format of the test question stem and the question analysis core logic is adjusted to be in accordance with the language style requirements in the expression specification of the prompt word frame, and formatted question stem and formatted core logic are obtained; And integrating the formatting knowledge point description, the formatting examination point, the formatting stem and the formatting core logic to obtain a formatting content element.
8. An optimization system for generating teaching content by a large model, comprising: The system comprises an acquisition module, a learning module and a learning module, wherein the acquisition module is used for acquiring authoritative education data and teacher-student feedback data in a teaching scene, the authoritative education data comprises a course teaching outline, a course target and knowledge point hierarchical relation data, and the teacher-student feedback data comprises error marking data of a teacher on history generation teaching content and understanding difficulty feedback data of students on history generation teaching content; The association module is used for carrying out association matching on the authoritative education data and the teacher-student feedback data by utilizing a multi-mode large model to generate initial teaching contents, wherein the initial teaching contents comprise test questions and question analysis; The construction module is used for constructing a teaching knowledge graph according to the authoritative education data, and calculating the degree of fit between the initial teaching content and the teaching knowledge graph based on the knowledge point hierarchical relationship and the course target in the teaching knowledge graph; The correction module is used for comparing the fit degree with a preset threshold value so as to correct the initial teaching content and obtain corrected teaching content; the optimizing module is used for selecting a prompting word frame from a preset prompting word template library, generating optimized teaching contents by utilizing a multi-mode large model in combination with the corrected teaching contents, wherein the prompting word frame comprises a description specification and an optimizing direction which are matched with a teaching scene; And the confirmation module is used for confirming the matching degree of the optimized teaching content and the course teaching outline and the student understanding difficulty score according to the teacher feedback data and the authoritative education data, and carrying out iterative optimization of the optimized teaching content by combining with a preset standard.
9. An electronic device, comprising: A memory for storing a computer program; A processor for implementing the steps of a method for optimizing the generation of teaching content for a large model according to any of claims 1 to 7 when executing said computer program.
10. A computer readable storage medium, wherein a computer program is stored in the computer readable storage medium, and when executed by a processor, the computer program is capable of implementing a method for optimizing the generation of teaching content by a large model according to any one of claims 1 to 7.

Description

Optimization method and system for generating teaching content by large model Technical Field The application relates to the technical field of education, in particular to an optimization method and system for generating teaching contents by using a large model. Background Along with the deep digital transformation of education, the large model is increasingly widely applied to the generation of teaching contents, and particularly when complex abstract knowledge is processed, the presentation form of the contents needs to be dynamically adjusted according to the understanding state of students so as to improve the learning efficiency. Currently, the existing mainstream scheme evaluates the understanding degree of the large model generated content by analyzing the interactive behavior data of students in the learning process, and when the understanding bottleneck is detected, a 3D modeling engine is called to convert two-dimensional teaching materials into three-dimensional graphics, and the three-dimensional teaching materials are pushed to the devices such as a tablet or AR glasses of the students for display through a wireless network. However, the existing scheme has certain defects, for example, the three-dimensional graph has a sense of space better than a two-dimensional image, but lacks real space depth and interactive immersion sense, students still need stronger space imagination to understand complex structures, the system adjusts the content only according to surface interaction data, and the system fails to deeply fuse a course knowledge system and a teaching target, so that the visualized content is disjointed from teaching logic, and the optimization of a continuous feedback content generation model is difficult. Disclosure of Invention The application aims to provide an optimization method and system for generating teaching contents by a large model, which are used for solving the problems that the prior art lacks real space depth and interactive immersion sense, visual contents and teaching logics are disjointed, and continuous feedback content generation model optimization is difficult. In order to solve the above technical problems, in a first aspect, the present application provides an optimization method for generating teaching contents by using a large model, including: the method comprises the steps of acquiring authoritative education data and teacher-student feedback data in a teaching scene, wherein the authoritative education data comprise course teaching outline, course targets and knowledge point hierarchical relation data, and the teacher-student feedback data comprise error marking data of a teacher on history generation teaching content and understanding difficulty feedback data of students on history generation teaching content; performing association matching on the authoritative education data and the teacher-student feedback data by using a multi-mode large model to generate initial teaching contents, wherein the initial teaching contents comprise test questions and question analysis; Constructing a teaching knowledge graph according to the authoritative education data, and calculating the degree of agreement between the initial teaching content and the teaching knowledge graph based on the knowledge point hierarchical relationship and the course target in the teaching knowledge graph; comparing the fit degree with a preset threshold value to correct the initial teaching content, and obtaining corrected teaching content; Selecting a prompting word frame from a preset prompting word template library, combining the corrected teaching content, and generating optimized teaching content by utilizing a multi-mode large model, wherein the prompting word frame comprises a description specification and an optimization direction which are matched with a teaching scene; And confirming the matching degree of the optimized teaching content and the course teaching outline and the student understanding difficulty score according to the teacher-student feedback data and the authoritative education data, and carrying out iterative optimization of the optimized teaching content by combining with a preset standard. Optionally, the authoritative education data and the teacher-student feedback data are associated and matched by using a multi-mode big model, so as to generate initial teaching content, wherein the initial teaching content comprises test questions and question analysis, and the method comprises the following steps: matching the core knowledge points of the course teaching outline with the error types of the error labeling data by using a multi-mode big model to generate a knowledge point error table; Matching the capability culture requirement of the course target with the difficulty interval of the understanding difficulty feedback data to generate a capability difficulty table; matching the knowledge point association path of the knowledge point hierarchical relation data with