Large model construction method supporting English subject education whole process scene

Abstract

The invention relates to the technical field of wisdom education, and discloses a large model construction method supporting an English subject education whole process scene. According to the method, a database comprising a knowledge point library, a text material library, a knowledge point explanation courseware library and a test question library and a task corpus for teaching material editing before teaching, courseware generation in teaching and test paper making after teaching are constructed, and three index evaluation systems of a knowledge point recommendation index, an important index and an interest index and three scene matching degree calculation models F1, F2 and F3 are combined; and intelligent generation of teaching content is realized. Through the content generation mode and the dynamic optimization mechanism based on the accurate mathematical model, pertinence, accuracy and individuation degree of teaching content generation are improved, and effective technical support is provided for improvement of teaching efficiency and realization of individualized teaching.

Description

Technical Field

[0001] The present application relates to the field of smart education technology, and in particular to the field of intelligent generation of English teaching content based on large models. Background Art

[0002] With the development of technologies such as artificial intelligence and big data, big model technology has begun to be widely used in the field of smart education. In English subject education, the application of big model technology helps to optimize educational teaching and management decisions and achieve deeper human-computer interaction. At present, big models have shown good application prospects in the field of English education, such as translation assistance and oral dialogue training.

[0003] However, the existing education model still has the following technical problems in supporting English subject education:

[0004] First, the existing education model mainly focuses on a single or limited autonomous learning scenario. For example, it provides assistance in question-answering and content generation in translation practice and oral dialogue, but lacks systematic support for the entire process of English subject education. This results in the core links of the teaching process, such as textbook compilation, courseware generation, and test paper production, being unable to be effectively supported.

[0005] Second, existing technologies lack in-depth analysis of the interactive needs of teaching content. In the actual teaching process, the needs of different teaching stages and different learning groups for teaching content vary significantly, and existing technologies cannot provide accurate content support for these differentiated needs.

[0006] Third, the teaching content generation is not targeted and adaptable enough. The teaching content generated by existing technologies often fails to meet the needs of all aspects of teaching. There are obvious deficiencies in the targeting, accuracy, and adaptability of the content to the teaching process, which affects the improvement of teaching effects.

[0007] Fourth, there is a lack of effective teaching content evaluation mechanism. Existing technologies make it difficult to scientifically evaluate the teaching value of generated content, cannot guarantee the fit between generated content and teaching objectives, and cannot achieve continuous optimization and improvement of content generation.

[0008] Therefore, there is an urgent need for a large model construction method that can support the entire process of English subject education and has the ability to generate accurate content in order to improve teaching efficiency and achieve personalized teaching. Summary of the invention

[0009] The purpose of this application is to provide a method for constructing a large model that supports the entire process of English subject education, so as to solve the problems raised in the above-mentioned background technology.

[0010] This application discloses a method for constructing a large model that supports the entire process of English subject education, including the following steps:

[0011] S1. Constructing a database, including: constructing a knowledge point library containing knowledge point attribute information, and marking the knowledge points involved in the teaching content therein; wherein the knowledge point attribute information includes at least two indexes for evaluating the importance of knowledge points; constructing a material library containing teaching content, wherein the material library includes a text material library, a knowledge point explanation courseware library, and a test question library;

[0012] S2. Construct a corpus for large model learning and training, wherein the corpus contains task corpora of multiple teaching scenarios, each task corpus contains question instruction information and its corresponding implementation method;

[0013] S3. Get the command information entered by the user;

[0014] S4. Generate corresponding teaching content based on the knowledge point attribute information by matching the user input instruction information obtained in step S3 with the learning and training results of the large model on the database constructed in step S1 and the corpus constructed in step S2;

[0015] S5. Output the teaching content generated in step S4 to the user.

[0016] In a preferred example, the knowledge point attribute information includes teaching stage, mastery requirements, recommendation index, importance index and interest index; wherein the recommendation index is calculated based on the matching between the mastery requirements and the difficulty level of the knowledge point; the importance index is calculated based on the matching between the mastery requirements of the knowledge point and the teaching group's degree of cognition of the knowledge point; the interest index is calculated based on the matching between the mastery requirements of the knowledge point and the degree of attention paid to the knowledge point by teachers and students in the teaching group.

[0017] In a preferred example, the task corpus includes: the task corpus of teaching material editing before teaching, which is composed of question instruction information for teaching material editing and the corresponding teaching material editing implementation method; the task corpus of courseware generation during teaching, which is composed of question instruction information for generating courseware and the corresponding courseware generation implementation method; the task corpus of test paper making after teaching, which is composed of question instruction information for making examination papers and the corresponding examination paper making implementation method.

[0018] In a preferred example, step S4 specifically includes: when the question instruction information is a textbook editing instruction, the textbook content required by the user is generated based on the attribute information of the knowledge points in combination with the data of the knowledge point library, the text material library and the reply method of the textbook editing task corpus before teaching; when the question instruction information is a courseware generation instruction, the courseware content required by the user is generated based on the attribute information of the knowledge points in combination with the data of the knowledge point library, the knowledge point explanation courseware library and the reply method of the courseware generation task corpus during teaching; when the question instruction information is a test paper making instruction, the test paper content required by the user is generated based on the attribute information of the knowledge points in combination with the data of the knowledge point library, the question library and the reply method of the test paper making task corpus after teaching.

[0019] In a preferred embodiment, the calculation formula of the recommendation index is:

[0020] x(a)=Match{(R a ), (Y a )}...(1)

[0021] Among them, x(a) is the recommendation index of the knowledge point; R a is the mastery requirement of the knowledge point, which is divided into three levels R′, R″, and R''', and R′>R″>R'''>0; Y a It is the difficulty level of the knowledge point, which is divided into three levels according to the difficulty coefficient: Y′, Y", Y''', and Y′>Y">Y'''>0.

[0022] In a preferred embodiment, the calculation formula of the importance index is:

[0023]

[0024] Among them, x(b) is the importance index of the knowledge point; R b is the requirement for mastering the knowledge point; m is the number of students in the teaching group; h is the number of listening and speaking questions involving the knowledge point; d ig is the score rate of the i-th student on this knowledge point in the g-th listening and speaking test; f ig is the value of whether the question is a listening and speaking question. If it is a listening and speaking question, it takes 1, otherwise it takes 0; l is the number of reading and writing questions involving this knowledge point; e ik is the score rate of the i-th student on this knowledge point in the k-th reading and writing test; f ik It is the value of whether the question is a reading and writing question. If it is a reading and writing question, it takes 1, otherwise it takes 0; α1 and α2 are weight coefficients.

[0025] In a preferred embodiment, the calculation formula of the interest index is:

[0026]

[0027] Among them, x(c) is the interest index of the knowledge point; R c is the requirement for mastering the knowledge point; m is the number of students in the teaching group; q is the number of materials containing the knowledge point used by students in the teaching group; r ip The number of times the i-th student applies the material p containing this knowledge point; w ip is the weight value of the student's application behavior at that time; n is the number of teachers in the teaching group; v is the number of materials containing this knowledge point used by teachers in the teaching group; s ju is the number of times the jth teacher uses the material u containing this knowledge point; w ju is the weight value of the teacher’s application behavior at that time; β1 and β2 are weight coefficients.

[0028] In a preferred example, when the question instruction information is a teaching material editing instruction, the calculation formula for determining the matching degree of the teaching material content is:

[0029]

[0030] Among them, F1 is the matching result of the current main text; x′ is the number of knowledge points contained in the current main text; F(x) (1) is the matching value of the knowledge point in accordance with the learning order of the syllabus; F(x) (2) is the requirement for mastering this knowledge point; F(x) (3) is the recommendation index of this knowledge point; F(x) (4) is the importance index of the knowledge point; F(x) (5) is the interest index of this knowledge point; μ1, μ2, μ3, μ4, μ5 are weight coefficients.

[0031] In a preferred example, when the question instruction information is a courseware generation instruction, the calculation formula for determining the matching degree of the courseware content is:

[0032] F2=ρ1F(y) (1) +ρ2(σ1F(y) (2) +σ2F(y) (3) +σ3F(y) (4) +σ4F(y) (5) )...(5)

[0033] Among them, F2 is the matching result of the current knowledge point courseware; F(y) (1) F(y) is the level of detail of the content of the knowledge point explained in the courseware; (2) is the mastering requirement of this knowledge point; F(y) (3) is the recommendation index of this knowledge point; F(y) (4) is the importance index of the knowledge point; F(y) (5)is the interest index of this knowledge point; ρ1, ρ2, σ1, σ2, σ3, σ4 are weight coefficients.

[0034] In a preferred example, when the question instruction information is a test paper making instruction, the calculation formula for determining the matching degree of the test paper content is:

[0035]

[0036] Among them, F3 is the matching result of the current question; F(z) (1) is the difficulty of the question type to which the question belongs; F(z) (2) is the discrimination of the question; z′ is the number of knowledge points included in the question; F(z) (3) is the mastery requirement of the zth knowledge point; F(z) (4) is the recommendation index of the zth knowledge point; F(z) (5) is the importance index of the zth knowledge point; F(z) (6) is the interest index of the zth knowledge point; ω1, ω2, ω3, is the weight coefficient.

[0037] The implementation methods of this application have the following technical effects:

[0038] First, the present invention constructs a database system that comprehensively covers all aspects of English teaching, including a knowledge point library, a text material library, a knowledge point explanation courseware library, and a test question library. The knowledge point library introduces multidimensional attributes such as teaching stages, mastering requirements, recommendation indexes, importance indexes, and interest indexes, and is dynamically calculated and generated based on student learning behavior data and teacher teaching behavior data through precise mathematical models such as Match{(Ra), (Ya)}. This database construction method based on precise mathematical models can more comprehensively and dynamically depict teaching elements, providing a solid data foundation for subsequent intelligent teaching.

[0039] Secondly, the present invention constructs a task corpus that includes pre-teaching teaching material editing, courseware generation during teaching, and post-teaching test paper production. Each task corpus contains question instruction information and corresponding implementation methods, forming a complete mapping from user needs to teaching content generation. By learning these corpora, the big model has mastered how to use knowledge point libraries, material libraries, etc. to intelligently generate teaching content according to user needs. This end-to-end corpus construction idea can greatly improve the automation and efficiency of content generation.

[0040] Thirdly, the present invention designs three matching degree calculation models F1, F2 and F3 in the scenarios of textbook editing, courseware generation and test paper production. Among them, F1 ensures the rationality of textbook arrangement through weighted calculation of five dimensions (teaching sequence matching value, mastering requirements, recommendation index, importance index and interest index); F2 optimizes courseware generation by evaluating the detail of explanation and the importance of knowledge points; F3 comprehensively considers the difficulty, discrimination and knowledge point characteristics of the questions to compose the test papers. This content generation method based on precise mathematical models can significantly improve the pertinence and effectiveness of teaching.

[0041] In addition, the large model construction method proposed in the present invention can automatically match the corresponding data and corpus according to the teaching stage, teaching group, question instructions, etc. input by the user, and generate teaching materials, courseware, test papers, etc. that meet the requirements in real time. Through the precise control of a series of mathematical models and algorithms, the scientificity and reliability of the generated content are ensured. This full-process, end-to-end content generation capability can greatly reduce the preparation pressure of front-line teachers and improve the efficiency of teaching design.

[0042] Finally, the present invention adopts big data analysis and artificial intelligence technology, based on precise mathematical models, to automatically statistically analyze learning and teaching data, evaluate knowledge point attributes, build databases and corpora, and realize the automation and intelligence of the teaching content generation process. The system continuously collects and analyzes teaching data, continuously optimizes the calculation parameters of various indicators, and enables the generated teaching content to better meet teaching needs.

[0043] In summary, the present invention has established a complete set of intelligent generation methods for teaching content by innovatively introducing a knowledge point three-index evaluation system (recommendation index, importance index, interest index) and a three-scenario matching degree calculation model (F1, F2, F3). This method based on a precise mathematical model not only realizes the automatic generation of teaching content in a full process, precision, and intelligence, but also ensures the continuous improvement of the generated content through dynamic optimization and adjustment, providing reliable technical support for improving the efficiency of teaching design and promoting personalized teaching in accordance with students' aptitude.

[0044] A large number of technical features are recorded in the specification of this application, which are distributed in various technical solutions. If all possible combinations of technical features of this application (i.e., technical solutions) are to be listed, the specification will be too long. In order to avoid this problem, the various technical features disclosed in the above-mentioned invention content of this application, the various technical features disclosed in the various embodiments and examples below, and the various technical features disclosed in the accompanying drawings can be freely combined with each other to form various new technical solutions (these technical solutions are all deemed to have been recorded in this specification), unless the combination of such technical features is technically infeasible. For example, in one example, feature A+B+C is disclosed, and in another example, feature A+B+D+E is disclosed, and features C and D are equivalent technical means that play the same role. Technically, only one can be used, and it is impossible to use them at the same time. Feature E can be combined with feature C technically. Then, the solution of A+B+C+D should not be deemed to have been recorded because it is technically infeasible, and the solution of A+B+C+E should be deemed to have been recorded. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A schematic diagram of the overall functional concept of a large model construction method supporting the entire process scenario of English subject education according to the first embodiment of the present application is shown.

[0046] Figure 2 A flowchart of a method for constructing a large model supporting the entire process of English subject education according to the first embodiment of the present application is shown. DETAILED DESCRIPTION

[0047] In the following description, many technical details are provided to help readers better understand the present application. However, those skilled in the art can understand that the technical solution claimed in the present application can be implemented even without these technical details and various changes and modifications based on the following embodiments.

[0048] Description of some concepts:

[0049] Knowledge point: A knowledge point is a relatively independent and complete unit in the teaching content and is the basic content for students to learn. In the present invention, the knowledge point not only includes basic information such as the name and content of the knowledge point, but also introduces multiple attribute dimensions such as teaching stage, mastery requirements, recommendation index, importance index, interest index, etc., to more comprehensively characterize the teaching characteristics of the knowledge point.

[0050] Mastering requirements: Mastering requirements refer to the level or degree of students' ability to understand and apply knowledge points. The present invention divides them into three levels: mastering application, understanding explanation, and knowing and understanding. They are used to describe the teaching requirements of each knowledge point in a refined manner and are an important basis for teaching design.

[0051] Recommendation Index: The recommendation index is an indicator calculated based on the knowledge point's mastery requirements and difficulty level, and is used to evaluate the recommended teaching priority of a knowledge point for students. The higher the recommendation index, the more suitable the knowledge point is for priority teaching. This indicator can be used to plan the teaching progress.

[0052] Importance Index: The importance index reflects the importance of a knowledge point to a specific teaching group. It takes into account the mastery requirements of the knowledge point and the actual cognitive level of the student group. The higher the importance index, the higher the requirement for mastering the knowledge point, but the students' actual mastery is insufficient, and teaching needs to be strengthened.

[0053] Interest Index: The interest index depicts the interest and attention of a specific group to a knowledge point from the perspective of the actual attention behavior of teachers and students. It is mainly calculated based on the actual frequency of use of relevant teaching materials by teachers and students. The higher the interest index, the more popular the knowledge point is among teachers and students, and it can be used as an important reference for selecting teaching content.

[0054] Task corpus: Task corpus refers to the corpus of question instructions and corresponding implementation methods for specific teaching tasks. The present invention constructs three types of task corpora: textbook editing, courseware generation, and test paper production. By learning these corpora, the big model masters the method of converting user needs into teaching content, which is the key to achieving end-to-end content automatic generation.

[0055] Matching degree: Matching degree is an indicator for comprehensively evaluating the pertinence and effectiveness of candidate teaching content. It integrates multiple factors such as the attributes of knowledge points, material characteristics, and learning conditions. The higher the matching degree score, the more the teaching content fits the teaching needs. The present invention proposes corresponding matching degree calculation models in multiple teaching scenarios.

[0056] Learning situation: Learning situation is a comprehensive reflection of students' learning status, including their cognitive level, learning ability, interests and hobbies, etc. The present invention constructs important indexes and interest indexes reflecting learning situation by analyzing students' test, homework, learning activities and other data, which are used to guide teaching students in accordance with their aptitude.

[0057] Teaching situation: Teaching situation refers to the actual teaching situation and the characteristics of teachers' teaching behavior. The present invention captures and analyzes the behavior data of teachers' preparation, teaching, and selection of teaching materials to form indicators such as interest index that reflect the degree of teachers' attention, which can be used to optimize the recommendation of teaching resources.

[0058] Big model: Big model is a natural language processing model based on deep learning, with massive parameters and powerful semantic understanding and knowledge learning capabilities. The present invention uses the domain knowledge base and task corpus constructed by big model learning to realize the automatic generation of teaching content in multiple scenarios. Big model is the core technical support of the present invention.

[0059] The following is a brief description of some of the innovative features of this application:

[0060] The present invention proposes a large model construction method that supports the entire process of English subject education, aiming to use artificial intelligence technology to realize the automatic generation of content in all aspects of English teaching and promote the precision, personalization and intelligence of teaching.

[0061] The basic idea of ​​this method is: first, comprehensively build a domain knowledge base covering all aspects of English teaching, including a knowledge point library, a text material library, a knowledge point explanation courseware library, a test question library, etc. Among them, the knowledge point library introduces attributes such as teaching stages, mastering requirements, recommendation index, importance index, interest index, etc., forming a multi-dimensional description system of knowledge points; other libraries store a wealth of teaching materials such as texts, courseware, test questions, etc., and annotate the related knowledge point information. These libraries form the basic data resources for the generation of English teaching content.

[0062] Secondly, we build a task corpus covering the entire teaching process. For three typical tasks, namely textbook editing, courseware generation, and test paper production, we collect and organize question instructions and corresponding implementation methods to form task corpora. These corpora depict the logic and methods of how to organize and generate teaching content according to needs in different teaching scenarios.

[0063] Secondly, based on the constructed knowledge base and corpus, the educational big model is trained. The big model masters the domain knowledge system of the English subject by learning the information of the knowledge base; by learning the task corpus, it masters the method of organizing and generating teaching content according to needs. This process is the key to the big model's ability to generate intelligent content in the field of education.

[0064] Finally, based on the trained big model, end-to-end teaching content generation is achieved. For the teaching needs input by the user (such as teaching stage, target student group, specific content needs, etc.), the system uses the big model to automatically retrieve and match the relevant data in the knowledge base, and combines the generation logic formed by the task corpus to automatically output teaching materials, courseware, test papers and other content that meet the needs. In this process, the system makes full use of the multi-dimensional attributes of knowledge points to achieve more accurate and personalized teaching content generation.

[0065] In addition, the present invention also designs a series of matching degree calculation models to evaluate the pertinence and effectiveness of generated content. In scenarios such as textbook editing, courseware generation, and test paper production, the model calculates the matching degree between content and teaching needs from the perspectives of knowledge point attributes, teaching requirements, student cognition, and actual attention. The matching degree score provides an important basis for the selection of teaching content.

[0066] In general, the present invention realizes the automatic content generation of the whole process of English teaching through a series of methods such as knowledge base construction, task corpus organization, large model training, content generation evaluation, etc. This large model-driven teaching content generation paradigm is expected to significantly improve the efficiency of teaching design, promote the intelligent and precise development of education, and provide a new technical path for the modernization of education.

[0067] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below in conjunction with the accompanying drawings.

[0068] The first implementation mode of the present application relates to a method for constructing a large model that supports the entire process of English subject education. The process is as follows: Figure 1 and Figure 2 As shown, the method comprises the following steps:

[0069] S1. Constructing a database, including: constructing a knowledge point library containing knowledge point attribute information; constructing a material library containing teaching contents, and marking the knowledge points involved in the teaching contents; wherein the knowledge point attribute information includes at least two indexes for evaluating the importance of knowledge points.

[0070] S2. Based on the database constructed in step S1, a corpus for large model learning and training is constructed, wherein the corpus contains task corpora of multiple teaching scenarios, and each task corpus contains question instruction information and its corresponding implementation method.

[0071] S3. Obtain the instruction information input by the user.

[0072] S4. Generate corresponding teaching content based on the knowledge point attribute information by matching the user input instruction information obtained in step S3 with the learning and training results of the large model on the database constructed in step S1 and the corpus constructed in step S2.

[0073] S5. Output the teaching content generated in step S4 to the user.

[0074] Optionally, the knowledge point attribute information includes teaching stage, mastery requirement, recommendation index, importance index and interest index.

[0075] Optionally, the recommendation index is calculated based on the matching between the mastery requirements of the knowledge point and its difficulty level; the importance index is calculated based on the matching between the mastery requirements of the knowledge point and the teaching group's level of cognition of the knowledge point; the interest index is calculated based on the matching between the mastery requirements of the knowledge point and the level of attention paid to the knowledge point by teachers and students in the teaching group.

[0076] Specifically, in addition to basic information such as the name and content of knowledge points, the present invention also introduces attributes such as teaching stage, mastery requirements, recommendation index, importance index, interest index, etc. These attributes characterize the importance and teaching requirements of knowledge points from different dimensions and are the key inputs of the subsequent intelligent generation algorithm. Designing them as optional provides flexibility for the construction of the knowledge point library.

[0077] The recommendation index reflects the recommended teaching priority of a knowledge point. It is obtained by calculating the matching between the mastery requirements and the difficulty level of the knowledge point. The mastery requirements are divided into three levels: mastery of application, understanding of explanation, and recognition of knowledge. The difficulty level is also divided into three levels. The higher the matching, the higher the recommendation index. This calculation method can statically evaluate the suitability of knowledge points for students.

[0078] The importance index reflects the importance of a knowledge point to the current teaching group. It is obtained by calculating the matching between the mastery requirements of a knowledge point and the actual degree of cognition of the student group on the knowledge point. The degree of cognition is based on the statistics of students' answers to relevant listening, speaking, reading and writing questions. The importance index is high for knowledge points with high mastery requirements but low student cognition. This calculation method can dynamically evaluate the importance of a knowledge point to a specific group.

[0079] The interest index reflects the current teaching group's interest and attention to the knowledge point. It is obtained by calculating the matching between the mastery requirements of the knowledge point and the actual frequency of use of the relevant teaching materials by teachers and students. The higher the frequency of teachers and students using the teaching materials containing the knowledge point, the higher the interest index. This calculation method dynamically evaluates the degree of attention paid to the knowledge point from the perspective of teacher and student behavior.

[0080] In summary, the recommendation index, importance index, and interest index comprehensively evaluate the importance of knowledge points from three dimensions: the characteristics of the knowledge points themselves, the cognitive level of students, and the attention of teachers and students. They are applied to the subsequent textbook editing, courseware generation, and test paper generation algorithms to make the generated teaching content more accurate and personalized. The calculation of these indexes is based on objective data, and by matching them with the requirements for mastery, they reflect the teaching concept of teaching students in accordance with their aptitude and teaching students in accordance with their groups.

[0081] In terms of technical implementation, big data analysis technology can be used to analyze and process data such as teaching syllabus, examination syllabus, student scores, teacher and student user behavior logs, etc., automatically calculate various indicators, save them in the knowledge point library, and update them dynamically. This not only reduces the workload of experts and teachers, but also ensures the objectivity of the evaluation process.

[0082] Optionally, the material library includes a text material library, a knowledge point explanation courseware library and a test question library.

[0083] Optionally, the task corpus includes: pre-teaching teaching material editing task corpus, which is composed of question instruction information for teaching material editing and corresponding teaching material editing implementation methods; courseware generation task corpus during teaching, which is composed of question instruction information for generating courseware and corresponding courseware generation implementation methods; post-teaching test paper production task corpus, which is composed of question instruction information for producing examination papers and corresponding examination paper production implementation methods.

[0084] Optionally, the instruction information input by the user includes teaching stage, teaching group information and question instruction information.

[0085] Specifically, the material library of the present invention mainly includes three types: a text material library, a knowledge point explanation courseware library and a test question library.

[0086] Text material library: It stores a large amount of English text materials and marks the knowledge points involved in the texts, providing candidate materials for textbook editors.

[0087] Knowledge point explanation courseware library: It stores the explanation courseware for each knowledge point and marks the corresponding knowledge points to provide reference materials for courseware generation.

[0088] Question bank: It stores a large number of test questions and marks the knowledge points to be tested, providing a source of questions for test paper composition. Dividing the material library into these three types can correspond to the content generation needs of the three scenarios before teaching, during teaching, and after teaching. At the same time, the material library and the knowledge point library are interconnected, which makes it easier for intelligent algorithms to filter materials based on knowledge point attributes.

[0089] Furthermore, in the corpus constructed by the present invention, the task corpus is divided into three types, namely:

[0090] Pre-teaching textbook editing corpus: consists of question instruction information for textbook editing and corresponding textbook editing implementation methods.

[0091] The courseware generation corpus in teaching consists of the question instruction information for generating courseware and the corresponding courseware generation implementation method.

[0092] Post-teaching test paper production corpus: It consists of question instruction information for producing test papers and corresponding test paper production implementation methods. Each task corpus contains question instruction information and corresponding implementation methods. By learning these corpora, the big model has mastered how to generate specific teaching content based on question instruction information and using relevant material libraries. This way of constructing corpus is the key to achieving end-to-end content generation.

[0093] Further, according to the present invention, the instruction information input by the user mainly includes three aspects: teaching stage, teaching group information, and question instruction information.

[0094] Teaching stage: Indicate the teaching stage corresponding to the required teaching content, such as high school, junior high school, etc., to facilitate the location of relevant knowledge points.

[0095] Teaching group information: specifies the target student group for the required teaching content, so as to facilitate the generation of content based on the learning situation.

[0096] Question instruction information: Indicates specific content generation requirements, such as editing teaching materials, generating courseware, and making test papers. By inputting these three pieces of information, users can guide the big model to generate the required personalized teaching content. The big model will automatically complete the content arrangement based on the specified teaching background, matching the corresponding task corpus, and using the associated material library and knowledge point information.

[0097] In summary, the present invention constructs a rich material library and task corpus, and by introducing structured user instructions, realizes the full-process, personalized, end-to-end intelligent generation of teaching content. In terms of technical implementation, artificial intelligence technologies such as natural language processing, information retrieval, and knowledge graphs can be used to perform semantic analysis, knowledge point extraction, automatic annotation and other pre-processing on the material library to construct task corpus. And based on the few-sample learning and transfer learning capabilities of the large model, automatic mapping from task corpus to teaching content is realized. The whole process has a high degree of automation and high generation efficiency. At the same time, the model can be optimized through continuous learning to continuously improve the pertinence and effectiveness of the teaching content.

[0098] Optionally, step S4 specifically includes: when the question instruction information is a textbook editing instruction, combining the data of the knowledge point library, the text material library and the reply method of the textbook editing task corpus before teaching, generating the textbook content required by the user based on the attribute information of the knowledge points; when the question instruction information is a courseware generation instruction, combining the data of the knowledge point library, the knowledge point explanation courseware library and the reply method of the courseware generation task corpus during teaching, generating the courseware content required by the user based on the attribute information of the knowledge points; when the question instruction information is a test paper making instruction, combining the data of the knowledge point library, the test question library and the reply method of the test paper making task corpus after teaching, generating the test paper content required by the user based on the attribute information of the knowledge points.

[0099] Specifically, the specific implementation method of S4 elaborates in detail how the method of the present invention utilizes the constructed database, corpus, and knowledge point attribute information in different teaching scenarios to dynamically generate teaching content through learning and reasoning of a large model.

[0100] 1) Textbook editing scenario: When the user instructs to edit textbooks, the system first extracts the knowledge point information and its attributes related to the teaching stage and target knowledge point from the knowledge point library, such as recommendation index, importance index, interest index, etc. Then, the text related to the target knowledge point is retrieved from the text material library.

[0101] Next, the big model combines the experience learned from the textbook editing task corpus, comprehensively considers the teaching order, recommendation index, importance index, and interest index of the knowledge points involved in each candidate text, and determines the best text sequence through weighted calculation. This process uses the text matching function F1 defined by formula (4). The higher the matching degree of the text combination, the more likely it is to be selected as the textbook content.

[0102] Finally, the large model generates a complete textbook manuscript based on the selected text combination and the textbook arrangement knowledge learned from the task corpus, thus realizing the automatic editing of textbook content.

[0103] 2) Courseware generation scenario: When the user instructs to generate courseware, the system first extracts the target knowledge point and its attribute information from the knowledge point library, and then retrieves the courseware materials related to the target knowledge point from the knowledge point explanation courseware library.

[0104] The large model uses the experience learned from the courseware generation task corpus to analyze the degree of explanation of each candidate courseware on the knowledge points. It combines the mastery requirements, recommendation index, importance index, and interest index of the knowledge points, and calculates the matching score of each candidate courseware through the courseware matching function F2 defined by formula (5).

[0105] The courseware with the highest matching score generally performs well in terms of explanation level, importance of knowledge points, students' cognitive level, and attention of teachers and students, and is more suitable as teaching courseware. The large model organizes the courseware content according to the selected courseware and the generation mode learned in the task corpus to generate complete multimedia courseware.

[0106] 3) Test paper making scenario: When the user instructs to make a test paper, the system first extracts the knowledge points and their attribute information that match the test scope from the knowledge point database, and then retrieves the test questions of the relevant knowledge points from the test question database.

[0107] The large model analyzes the test paper composition logic learned from the test paper making task corpus, analyzes the difficulty and discrimination of the test questions, combines the mastery requirements of the test knowledge points, the recommendation index, the importance index, and the interest index, and uses the test question matching function F3 defined in formula (6) to calculate the suitability of the test questions.

[0108] At the same time, the big model will also control the proportion of question types, difficulty gradient, knowledge point coverage, etc. of the test paper based on the test paper composition rules learned from the task corpus, select questions from the candidate questions, and automatically generate a complete test paper.

[0109] In summary, the present invention makes full use of the multi-dimensional attribute information of knowledge points, maps them with teaching materials such as questions, texts, and courseware, and forms a related knowledge network. And by constructing task corpora covering the entire teaching process, the large model is trained to master the generation logic and optimization strategies of various teaching contents.

[0110] In specific applications, users only need to describe their teaching needs in natural language. The system will dynamically extract relevant information from the knowledge network based on the needs, and combine the generation templates and optimization strategies learned from the task corpus to generate teaching content that meets the needs in real time.

[0111] Compared with the traditional teaching content generation method, the biggest feature of this invention is that it realizes the end-to-end connection of knowledge, materials, and teaching logic, and introduces multi-dimensional factors such as knowledge point difficulty, student cognition, and teacher-student attention for comprehensive optimization, making the teaching content generation process more intelligent, personalized, and precise. This is of great significance to improving teaching efficiency and learning effects.

[0112] Optionally, the calculation formula of the recommendation index is:

[0113] x(a)=Match{(R a ), (Y a )}...(1)

[0114] Among them, x(a) is the recommendation index of the knowledge point; R a is the mastery requirement of the knowledge point, which is divided into three levels according to the mastery requirement: R′, R", and R''', and R′>R">R'''>0; Y a It is the difficulty level of the knowledge point, which is divided into three levels according to the difficulty coefficient: Y′, Y", and Y''', and Y′>Y″>Y'''>0.

[0115] Optionally, the calculation formula of the importance index is:

[0116]

[0117] Among them, x(b) is the importance index of the knowledge point; R b is the requirement for mastering the knowledge point; m is the number of students in the teaching group; h is the number of listening and speaking questions involving the knowledge point; d ig is the score rate of the i-th student on this knowledge point in the g-th listening and speaking test; f ig is the value of whether the question is a listening and speaking question. If it is a listening and speaking question, it takes 1, otherwise it takes 0; l is the number of reading and writing questions involving this knowledge point; e ik is the score rate of the i-th student on this knowledge point in the k-th reading and writing test; f ik It is the value of whether the question is a reading and writing question. If it is a reading and writing question, it takes 1, otherwise it takes 0; α1 and α2 are weight coefficients.

[0118] Optionally, the calculation formula of the interest index is:

[0119]

[0120] Among them, x(c) is the interest index of the knowledge point; R c is the requirement for mastering the knowledge point; m is the number of students in the teaching group; q is the number of materials containing the knowledge point used by students in the teaching group; r ip The number of times the i-th student applies the material p containing this knowledge point; w ip is the weight value of the student's application behavior at that time; n is the number of teachers in the teaching group; v is the number of materials containing this knowledge point used by teachers in the teaching group; s ju is the number of times the jth teacher uses the material u containing this knowledge point; w ju is the weight value of the teacher’s application behavior at that time; β1 and β2 are weight coefficients.

[0121] The above content is a detailed explanation of the calculation formulas and parameter meanings of the three key indexes (recommendation index, importance index, interest index) in the present invention.

[0122] First, the calculation formula of the recommendation index x(a) is:

[0123] x(a)=Match{(R a ), (Y a )}...(1)

[0124] The formula shows that the recommendation index is determined by two factors: the mastery requirements and the degree of difficulty of the knowledge point. The mastery requirements and the difficulty coefficient are divided into three levels. The larger the value, the higher the mastery requirements and the greater the difficulty. The recommendation index is calculated by R a and Y a The higher the matching degree, the higher the recommendation index. This index can be used to filter and sort teaching knowledge points.

[0125] Secondly, the calculation formula of the importance index x(b) is:

[0126]

[0127] The formula shows that the importance index is determined by two factors: the mastery requirements of the knowledge point and the degree of cognition of the student group on the knowledge point. By counting the scores of students on relevant listening, speaking, reading and writing questions, and combining the weights of the question types, the overall degree of cognition of the student group on the knowledge point can be evaluated. The lower the degree of match between the degree of cognition and the mastery requirements of the knowledge point, the less ideal the student's mastery, and the higher the importance index of the knowledge point. This index can be used to discover students' weak knowledge points.

[0128] And, the calculation formula of interest index x(c) is:

[0129]

[0130] The formula shows that the interest index is determined by two factors: the mastery requirements of the knowledge point and the frequency of application of relevant materials by teachers and students. By counting the frequency of students and teachers applying materials containing target knowledge points, and combining the weights of different application behaviors (such as retrieval, downloading, editing, etc.), the overall attention of teachers and students to knowledge points can be evaluated. The higher the match between attention and mastery requirements, the stronger the learning interest of teachers and students, and the higher the interest index of knowledge points. This index can be used to discover knowledge points that teachers and students are interested in.

[0131] In summary, the recommendation index, importance index, and interest index characterize the importance of knowledge points from three dimensions: the difficulty of the knowledge points themselves, the cognitive level of students, and the attention of teachers and students. They are the key basis for intelligent teaching. The present invention automatically calculates these three indexes through big data analysis of student learning, examination, and activity data and teacher preparation and lecture data, which greatly simplifies the teaching design process, improves the accuracy of personalized teaching, and is of great value in improving teaching results.

[0132] Optionally, when the question instruction information is a textbook editing instruction, the calculation formula for determining the matching degree of the textbook content is:

[0133]

[0134] Among them, F1 is the matching result of the current main text; x′ is the number of knowledge points contained in the current main text; F(x) (1) is the matching value of the knowledge point in accordance with the learning order of the syllabus; F(x) (2) is the requirement for mastering this knowledge point; F(x) (3) is the recommendation index of this knowledge point; F(x) (4) is the importance index of the knowledge point; F(x) (5) is the interest index of this knowledge point; μ1, μ2, μ3, μ4, μ5 are weight coefficients.

[0135] Optionally, when the question instruction information is a courseware generation instruction, the calculation formula for determining the matching degree of the courseware content is:

[0136] F2=ρ1F(y) (1) +ρ2(σ1F(y) (2) +σ2F(y) (3) +σ3F(y) (4) +σ4F(y) (5) ...(5)

[0137] Among them, F2 is the matching result of the current knowledge point courseware; F(y) (1) F(y) is the level of detail of the content of the knowledge point explained in the courseware; (2) is the mastering requirement of this knowledge point; F(y) (3)is the recommendation index of this knowledge point; F(y) (4) is the importance index of the knowledge point; F(y) (5) is the interest index of this knowledge point; ρ1, ρ2, σ1, σ2, σ3, σ4 are weight coefficients.

[0138] Optionally, when the question instruction information is a test paper making instruction, the calculation formula for determining the matching degree of the test paper content is:

[0139]

[0140] Among them, F3 is the matching result of the current question; F(z) (1) is the difficulty of the question type to which the question belongs; F(z) (2) is the discrimination of the question; z′ is the number of knowledge points included in the question; F(z) (3) is the mastery requirement of the zth knowledge point; F(z) (4) is the recommendation index of the zth knowledge point; F(z) (5) is the importance index of the zth knowledge point; F(z) (6) is the interest index of the zth knowledge point; ω1, ω2, ω3, is the weight coefficient.

[0141] The above-mentioned specific technical means is a detailed explanation of the teaching content matching degree calculation formula in the three scenarios of textbook editing, courseware generation, and test paper production of the present invention.

[0142] First, in the textbook editing scenario, the calculation formula for the main text matching degree F1 is:

[0144]

[0145] The formula shows that when editing textbooks, the comprehensive evaluation of candidate texts needs to consider five factors: the degree to which the knowledge points conform to the syllabus, the requirements for mastering the knowledge points, the recommendation index, the importance index, and the interest index. The matching score of each candidate text is calculated by weighted average. The higher the matching score, the more the content of the text meets the teaching needs and the more suitable it is to be included in the textbook. This formula is used to select the optimal text combination.

[0146] Secondly, in the courseware generation scenario, the calculation formula for the knowledge point courseware matching degree F2 is:

[0147] F2=ρ1F(y) (1) +ρ2(σ1F(y) (2) +σ2F(y) (3) +σ3F(y) (4) +σ4F(y)(5) ...(5)

[0148] This formula shows that when generating courseware for explaining knowledge points, two factors should be considered comprehensively: one is whether the courseware explains the knowledge points in detail, and the other is the importance of the knowledge points to students, including mastery requirements, recommendation index, importance index and interest index. The degree of detail of the explanation and the importance of the knowledge points are assigned two weights, ρ1 and ρ22, respectively, and the four importance indicators are assigned four weights, σ1, σ2, σ3 and σ4. The matching score is calculated by weighted average. The higher the score, the more the courseware content fits the teaching needs.

[0149] In addition, in the test paper production scenario, the calculation formula for the test question matching degree F3 is:

[0150]

[0151] This formula shows that when evaluating the quality of test questions, three factors need to be considered comprehensively: the difficulty and discrimination of the questions themselves, the requirements for mastering the knowledge points contained in the questions, and the recommendation index, importance index and interest index of the knowledge points. The three indicators of question difficulty, discrimination and knowledge point importance are assigned three weights of ω1, ω2 and ω3 respectively, and the four knowledge point indicators are further assigned Four weights. Take the weighted average of each item and calculate the matching score. The higher the score, the more the question meets the teaching requirements and is more suitable for inclusion in the test paper.

[0152] In summary, the present invention constructs a complete matching degree calculation model for three typical teaching content generation scenarios. These models make full use of the multidimensional attributes of knowledge points, associate them with the content characteristics of questions, texts, and courseware, and form a set of effective teaching content evaluation and screening mechanisms. Compared with the traditional manual production method, this method is not only more efficient, but also can flexibly adjust various parameters and adapt to the changes in the needs of different teaching scenarios to generate more personalized and precise teaching content. This is of great significance for improving the quality of teaching and promoting teaching students in accordance with their aptitude.

[0153] In order to better understand the technical solution of the present application, a specific example is provided below for illustration. The details listed in the example are mainly for ease of understanding and are not intended to limit the scope of protection of the present application.

[0154] This example proposes a method for constructing a large model that supports the entire process of English subject education. First, based on the data information required for the entire process of English subject education, this method constructs a database including a knowledge point library, a text material library, a knowledge point explanation courseware library, and a test question library; secondly, based on the corpus information required for the entire process of English subject education, a corpus including pre-teaching textbook editing task corpus, in-teaching courseware generation task corpus, and post-teaching test paper production task corpus is constructed; finally, according to the data required for inputting instructions in the database trained by the large model learning and the reply method of the input instructions in the corpus, the content output of the reply instruction is realized. Figure 1 .

[0155] The specific content of this example includes the following three parts:

[0156] (I) Building a database

[0157] Based on the data information required for the entire process of English subject education, a database including a knowledge point library, a text material library, a knowledge point explanation courseware library, and a test question library is constructed. The constructed database can be used for the data required for large model learning training instructions. The specific database construction is as follows:

[0158] 1. Knowledge point database: Based on the knowledge point information in the teaching syllabus and examination syllabus of each teaching stage of the English subject, the knowledge point domain of each teaching stage is constructed, and the attributes such as the teaching stage, mastery requirements, recommendation index, importance index, interest index, etc. of each knowledge point are marked. Specifically, the knowledge point domain of each teaching stage is:

[0159]

[0160] Among them, the mastery requirements are divided into three levels, namely mastering application, understanding explanation, and knowing and understanding; the knowledge point recommendation index, knowledge point importance index, and knowledge point interest index are obtained by analyzing students' learning behavior data and teachers' teaching behavior data. The specific generation of each index is as follows:

[0161] (1) Knowledge point recommendation index

[0162] By analyzing the matching between the mastery requirements and the difficulty level of a knowledge point, a recommendation index for that knowledge point is generated, which can be used to prioritize knowledge points with a high recommendation index when generating teaching materials, courseware, and test papers in the entire English education process. Specifically, the recommendation index is calculated as follows:

[0163] x(a)=Match{(R a ), (Y a )} (1)

[0164] In formula (1), x(a) is the recommendation index of a certain knowledge point. The lower the mastering requirement of the knowledge point and the higher the difficulty, the lower the recommendation index of the knowledge point; R a is the mastery requirement of this knowledge point, and its value is divided into three levels according to the mastery requirement: R′, R", and R''', and R′>R">R'''>0; Y a is the difficulty of this knowledge point, and its value is divided into Y′, Y", Y''' according to the three levels of difficulty coefficient (i.e. high difficulty, medium difficulty and low difficulty), and Y′>Y">Y'''>0.

[0165] It should be pointed out that the difficulty of a knowledge point is determined by the average score of previous candidates' answers to the questions.

[0166] (2) Knowledge point importance index

[0167] By analyzing the matching between the requirements for mastering a knowledge point and the degree of cognition of the teaching group on the knowledge point, the importance index of the knowledge point is generated, which can be used to prioritize the knowledge points with high importance index when generating teaching materials, courseware, test papers, etc. in the whole process of English education. Specifically, the calculation method of the importance index is:

[0168]

[0169] In formula (2), x(b) is the importance index of a certain knowledge point. The higher the mastering requirement of the knowledge point and the lower the student's cognitive level, the higher the importance index of the knowledge point. b is the mastery requirement of the knowledge point, and its value is divided into three levels according to the mastery requirement: R′, R″, and R''', and R′>R″>R'''>0; m refers to the number of students in the teaching group, h refers to the number of listening and speaking questions involving this knowledge point, and d ig refers to the score rate of the i-th student on this knowledge point in the g-th listening and speaking test; f ig It refers to whether the question is a listening and speaking question. If it is a listening and speaking question, it takes 1, otherwise it takes 0; l refers to the number of reading and writing questions involving this knowledge point, e ik The score rate of the i-th student on this knowledge point in the g-th reading and writing test; f ik It refers to whether the question is a reading and writing question. If it is a reading and writing question, it takes 1, otherwise it takes 0. α1 and α2 are weight coefficients.

[0170] It should be pointed out that the degree of cognition of knowledge points is obtained through comprehensive analysis of the teaching group's answers to listening, speaking, reading and writing questions.

[0171] (3) Knowledge point interest index

[0172] By analyzing the matching between the requirements for mastering a knowledge point and the degree of attention paid by teachers and students in the teaching group to the knowledge point, the interest index of the knowledge point is generated, which can be used to prioritize the knowledge points with high interest index when generating teaching materials, courseware, test papers, etc. in the whole process of English education. Specifically, the calculation method of the interest index is:

[0173]

[0174] In formula (3), x(c) is the interest index of a certain knowledge point. The higher the mastering requirement of the knowledge point and the higher the attention of teachers and students, the higher the interest index of the knowledge point. c is the mastery requirement of the knowledge point, and its value is divided into three levels according to the mastery requirement: R′, R″, and R''', and R′>R″>R'''>0; m refers to the number of students in the teaching group, q is the number of materials containing the knowledge point applied by the students in the teaching group, and r ip is the number of times the i-th student applies the material p containing this knowledge point, w ip is the weight value of the student's application behavior at that time (where the search behavior weight value is recorded as 1, the download behavior weight value is recorded as 2...); n refers to the number of teachers in the teaching group, v is the number of materials containing the knowledge point applied by the teachers in the teaching group, and s ju is the number of times the jth teacher uses the material u containing this knowledge point, w ju is the weight value of the teacher's application behavior at that time (where the editing behavior weight value is 1, the selection behavior weight value is 2...); β1 and β2 are weight coefficients.

[0175] It should be pointed out that the degree of attention paid to knowledge points is obtained through statistical analysis of the frequency of application of materials by teachers and students in the teaching group.

[0176] 2. Text material library, knowledge point explanation courseware library, and test question library: According to the teaching requirements of different teaching stages, collect a large number of different types of text materials and mark the knowledge points in the texts; create a courseware library for explaining each knowledge point and mark the knowledge points explained in the courseware; collect a large number of test questions of different types and mark the knowledge points in the test questions.

[0177] (II) Building a corpus

[0178] Based on the corpus information required for the entire process of English subject education, a corpus is constructed that includes the corpus for pre-teaching textbook editing tasks, the corpus for teaching courseware generation tasks, and the corpus for post-teaching test paper production tasks. The constructed corpus can be used to respond to large model learning training instructions.

[0179] (1) Pre-teaching textbook editing task corpus

[0180] The textbook editing task corpus consists of the question instruction information for textbook editing and the corresponding textbook editing implementation method. The specific textbook editing implementation method is as follows:

[0181] ① Based on the teaching syllabus, obtain the target knowledge point domain and the beyond-syllabus knowledge point domain of the preset teaching stage, determine the overall content of the textbook, and determine the number of main texts according to the class requirements.

[0182] ② Combine the relevance of the knowledge learning sequence of each alternative text in the text material library, the importance of the knowledge point mastery requirements, the knowledge point recommendation index, the knowledge point importance index, and the knowledge point interest index to determine the current main text.

[0183]

[0184] In formula (4), F1 is the matching result of the current main text; x′ refers to the number of knowledge points contained in the current main text; F(x) (1) It refers to the matching value of the knowledge point in accordance with the learning order of the teaching syllabus. The earlier the knowledge point is learned in this teaching stage, the larger the value is. If there is no learning order requirement (generally, it is a knowledge point not learned in this stage), it is 0. F(x) (2) Refers to the mastery requirements of this knowledge point; F(x) (3) Refers to the recommendation index of this knowledge point; F(x) (4) Refers to the importance index of the knowledge point; F(x) (5) It refers to the interest index of this knowledge point; μ1, μ2, μ3, μ4, and μ5 are the influencing factors of the five dimensions.

[0185] ③Based on the confirmed main texts, update the knowledge points in the knowledge point domain of this teaching stage until all the main texts are determined and the textbook editing is completed.

[0186] (2) Generating task corpus in teaching courseware

[0187] The courseware generation task corpus consists of the question instruction information for generating courseware and the corresponding courseware generation implementation method. The specific courseware generation implementation method is as follows:

[0188] ① Based on the teaching syllabus, obtain the key points and general knowledge points of the main text, determine the overall content of the courseware, and determine the number of knowledge point courseware according to the course duration.

[0189] ② Determine the current knowledge point courseware based on the matching degree between the detailed explanation of the knowledge point by each candidate knowledge point courseware in the knowledge point explanation courseware library and the importance of the knowledge point mastery requirement, the knowledge point recommendation index, the knowledge point importance index, and the knowledge point interest index.

[0190] F2=ρ1F(y)(1) +ρ2(σ1F(y) (2) +σ2F(y) (3) +σ3F(y) (4) +σ4F(y) (5) ) (5)

[0191] In formula (5), F2 is the matching result of the current knowledge point courseware; F(y) (1) F(y) is the level of detail of the content of the knowledge point explained in the courseware; (2) is the mastering requirement of this knowledge point; F(y) (3) is the recommendation index of this knowledge point; F(y) (4) is the importance index of the knowledge point; F(y) (5) is the interest index of this knowledge point; ρ1, ρ2, σ1, σ2, σ3, σ4 are weight coefficients.

[0192] ③Based on the confirmed knowledge point courseware, update the knowledge point domain contained in the main text until all knowledge point courseware are confirmed and the generation of classroom courseware is completed.

[0193] (3) Post-teaching test paper production task corpus

[0194] The test paper making task corpus consists of the question instruction information for making the test paper and the corresponding test paper making implementation method. The specific test paper making implementation method is as follows:

[0195] ① Based on the examination syllabus, obtain the target assessment knowledge point domain of the teaching stage, obtain the test score ratio of each knowledge point, and determine the type and number of examination questions based on the test paper template.

[0196] ② Determine the current test question based on the difficulty and discrimination of each alternative question in the question bank and the matching degree between the knowledge point content contained in the question (including the importance of knowledge point mastery requirements, knowledge point recommendation index, knowledge point importance index, and knowledge point interest index).

[0197]

[0198] In formula (6), F3 is the matching result of the current question; F(z) (1) is the difficulty of the question type to which the question belongs, F(z) (2) is the discrimination of the question, z′ is the number of knowledge points included in the question, and F(z) (3) is the mastery requirement of the zth knowledge point; F(z) (4) is the recommendation index of the zth knowledge point; F(z) (5) is the importance index of the zth knowledge point; F(z) (6) is the interest index of the zth knowledge point; ω1, ω2, ω3, is the weight coefficient.

[0199] ③Based on the confirmed test questions, update the knowledge points in the knowledge point domain of the test knowledge graph until all test questions are determined and the test paper generation is completed.

[0200] (III) Implementing command reply based on the input command information obtained

[0201] Step 1: Get input instructions

[0202] Get user input instructions (instructions can be in text or voice form), the input instructions should include the teaching stage, teaching group and corresponding question instruction information. For example: edit the senior high school textbook.

[0203] Step 2: Output the response instructions for the entire English education process scenario

[0204] By matching the user's input command information with the big model's learning and training results for the database and corpus, it outputs the response command information required for the entire process of English education, including professional multimedia teaching materials, targeted multimedia courseware, and test papers with appropriate difficulty and highlighted key points.

[0205] (1) Reply to the instruction for the pre-teaching teaching material editing task

[0206] The textbook editing instruction information input by the user is obtained, and the textbook required by the user is generated based on the data required for the textbook editing instructions in the database trained by the large model learning and the response method of the pre-teaching textbook editing task corpus in the corpus.

[0207] (2) Courseware generates task response instructions during teaching

[0208] Obtain the courseware generation instruction information input by the user, and generate the courseware required by the user based on the data required for the courseware generation instruction in the database trained by the large model learning and the response method of the courseware generation task corpus in the teaching in the corpus.

[0209] (3) Post-teaching test paper preparation task reply instructions

[0210] The test paper making instruction information input by the user is obtained, and the test paper required by the user is generated according to the data required for the test paper making instruction in the database trained by the large model learning and the answer method of the post-teaching test paper making task corpus in the corpus.

[0211] Technical effects:

[0212] First, the above embodiment constructs a database system that comprehensively covers all aspects of English teaching, including a knowledge point library, a text material library, a knowledge point explanation courseware library, and a test question library. The knowledge point library introduces multi-dimensional attributes such as teaching stage, mastery requirements, recommendation index, importance index, and interest index, and uses Match{(R a ), (Y a )}, etc., are dynamically calculated and generated based on students’ learning behavior data and teachers’ teaching behavior data. This database construction method based on precise mathematical models can more comprehensively and dynamically describe teaching elements, providing a solid data foundation for subsequent intelligent teaching.

[0213] Secondly, the above embodiment constructs a task corpus that includes pre-teaching teaching material editing, courseware generation during teaching, and post-teaching test paper production. Each task corpus contains question instruction information and corresponding implementation methods, forming a complete mapping from user needs to teaching content generation. By learning these corpora, the big model has mastered how to use knowledge point libraries, material libraries, etc. to intelligently generate teaching content according to user needs. This end-to-end corpus construction idea can greatly improve the automation and efficiency of content generation.

[0214] Again, the above embodiments respectively design three matching degree calculation models F1, F2, and F3 in the scenarios of textbook editing, courseware generation, and test paper production. Among them, F1 ensures the rationality of textbook arrangement through weighted calculation of five dimensions (teaching sequence matching value, mastering requirements, recommendation index, importance index, and interest index); F2 optimizes courseware generation by evaluating the detail of explanations and the importance of knowledge points; and F3 comprehensively considers the difficulty, discrimination, and knowledge point characteristics of questions to compose test papers. This content generation method based on precise mathematical models can significantly improve the pertinence and effectiveness of teaching.

[0215] In addition, the large model construction method proposed in the above embodiment can automatically match the corresponding data and corpus according to the teaching stage, teaching group, question instructions, etc. input by the user, and generate teaching materials, courseware, test papers, etc. that meet the requirements in real time. Through the precise control of a series of mathematical models and algorithms, the scientificity and reliability of the generated content are ensured. This full-process, end-to-end content generation capability can greatly reduce the preparation pressure of front-line teachers and improve the efficiency of teaching design.

[0216] Finally, the above embodiment uses big data analysis and artificial intelligence technology, based on precise mathematical models, to automatically statistically analyze learning and teaching data, evaluate knowledge point attributes, build databases and corpora, and realize the automation and intelligence of the teaching content generation process. The system continuously collects and analyzes teaching data and continuously optimizes the calculation parameters of various indicators, so that the generated teaching content can better meet teaching needs.

[0217] In summary, the above embodiment establishes a complete set of intelligent generation methods for teaching content by innovatively introducing the knowledge point three-index evaluation system (recommendation index, importance index, interest index) and three-scenario matching degree calculation models (F1, F2, F3). This method based on precise mathematical models not only realizes the automatic generation of teaching content in a full process, precision, and intelligence, but also ensures the continuous improvement of the generated content through dynamic optimization and adjustment, providing reliable technical support for improving the efficiency of teaching design and promoting personalized teaching in accordance with students' aptitude.

[0218] It should be noted that in the application documents of this patent, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "including one" do not exclude the existence of other identical elements in the process, method, article or device including the elements. In the application documents of this patent, if it is mentioned that an action is performed according to an element, it means that the action is performed at least according to the element, which includes two situations: performing the action only according to the element, and performing the action according to the element and other elements. Expressions such as multiple, multiple, and multiple include 2, 2 times, 2 kinds, and more than 2, more than 2 times, and more than 2 kinds.

[0219] All documents mentioned in this application are considered to be included in the disclosure of this application as a whole, so that they can be used as the basis for modification when necessary. In addition, it should be understood that after reading the above disclosure of this application, those skilled in the art can make various changes or modifications to this application, and these equivalent forms also fall within the scope of protection claimed in this application.

Claims

1. A method for constructing a large model that supports the entire process of English subject education, characterized in that: The following steps are involved: S1. Constructing a database, including: constructing a knowledge point library containing knowledge point attribute information, and marking the knowledge points involved in the teaching content therein; wherein the knowledge point attribute information includes at least two indexes for evaluating the importance of knowledge points; constructing a material library containing teaching content, wherein the material library includes a text material library, a knowledge point explanation courseware library, and a test question library; S2. Construct a corpus for large model learning and training, wherein the corpus contains task corpora of multiple teaching scenarios, each task corpus contains question instruction information and its corresponding implementation method; S3. Obtaining the instruction information input by the user; S4. Generate corresponding teaching content based on the knowledge point attribute information by matching the user input instruction information obtained in step S3 with the learning and training results of the large model on the database constructed in step S1 and the corpus constructed in step S2; S5. Output the teaching content generated in step S4 to the user.

2. The method according to claim 1, characterized in that: The knowledge point attribute information includes teaching stage, mastery requirements, recommendation index, importance index and interest index; wherein the recommendation index is calculated based on the matching between the mastery requirements and the difficulty level of the knowledge point; the importance index is calculated based on the matching between the mastery requirements of the knowledge point and the teaching group's degree of cognition of the knowledge point; the interest index is calculated based on the matching between the mastery requirements of the knowledge point and the degree of attention paid by teachers and students in the teaching group to the knowledge point.

3. The method according to claim 1, characterized in that: The task corpus includes: the task corpus of teaching material editing before teaching, which is composed of question instruction information for teaching material editing and the corresponding teaching material editing implementation method; the task corpus of courseware generation during teaching, which is composed of question instruction information for generating courseware and the corresponding courseware generation implementation method; the task corpus of test paper making after teaching, which is composed of question instruction information for making test papers and the corresponding test paper making implementation method.

4. The method according to claim 1, characterized in that: Step S4 specifically includes: when the question instruction information is a textbook editing instruction, combining the data of the knowledge point library, the text material library and the reply method of the textbook editing task corpus before teaching, the textbook content required by the user is generated based on the attribute information of the knowledge points; when the question instruction information is a courseware generation instruction, combining the data of the knowledge point library, the knowledge point explanation courseware library and the reply method of the courseware generation task corpus during teaching, the courseware content required by the user is generated based on the attribute information of the knowledge points; when the question instruction information is a test paper making instruction, combining the data of the knowledge point library, the test question library and the reply method of the test paper making task corpus after teaching, the test paper content required by the user is generated based on the attribute information of the knowledge points.

5. The method according to claim 2, characterized in that: The calculation formula of the recommendation index is: x(a)=Match{(R a ),(Y a )} ...(1) Among them, x(a) is the recommendation index of the knowledge point; R a is the mastery requirement of the knowledge point, which is divided into three levels according to the mastery requirement: R′, R"", and R''', and R′>R"">R'''>0; Y a It is the difficulty level of the knowledge point, which is divided into three levels according to the difficulty coefficient: Y′, Y"′, Y''', and Y′>Y"">Y'''>0.

6. The method according to claim 2, characterized in that: The calculation formula of the important index is: Among them, x(b) is the importance index of the knowledge point; R b is the requirement for mastering the knowledge point; m is the number of students in the teaching group; h is the number of listening and speaking questions involving the knowledge point; d ig is the score rate of the i-th student on this knowledge point in the g-th listening and speaking test; f ig is the value of whether the question is a listening and speaking question. If it is a listening and speaking question, it takes 1, otherwise it takes 0; l is the number of reading and writing questions involving this knowledge point; e ik is the score rate of the i-th student on this knowledge point in the k-th reading and writing test; f ik It is the value of whether the question is a reading and writing question. If it is a reading and writing question, it takes 1, otherwise it takes 0; α1 and α2 are weight coefficients.

7. The method according to claim 2, characterized in that: The calculation formula of the interest index is: Among them, x(c) is the interest index of the knowledge point; R c is the requirement for mastering the knowledge point; m is the number of students in the teaching group; q is the number of materials containing the knowledge point used by students in the teaching group; r ip The number of times the i-th student applies the material p containing this knowledge point; w ip is the weight value of the student's application behavior at that time; n is the number of teachers in the teaching group; v is the number of materials containing this knowledge point used by teachers in the teaching group; s ju is the number of times the jth teacher uses the material u containing this knowledge point; w ju is the weight value of the teacher’s application behavior at that time; β1 and β2 are weight coefficients.

8. The method according to claim 4, characterized in that: When the question instruction information is a textbook editing instruction, the calculation formula for determining the matching degree of the textbook content is: Among them, F1 is the matching result of the current main text; x′ is the number of knowledge points contained in the current main text; F(x) (1) is the matching value of the knowledge point in accordance with the learning order of the syllabus; F(x) (2) is the requirement for mastering this knowledge point; F(x) (3) is the recommendation index of this knowledge point; F(x) (4) is the importance index of the knowledge point; F(x) (5) is the interest index of this knowledge point; μ1, μ2, μ3, μ4, μ5 are weight coefficients.

9. The method according to claim 4, characterized in that: When the question instruction information is a courseware generation instruction, the calculation formula for determining the matching degree of the courseware content is: F2=ρ1F(y) (1) +ρ2(σ1F(y) (2) +σ2F(y) (3) +σ3F(y) (4) +σ4F(y) (5) ...(5) Among them, F2 is the matching result of the current knowledge point courseware; F(y) (1) F(y) is the level of detail of the content of the knowledge point explained in the courseware; (2) is the mastering requirement of this knowledge point; F(y) (3) is the recommendation index of this knowledge point; F(y) (4) is the importance index of the knowledge point; F(y) (5) is the interest index of this knowledge point; ρ1, ρ2, σ1, σ2, σ3, σ4 are weight coefficients.

10. The method according to claim 4, characterized in that: When the question instruction information is a test paper making instruction, the calculation formula for determining the matching degree of the test paper content is: Among them, F3 is the matching result of the current question; F(z) (1) is the difficulty of the question type to which the question belongs; F(z) (2) is the discrimination of the question; z′ is the number of knowledge points included in the question; F(z) (3) is the mastery requirement of the zth knowledge point; F(z) (4) is the recommendation index of the zth knowledge point; F(z) (5) is the importance index of the zth knowledge point; F(z) (6) is the interest index of the zth knowledge point; ω1, ω2, ω3, is the weight coefficient.

Images