Generative AI-assisted teaching plan automatic construction method and device, equipment and medium

Abstract

The invention relates to a generative AI-assisted teaching plan automatic construction method and device, equipment and a medium. The method comprises the following steps: analyzing a course standard document to extract a subject core concept; generating a customizable project-based learning situation through a generative AI model based on the subject core concept in combination with a preset situation database and an interdisciplinary knowledge graph; generating a structured teaching activity process template by using a preset teaching design model according to the customized project-based learning situation; based on the teaching activity process template, matching and obtaining teaching resources, cognitive tools and an evaluation template from a preset resource and tool library; and combining the teaching resources, the cognitive tool, the evaluation template and the teaching activity process template to generate an initial teaching plan. According to the invention, full-process automation of teaching plan design is realized, the technical problems of low efficiency of manual teaching plan construction and insufficient interdisciplinary integration are solved, and the standardization degree and implementability of teaching design are improved.

Description

Technical Field

[0001] The present invention belongs to the field of teaching design, and in particular relates to a generative AI-assisted teaching plan automatic construction method, device, equipment and medium. Background Art

[0002] With the rapid development of educational information technology, intelligent lesson plan generation technology has gradually been applied to instructional design. This technology uses digital tools to assist teachers in constructing teaching frameworks. Its hallmark is the standardized organization of lesson plan elements using structured templates. This has given rise to the current mainstream semi-automated lesson plan construction method. Traditionally, lesson plan construction relies primarily on a manual process: teachers manually parse curriculum standards documents and extract core knowledge points; design learning scenarios and teaching activities based on their personal experience; manually search and match teaching tools from decentralized resource repositories; and independently develop and integrate assessment plans into the final lesson plan. However, current semi-automated lesson plan construction methods still have significant flaws: weak interdisciplinary integration capabilities, making it difficult for manual design to effectively connect real-world contexts with multidisciplinary knowledge, resulting in a lack of contextual authenticity; low resource matching accuracy: the selection of teaching resources and cognitive tools relies on subjective experience and lacks quantitative correlation with the cognitive level of the teaching objectives; significant bottlenecks in generation efficiency, with the average time required for a single lesson plan development being long, unable to meet the needs of frequent teaching iterations; and a failure to integrate the REAL-4Ps framework, making it difficult to unleash teacher creativity, stimulate student initiative, or bridge resource gaps. Summary of the Invention

[0003] Based on this, it is necessary to provide a generative AI-assisted teaching plan automatic construction method, device, equipment and medium that can solve the above problems.

[0004] In a first aspect, the present application provides a generative AI-assisted method for automatically constructing teaching plans, comprising:

[0005] Parse curriculum standard documents and extract core concepts of the subject;

[0006] Based on the core concepts of the subject, combined with the preset context database and interdisciplinary knowledge graph, a generative AI model is used to generate customizable project-based learning scenarios;

[0007] Based on the customized project-based learning scenario, the preset teaching design model is used to generate the corresponding structured teaching activity process template;

[0008] Based on the teaching activity process template, obtain corresponding teaching resources, cognitive tools and assessment templates from the preset resource and tool library;

[0009] Combine teaching resources, cognitive tools, assessment templates, and teaching activity process templates to form an initial lesson plan.

[0010] In one embodiment, based on core disciplinary concepts, combined with a pre-set context database and interdisciplinary knowledge graph, a generative AI model is used to generate customizable project-based learning scenarios, including:

[0011] Retrieve real-world contextual data related to core concepts of the subject from a pre-set contextual database;

[0012] Based on the interdisciplinary knowledge graph, the core concepts of the discipline are mapped with real-world contextual data;

[0013] Generate semantics from the association mapping results through a generative AI model, and output multiple project-based learning scenarios;

[0014] In response to the user's selection operation of multiple project-based learning scenarios, a customized project-based learning scenario is generated.

[0015] In one embodiment, based on the customized project-based learning scenario, a corresponding structured teaching activity process template is generated using a preset teaching design model, including:

[0016] Utilize pre-set instructional design models to analyze the teaching objectives and activity elements of customized project-based learning scenarios;

[0017] Based on the teaching objectives and the preset teaching theory rule base, the teaching design model is used to generate a teaching stage sequence that matches the activity elements;

[0018] According to the sequence of teaching stages, a structured teaching activity process template is constructed, which includes time allocation, teacher-student interaction logic and tool calling nodes.

[0019] In one embodiment, based on the teaching activity process template, corresponding teaching resources, cognitive tools and evaluation templates are obtained from a preset resource and tool library, including:

[0020] Analyze the teaching stage types, activity elements and target cognitive levels in the structured teaching activity process template;

[0021] Based on the preset matching rules, the teaching stage type, activity elements and target cognitive level are mapped to the associated tags in the resource and tool library;

[0022] According to the mapping results of the associated tags, the corresponding teaching resources, cognitive tools and assessment templates are retrieved and extracted from the resource and tool library.

[0023] In one embodiment, the method further comprises:

[0024] Real-time collection of teacher-student interaction data, student tool operation logs, and periodic evaluation results during the implementation of the initial teaching plan to generate a multi-dimensional learning behavior dataset;

[0025] Perform spatiotemporal alignment and feature annotation on the learning behavior dataset to construct a dynamic learning process map;

[0026] Conduct time-series analysis on the learning process graph and, combined with a pre-set learning ability assessment indicator library, identify learning progress deviations and abnormal knowledge acquisition events;

[0027] Based on learning progress deviations and abnormal knowledge mastery events, the preset feedback template library is called, and a structured teaching plan feedback report is generated in combination with student cognitive model reasoning.

[0028] In one embodiment, the method further comprises:

[0029] Based on learning progress deviation and knowledge acquisition anomalies, the loss value is optimized by the following calculation model:

[0030]

[0031] Where L is the loss function, α is the learning progress deviation penalty coefficient, Γ(·) is the learning progress deviation function, Δ t is the learning accuracy deviation data, n is the total number of abnormal types, k is the knowledge abnormal event type index, λ k is the dynamic teaching weight coefficient, ψ(·) is the abnormal quantization function, Δ k is the set of k-th type of knowledge anomaly events, β is the cognitive alignment strength coefficient, D KL is the KL divergence operator, C m is the probability distribution of students’ cognitive states, M ideal It is the target cognitive model;

[0032] Based on the optimization loss value, the updated parameters of the generative AI model are calculated using the following formula:

[0033]

[0034] Among them, θ new is the updated generative AI model parameter vector, θ old is the parameter vector of the generative AI model before updating, η is the adaptive learning rate, is the gradient of the loss function;

[0035] The updated parameters are used to update the parameters of the generative AI model, and the project-based learning scenario is regenerated based on the updated generative AI model to update the initial lesson plan.

[0036] In one embodiment, parsing the curriculum standard document and extracting the core concepts of the subject include:

[0037] Access to standardized curriculum standards data;

[0038] Structural processing of curriculum standard data to identify key knowledge points;

[0039] Based on the preset subject label system, key knowledge points are mapped to core concepts of the subject.

[0040] In a second aspect, the present application also provides a generative AI-assisted teaching plan automatic construction device, comprising:

[0041] Teaching concept extraction module, used to parse curriculum standard documents and extract core concepts of the subject;

[0042] The learning scenario generation module is used to generate customizable project-based learning scenarios based on core subject concepts, combined with a preset scenario database and interdisciplinary knowledge graph, using a generative AI model;

[0043] The teaching process generation module is used to generate corresponding structured teaching activity process templates based on customized project-based learning scenarios using the preset teaching design model;

[0044] The resource tool matching module is used to obtain corresponding teaching resources, cognitive tools and assessment templates from the preset resource and tool library based on the teaching activity process template;

[0045] The lesson plan generation module is used to combine teaching resources, cognitive tools, assessment templates and teaching activity process templates to form an initial lesson plan.

[0046] In a third aspect, the present application also provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the steps of the above-mentioned generative AI-assisted teaching plan automatic construction method.

[0047] In a fourth aspect, the present application also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above-mentioned generative AI-assisted teaching plan automatic construction method are implemented.

[0048] The above-mentioned generative AI-assisted lesson plan automatic construction method, device, computer equipment and storage medium extract the core concepts of the subject by parsing the curriculum standard document, so that the lesson plan design is closely linked to the teaching objectives, avoiding the subjectivity and omissions of manual analysis; based on the core concepts of the subject combined with the preset context database and the interdisciplinary knowledge graph, a generative AI model is used to generate customizable project-based learning scenarios, and the real-world scenarios are mapped with multidisciplinary knowledge to solve the problems of weak interdisciplinary integration and insufficient context authenticity in traditional manual design; a preset teaching design model is used to generate a structured teaching activity process template, clarify the sequence of teaching stages, time allocation and teacher-student interaction logic, and improve the standardization of teaching design; based on the teaching activity process template, teaching resources, cognitive tools and evaluation templates are matched from the preset resource and tool library, and the quantitative association between resources and the cognitive level of teaching objectives is achieved through label mapping, solving the problem of low resource matching accuracy; the various elements are combined to generate an initial lesson plan, realizing the automation of the entire lesson plan design process, significantly shortening the lesson plan development time, meeting the needs of high-frequency teaching iteration, and breaking through the bottleneck of low lesson plan generation efficiency of traditional semi-automated construction methods. Using the REAL-4Ps framework, for students: shift from learning knowledge to practicing thinking, and establish major subject concepts in creativity; for teachers: minimalist technology reduces transactional work, focuses on teaching design and humanistic care, and uses AI and neuroscience to promote educational design from experience-driven to evidence-driven. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0050] Figure 1 This is a flow chart of a generative AI-assisted teaching plan automation construction method of the present invention;

[0051] Figure 2 This is a structural diagram of a generative AI-assisted teaching plan automation construction device of the present invention. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0053] In one embodiment, Figure 1As shown, a generative AI-assisted method for automatically constructing teaching plans is provided. This embodiment uses the method applied to a terminal as an example. It is understandable that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0054] S01, parse the curriculum standard documents and extract the core concepts of the subject.

[0055] Among them, the core concepts of the subject are extracted from the curriculum standard documents (such as the subject teaching syllabus issued by the Ministry of Education), and the content related to the subject practice is retained first through the keyword filtering mechanism. Natural language processing technology can be used to perform in-depth semantic analysis on the filtered curriculum standard documents to identify subject entities (such as extracting predicate understanding and object light reaction stage from understanding the light reaction stage of photosynthesis, and determining the logical relationship through dependency syntax analysis), and converting them into concept-relationship-attribute triples, such as photosynthesis and chloroplasts are connected through the place of occurrence relationship, and forward matching and reverse verification are performed through knowledge graph retrieval, such as physics-heat conduction automatically associated with chemistry-molecular motion, providing an anchor point for subsequent interdisciplinary context generation and forming the core concepts of the subject.

[0056] S02, based on the core concepts of the subject, combined with the preset context database and interdisciplinary knowledge graph, uses the generative AI model to generate customizable project-based learning scenarios.

[0057] Among them, based on real-life situational cases stored in a distributed graph database, each situation can be annotated with a relevance label and a real-world scenario category label. Based on the labels corresponding to the core concepts of the subject, a four-dimensional evaluation system (authenticity / education / narrative / interdisciplinarity, with scores for each dimension calculated based on historical data or manually determined) is used to screen and obtain situations with evaluation scores above a preset threshold. Based on the labels of the core concepts of the subject, situational nodes in the graph are retrieved, such as the chemical-oxidation reaction is associated with the antioxidant situation of apple slices through the food spoilage relationship. The interdisciplinary knowledge graph architecture consists of three layers: the subject concept layer, the interdisciplinary relationship layer, and the situational instance layer. The edge weight is determined by the interdisciplinary association strength (0-1) and the frequency of teaching practice. The situational nodes in the interdisciplinary knowledge graph are retrieved using the labels of the core concepts of the subject, and a generative AI model (such as one based on a large language model architecture such as GPT-4, pre-trained on an education corpus) is used to obtain a structured situational description. The generative AI model controls the generation diversity by adjusting the temperature parameter, outputting 5-8 situation variants each time, forming a customizable project-based learning situation, and realizing a collaborative cognitive closed loop of teacher decision-making and system assistance.

[0058] S03, based on the customized project-based learning scenario, uses the preset teaching design model to generate the corresponding structured teaching activity process template.

[0059] Among them, for the customized project-based learning situation, regular expressions are used to match the cognitive goal keywords in the situation, and the levels are automatically labeled in combination with Bloom's goal taxonomy, such as extracting creative hierarchical goals from making a strawberry fresh-keeping box prototype; according to the teaching objectives, the teacher's guidance points and students' cognitive tasks are labeled; according to the four-what question (what / why / how / what) feature classification, a structured teaching activity process template for the four matching teaching stages is generated, including: problem orientation (such as disassembling the preservation problem preset: known preservation methods / questions to be explored / learning plans) → evidence perception (such as microbial experiments: using standardized tools such as variable control tables and data trend charts to generate experimental reports) → hypothetical products (such as fresh-keeping box design) → speech presentation (such as program defense: new discoveries, problems to be solved, and application expansion).

[0060] S04, based on the teaching activity process template, obtain corresponding teaching resources, cognitive tools and evaluation templates from the preset resource and tool library.

[0061] Among them, different stage types, activity elements, and cognitive levels can be converted into one-hot vectors, input into the BERT model to generate semantic vectors, calculate the similarity with the label vector of each resource in the resource library and map it to the predefined association tags in the resource library, and based on the association tags, use the distributed retrieval engine to match the corresponding teaching resources (experimental videos, case reports), cognitive tools (such as mind map generators, TELOS evaluation tables), and evaluation templates (such as iterative comparison tables of viewpoints) from the resource tool library.

[0062] S05, combine teaching resources, cognitive tools, assessment templates and teaching activity process templates to form an initial lesson plan.

[0063] Among them, the teaching activity process template (such as problem orientation → evidence perception → hypothetical product → speech presentation) is used as the timeline skeleton, and teaching resources and cognitive tools are mounted according to stage nodes, such as mounting the microscope simulation software call entrance in the evidence perception stage; evaluation trigger points are embedded in the process nodes, such as automatically activating the TELOS evaluation form at the end of the hypothetical product stage; generate standardized initial lesson plans, including specific project output paths, such as each stage of the lunch box design project is associated with the corresponding prototype iteration version; integrate at least two digital tools (such as microbial simulator + Gantt chart tool) to form a physical / virtual fusion teaching field; arrange the generated initial lesson plans in chronological order with the four-part structure of real problem → evidence concept → product verification → reflection internalization, and realize the automatic generation of lesson plans.

[0064] The above-mentioned generative AI-assisted teaching plan automation construction method extracts the core concepts of the subject by parsing the curriculum standard documents, avoiding the subjectivity and omissions of manual analysis; based on the core concepts of the subject combined with the preset situation database and the interdisciplinary knowledge map, the generative AI model is used to generate customizable project-based learning situations, and the real-world situations are mapped with multidisciplinary knowledge to solve the problems of weak interdisciplinary integration and insufficient situational authenticity in traditional manual design; the preset teaching design model is used to generate a structured teaching activity process template, clarify the teaching stage sequence, time allocation and teacher-student interaction logic, and improve the standardization and Feasibility: Based on the teaching activity process template, teaching resources, cognitive tools and evaluation templates are matched from the preset resource and tool library, and the quantitative association between resources and the cognitive level of teaching objectives is achieved through label mapping, solving the problem of low resource matching accuracy; the various elements are combined to generate the initial lesson plan, realizing the automation of the entire lesson plan design process, significantly shortening the lesson plan development time, meeting the needs of high-frequency teaching iterations, and breaking through the bottleneck of low lesson plan generation efficiency in traditional semi-automated construction methods. This solution realizes the full chain of theory-design-practice through AI-enabled four-dimensional system design, effectively improving the standardization and feasibility of teaching design. It drives students to create interdisciplinary works (such as programming projects and scientific reports) through generative AI, forming a learning loop of interest activation → practical creation → evaluation iteration, strengthening deep learning and innovation capabilities, enabling teachers to focus on teaching design, reducing repetitive work, and promoting the transformation of their roles to learning ecosystem guides.

[0065] In one embodiment, based on core disciplinary concepts, combined with a pre-set context database and interdisciplinary knowledge graph, a generative AI model is used to generate customizable project-based learning scenarios, including:

[0066] S11, retrieve real-world context data associated with the core concepts of the subject from a preset context database;

[0067] S12, based on the interdisciplinary knowledge graph, associates and maps the core concepts of the discipline with real-world contextual data;

[0068] S13, semantically generates the association mapping results through a generative AI model and outputs multiple project-based learning scenarios;

[0069] In response to the user's selection operation of multiple project-based learning scenarios, a customized project-based learning scenario is generated.

[0070] Specifically, the association retrieval with the core concepts of the subject is realized through the distributed graph database architecture of the preset context database. The database has a large number of real context cases and constructs a three-dimensional label system (subject relevance label, real scene category label, and learning stage adaptation label). The cosine similarity between the core concept and the context label is calculated based on the BERT word vector model, and the real-world context data with an evaluation score ≥4.0 (scoring range 1-5 points, calculated by the arithmetic average of the scores of each dimension or weighted calculation based on historical teaching data) is screened through the four-dimensional evaluation system (authenticity / education / narrative / interdisciplinarity); with the help of the three-layer network architecture of the interdisciplinary knowledge graph (subject concept layer, interdisciplinary relationship layer, context instance layer), the core concept is used as the seed node to retrieve the context nodes within 2 hops. , the mapping logic is verified through the four-question generator (ensuring that the degree of interdisciplinary integration is ≥ 2 subjects) and adapted to the cognitive level of the learning stage; a generative AI model fine-tuned based on the GPT-4 architecture is used to input the core concept-context feature-interdisciplinary relationship triple, and 5-8 context variants (interdisciplinary dimension difference ≥ 1) are generated through the parameter control of Temperature = 0.8; context cards are presented through a visual interactive interface (with a subject correlation radar chart and an interdisciplinary tag cloud), supporting teachers to adjust three-dimensional parameters such as interdisciplinary depth, context complexity, and real-world scenario migration, and generate AI optimization suggestions based on operation history (such as adding interdisciplinary concept associations), ultimately forming a customized context that conforms to the dual-loop structure (linking explicit project processes with implicit knowledge construction). This embodiment reduces the time consumption of context design and improves interdisciplinary integration through the collaboration of AI semantic generation and interdisciplinary graphs, solving the technical problems of insufficient context authenticity and fragmented interdisciplinary integration in manual design.

[0071] In one embodiment, based on the customized project-based learning scenario, a corresponding structured teaching activity process template is generated using a preset teaching design model, including:

[0072] S21, using the pre-set instructional design model, analyze the teaching objectives and activity elements of the customized project-based learning scenario;

[0073] S22, based on the teaching objectives and the preset teaching theory rule base, generates a teaching stage sequence that matches the activity elements through the teaching design model;

[0074] S23, based on the sequence of teaching stages, construct a structured teaching activity process template that includes time allocation, teacher-student interaction logic, and tool calling nodes.

[0075] For example, through the semantic parsing engine of the preset teaching design model, the teaching objectives in the customized context are deconstructed in three dimensions (knowledge and skills / process and methods / emotions, attitudes and values), and the BERT named entity recognition technology is used to extract activity elements (such as experimental exploration and program design) and map them to the preset activity classification system (cognitive / practical / social); based on the Bloom's taxonomy level of teaching objectives (such as the creative level), the matching constructivist teaching strategies (such as scaffolding teaching and anchoring teaching) are retrieved from the teaching theory rule library, and the relationship between activity elements and the four stages of teaching (problem orientation / evidence perception / hypothesis product / speech presentation) is calculated through the Neo4j graph database. The degree of connection (threshold ≥ 0.8) is used to generate a sequence of 5-8 teaching stages; a Gantt chart-style time allocation algorithm is used to automatically allocate the duration of each stage (e.g., the evidence perception stage accounts for 30%-40%) according to the characteristics of the learning stage (40 minutes for elementary school / 90 minutes for high school), and embed teacher-student interaction nodes (e.g., teacher demonstration → student practice → group discussion), with each node associated with a preset cognitive tool (e.g., KWL table generator, CER argumentation tool); a four-dimensional verification mechanism is used to ensure that the process complies with the teaching theory rules (e.g., each stage must include clear cognitive goals and evaluation methods), and a structured template that conforms to the double-loop structure (linking the explicit project process with the implicit knowledge construction) is generated. The AI-driven teaching design model solves the problems of difficulty in implementing teaching theory and fragmented activity logic in manual design, achieving a seamless connection between teaching process and resource tools.

[0076] In one embodiment, based on the teaching activity process template, corresponding teaching resources, cognitive tools and evaluation templates are obtained from a preset resource and tool library, including:

[0077] S31, analyzing the teaching stage types, activity elements and target cognitive levels in the structured teaching activity process template;

[0078] S32, based on the preset matching rules, mapping the teaching stage type, activity elements and target cognitive level to the associated tags in the resource and tool library;

[0079] S33, according to the mapping result of the associated tags, the corresponding teaching resources, cognitive tools and evaluation templates are retrieved and extracted from the resource and tool library.

[0080] For example, the structured process template is semantically deconstructed through a three-dimensional feature parsing engine, and regular expression matching and teaching stage types are used to extract activity elements (such as microbial experiments and prototyping) with the help of named entity recognition technology. The target cognitive level (such as creation and analysis) is automatically labeled based on the Bloom's goal taxonomy, and the subject cognitive requirements of each teaching stage (which can be divided into requirements on the teacher side and the student side) are simultaneously associated; based on the preset matching rule engine, the parsed stage type, activity elements and cognitive level are converted into one-hot vectors, input into the BERT model to generate semantic vectors, and the cosine similarity is calculated with the four-dimensional label vector (authenticity / education / narrative / interdisciplinary) of each resource in the resource tool library, and mapped to the predefined associated tags in the resource library (such as experimental tools "engineering design, process evaluation), where the matching rule library The system integrates the REAL-4Ps-based discipline-focused principle (more than 60% of resources must correspond to disciplinary practice) and the four-stage teaching activity designer specification (e.g., only matching tools such as KWL tables in the problem-oriented stage). Based on the label mapping results, a distributed search engine extracts corresponding teaching resources (e.g., experimental videos and case reports in the disciplinary general template library), cognitive tools (e.g., mind map generators, TELOS evaluation forms), and evaluation templates (e.g., iterative viewpoint comparison forms, 3-2-1 reflection forms) from the resource tool library. A four-dimensional evaluation screening mechanism (resource authenticity score ≥ 3.5) is embedded in the retrieval process, and related resources are expanded through interdisciplinary knowledge graphs (e.g., a chemical antioxidant database recommended for simultaneous biological experiments). AI-driven label mapping and intelligent retrieval address the technical issues of subjective resource matching and fragmented interdisciplinary resources in traditional manual retrieval.

[0081] In one embodiment, the method further comprises:

[0082] S41, real-time collection of teacher-student interaction data, student tool operation logs, and periodic evaluation results during the implementation of the initial teaching plan to generate a multi-dimensional learning behavior dataset;

[0083] S42, perform spatiotemporal alignment and feature annotation on the learning behavior dataset to construct a dynamic learning process map;

[0084] S43, performing a time series analysis on the learning process graph, and identifying learning progress deviations and abnormal knowledge mastery events by combining it with a preset learning ability assessment indicator library;

[0085] S44, based on learning progress deviations and abnormal knowledge mastery events, calls the preset feedback template library and combines the student cognitive model reasoning to generate a structured teaching plan feedback report.

[0086] Specifically, the distributed data collection engine collects teacher-student interaction text data (such as speeches in the discussion area, question-and-answer records), student tool operation logs (such as microscope simulation software usage trajectories, frequency of filling out experimental report forms) and phased evaluation results (such as viewpoint iteration table, 3-2-1 reflection table) in real time, and standardizes and packages them according to timestamps and spatial nodes (such as classrooms / virtual laboratories) to generate a multi-dimensional learning behavior dataset containing behavior sequences, tool call chains, and evaluation scores; the spatiotemporal alignment algorithm is used to calibrate the data set in time sequence and perform feature annotation (such as hypothesis verification, social negotiation, and behavior labels); and the Neo4j graph database is used to construct a dynamic learning process map containing the four-dimensional relationship of students, concepts, tools, and interactions. The node weights are determined by the frequency of behavior and the degree of cognition. The contribution is jointly determined; a sliding window time series analysis is performed on the graph (window size = teaching phase duration), matching the preset learning ability assessment indicator library (including Bloom's goal achievement, knowledge construction progress, and other indicators). The threshold comparison method (such as progress deviation > 15% and abnormal event frequency ≥ 3 times / phase) is used to identify learning progress deviations and knowledge mastery abnormal events; based on the abnormal event type, the preset feedback template library (including progress warnings, concept clarifications, etc.) is called. Combined with the divergence analysis of the probability distribution of students' cognitive states and the target cognitive model, a structured lesson plan feedback report containing deviation location, cause analysis, and optimization suggestions is generated through Bayesian reasoning. The report content adheres to the dual-subject principle of REAL-4Ps (also marking the key points of teacher intervention and the path of student cognitive improvement). Through multi-dimensional data collection and graph analysis, teaching feedback is generated in real time, improving the accuracy of abnormal event identification, providing data support for lesson plan iteration and generative AI model optimization, and realizing closed-loop management of teaching implementation and feedback optimization.

[0087] In one embodiment, the method further comprises:

[0088] S51, based on learning progress deviation and knowledge acquisition abnormal events, optimize the loss value through the following calculation model:

[0089]

[0090] Where L is the loss function, α is the learning progress deviation penalty coefficient, Γ(·) is the learning progress deviation function, Δ t is the learning accuracy deviation data, n is the total number of abnormal types, k is the knowledge abnormal event type index, λ k is the dynamic teaching weight coefficient, ψ(·) is the abnormal quantization function, Δ k is the set of k-th type of knowledge anomaly events, β is the cognitive alignment strength coefficient, D KL is the KL divergence operator, C m is the probability distribution of students’ cognitive states, M ideal It is the target cognitive model;

[0091] S52, based on the optimized loss value, uses the following formula to calculate the updated parameters of the generative AI model:

[0092]

[0093] Among them, θ new is the updated generative AI model parameter vector, θ old is the parameter vector of the generative AI model before updating, η is the adaptive learning rate, is the gradient of the loss function;

[0094] S53, using the updated parameters to update the parameters of the generative AI model, and regenerate the project-based learning scenario based on the updated generative AI model to update the initial lesson plan.

[0095] For example, the loss function L is used to quantify the deviation between teaching practice and model expectation, where the learning progress deviation term α·Γ(Δ t ) Use piecewise function to calculate the time difference penalty between actual progress and preset milestone (for example, if the progress lags by more than 15%, the penalty is calculated by Γ(Δ t ) triggers exponential penalty); knowledge abnormal event items are adjusted by dynamic teaching weight coefficient λ k Weighted ψ(Δ k ) Quantify the impact of various types of anomalies (such as the lambda of concept comprehension error events) k Higher than tool operation error, cognitive alignment term β·D KL (C m ||M ideal ) Use KL divergence to measure the deviation between the distribution of students' cognitive states and the target model; based on the gradient descent algorithm The model's update parameters are calculated, with the adaptive learning rate η dynamically adjusted via the AdaGrad algorithm (initial value 0.01, decaying with the number of iterations). Gradient calculations utilize backpropagation combined with automatic differentiation. When the updated model regenerates project-based learning scenarios, it prioritizes correcting contextual factors that cause deviations (e.g., adjusting context complexity when learning progress lags, strengthening interdisciplinary conceptual connections when knowledge is unusually concentrated). The context descriptions and activity designs in the initial lesson plan are also updated simultaneously. Through model iteration driven by teaching feedback, the accuracy of generative AI's scenario generation is iteratively improved, the rate of interdisciplinary connection deviation is reduced, and a closed-loop evolution based on theory, practice, and optimization, based on REAL-4Ps, is achieved.

[0096] In one embodiment, parsing the curriculum standard document and extracting the core concepts of the subject include:

[0097] S61, obtain standardized curriculum standards data;

[0098] S62, structure the curriculum standard data and identify key knowledge points;

[0099] S63, based on the preset subject label system, maps key knowledge points to core subject concepts.

[0100] Specifically, the curriculum parsing engine can be used to connect to the standardized curriculum documents of the education department, and the latest data can be synchronized in real time. After triple verification (format compliance / content integrity / version consistency), a standardized data set can be formed; the Bi-LSTM+CRF model can be used for word segmentation and part-of-speech tagging, and the named entity recognition (NER) technology can be combined to extract subject terms. The TF-IDF and TextRank algorithms can be used to calculate the term weights, screen high-frequency core expressions and convert them into concept-relationship-attribute triplets and store them in the Neo4j graph database; based on the three-layer ontology structure of the subject label system (subject field → core literacy → knowledge point), the BERT word vector model can be used to calculate the cosine similarity between key knowledge points and labels, and the rule engine can be used to reversely verify the label coverage (each concept matches at least 2 sub-labels), and at the same time, interdisciplinary knowledge graph related concepts (such as physics-heat conduction related chemistry-molecular motion) can be pre-retrieved to generate structured subject core concepts containing basic attributes, cognitive levels, contextual associations and dual-subject annotations. Reduce the time spent on extracting core concepts, improve the completeness of extracted concepts, and increase the accuracy of interdisciplinary associations, so as to meet the requirements of the REAL-4Ps discipline for extracting practical knowledge and provide a precise conceptual foundation for the subsequent generation of project-based learning scenarios.

[0101] The above-mentioned generative AI-assisted teaching plan automatic construction method extracts the core concepts of the subject by parsing the curriculum standard documents to closely follow the teaching objectives, avoiding the subjectivity and omissions of manual analysis; based on the core concepts of the subject combined with the preset situation database and the interdisciplinary knowledge map, the generative AI model is used to generate customizable project-based learning situations, and the real-world situations are mapped with multidisciplinary knowledge to solve the problems of weak interdisciplinary integration and insufficient situational authenticity in traditional manual design; the preset teaching design model is used to generate a structured teaching activity process template including time allocation, teacher-student interaction logic and tool call nodes, clarifying the sequence of teaching stages and improving the standardization and feasibility of teaching design; based on teaching activities The process template uses label mapping to match teaching resources, cognitive tools, and assessment templates from a pre-set resource and tool library, achieving a quantitative correlation between resources and the cognitive level of teaching objectives, addressing the issue of low resource matching accuracy. It combines various elements to generate initial lesson plans, automating the entire lesson plan design process, significantly reducing development time to meet the needs of high-frequency teaching iterations and breaking through the efficiency bottleneck of traditional semi-automated construction methods. Furthermore, by collecting lesson plan implementation data in real time, it constructs a dynamic learning process map, identifies learning progress deviations and knowledge mastery anomalies, and optimizes the parameters of the generative AI model using a loss function, forming a closed-loop teaching feedback loop. This continuously improves the quality and adaptability of lesson plan generation, and enhances the standardization and implementability of teaching design. Using the REAL-4Ps framework, a structured and layered deconstruction of educational complexity is implemented, enabling the continuous evolution of teaching intelligence through dynamic feedback, ultimately building a new educational ecosystem driven by creativity and human-machine collaboration.

[0102] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0103] Based on the same inventive concept, the embodiments of the present application also provide a generative AI-assisted lesson plan automation construction device for implementing the aforementioned generative AI-assisted lesson plan automation construction method. The implementation solution provided by this device is similar to the implementation solution described in the aforementioned method. Therefore, the specific limitations of one or more embodiments of a generative AI-assisted lesson plan automation construction device provided below can be found in the above-mentioned limitations of a generative AI-assisted lesson plan automation construction method, and will not be repeated here.

[0104] In an exemplary embodiment, Figure 2 As shown, a generative AI-assisted teaching plan automatic construction device is provided, including:

[0105] The teaching concept extraction module 101 is used to parse the curriculum standard document and extract the core concepts of the subject;

[0106] The learning scenario generation module 102 is used to generate customizable project-based learning scenarios using a generative AI model based on core concepts of the subject, combined with a preset scenario database and an interdisciplinary knowledge graph;

[0107] The teaching process generation module 103 is used to generate a corresponding structured teaching activity process template based on the customized project-based learning scenario using a preset teaching design model;

[0108] The resource tool matching module 104 is used to obtain corresponding teaching resources, cognitive tools and evaluation templates from a preset resource and tool library based on the teaching activity process template;

[0109] The lesson plan generation module 105 is used to combine teaching resources, cognitive tools, evaluation templates and teaching activity process templates to form an initial lesson plan.

[0110] In one embodiment, the learning scenario generation module 102 is further configured to:

[0111] Retrieve real-world contextual data related to core concepts of the subject from a pre-set contextual database;

[0112] Based on the interdisciplinary knowledge graph, the core concepts of the discipline are mapped with real-world contextual data;

[0113] Generate semantics from the association mapping results through a generative AI model, and output multiple project-based learning scenarios;

[0114] In response to the user's selection operation of multiple project-based learning scenarios, a customized project-based learning scenario is generated.

[0115] In one embodiment, the teaching process generating module 103 is further configured to:

[0116] Utilize pre-set instructional design models to analyze the teaching objectives and activity elements of customized project-based learning scenarios;

[0117] Based on the teaching objectives and the preset teaching theory rule base, the teaching design model is used to generate a teaching stage sequence that matches the activity elements;

[0118] According to the sequence of teaching stages, a structured teaching activity process template is constructed, which includes time allocation, teacher-student interaction logic and tool calling nodes.

[0119] In one embodiment, the resource tool matching module 104 is further configured to:

[0120] Analyze the teaching stage types, activity elements and target cognitive levels in the structured teaching activity process template;

[0121] Based on the preset matching rules, the teaching stage type, activity elements and target cognitive level are mapped to the associated tags in the resource and tool library;

[0122] According to the mapping results of the associated tags, the corresponding teaching resources, cognitive tools and assessment templates are retrieved and extracted from the resource and tool library.

[0123] In one embodiment, a teaching plan feedback module is further included for:

[0124] Real-time collection of teacher-student interaction data, student tool operation logs, and periodic evaluation results during the implementation of the initial teaching plan to generate a multi-dimensional learning behavior dataset;

[0125] Perform spatiotemporal alignment and feature annotation on the learning behavior dataset to construct a dynamic learning process map;

[0126] Conduct time-series analysis on the learning process graph and, combined with a pre-set learning ability assessment indicator library, identify learning progress deviations and abnormal knowledge acquisition events;

[0127] Based on learning progress deviations and abnormal knowledge mastery events, the preset feedback template library is called, and a structured teaching plan feedback report is generated in combination with student cognitive model reasoning.

[0128] In one embodiment, the teaching plan feedback module is further configured to:

[0129] Based on learning progress deviation and knowledge acquisition anomalies, the loss value is optimized by the following calculation model:

[0130]

[0131] Where L is the loss function, α is the learning progress deviation penalty coefficient, Γ(·) is the learning progress deviation function, Δ tis the learning accuracy deviation data, n is the total number of abnormal types, k is the knowledge abnormal event type index, λ k is the dynamic teaching weight coefficient, ψ(·) is the abnormal quantization function, Δ k is the set of k-th type of knowledge anomaly events, β is the cognitive alignment strength coefficient, D KL is the KL divergence operator, C m is the probability distribution of students’ cognitive states, M ideal It is the target cognitive model;

[0132] Based on the optimization loss value, the updated parameters of the generative AI model are calculated using the following formula:

[0133]

[0134] Among them, θ new is the updated generative AI model parameter vector, θ old is the parameter vector of the generative AI model before updating, η is the adaptive learning rate, is the gradient of the loss function;

[0135] The updated parameters are used to update the parameters of the generative AI model, and the project-based learning scenario is regenerated based on the updated generative AI model to update the initial lesson plan.

[0136] In one embodiment, the teaching concept extraction module 101 is further configured to:

[0137] Access to standardized curriculum standards data;

[0138] Structural processing of curriculum standard data to identify key knowledge points;

[0139] Based on the preset subject label system, key knowledge points are mapped to core concepts of the subject.

[0140] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the generative AI-assisted teaching plan automatic construction method as described above are implemented.

[0141] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0142] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The device embodiments described above are merely illustrative, wherein the components described as separate parts may or may not be physically separated, and the parts displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the disclosed solution. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0143] The above-described embodiments merely represent several implementation methods of the embodiments of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person skilled in the art may make various modifications and improvements without departing from the concept of the embodiments of the present application, and these modifications and improvements fall within the scope of protection of the embodiments of the present application.

Claims

1. A generative AI-assisted teaching plan automation construction method, characterized in that: The method comprises: Parse curriculum standard documents and extract core concepts of the subject; Based on the core concepts of the discipline, combined with the preset context database and interdisciplinary knowledge graph, a generative AI model is used to generate customizable project-based learning scenarios; Based on the customized project-based learning scenario, the preset teaching design model is used to generate the corresponding structured teaching activity process template; Based on the teaching activity process template, obtain corresponding teaching resources, cognitive tools and evaluation templates from a preset resource and tool library; Combine the teaching resources, cognitive tools, assessment templates, and teaching activity process templates to form an initial lesson plan.

2. The method according to claim 1, characterized in that Based on the core concepts of the discipline, combined with a preset context database and interdisciplinary knowledge graph, a generative AI model is used to generate customizable project-based learning scenarios, including: Retrieving real-world context data associated with the core concepts of the subject from the preset context database; Based on the interdisciplinary knowledge graph, the core concepts of the discipline are associated with the real-world context data; Perform semantic generation on the association mapping results through a generative AI model to output multiple project-based learning scenarios; In response to the user's selection operation of the multiple project-based learning scenarios, a customized project-based learning scenario is generated.

3. The method according to claim 1, characterized in that The customized project-based learning scenario uses a preset teaching design model to generate a corresponding structured teaching activity process template, including: Utilize the preset instructional design model to analyze the instructional objectives and activity elements of the customized project-based learning scenario; Based on the teaching objectives and a preset teaching theory rule library, generating a teaching stage sequence matching the activity elements through the teaching design model; According to the teaching stage sequence, the structured teaching activity process template including time allocation, teacher-student interaction logic and tool calling nodes is constructed.

4. The method according to claim 1, wherein The step of obtaining corresponding teaching resources, cognitive tools, and evaluation templates from a preset resource and tool library based on the teaching activity process template includes: Analyze the teaching stage types, activity elements and target cognitive levels in the structured teaching activity process template; Based on preset matching rules, the teaching stage type, activity elements and target cognitive level are mapped to associated tags in the resource and tool library; According to the mapping result of the associated tags, corresponding teaching resources, cognitive tools and evaluation templates are retrieved and extracted from the resource and tool library.

5. The method according to claim 1, wherein The method further comprises: Real-time collection of teacher-student interaction data, student tool operation logs, and periodic evaluation results during the implementation of the initial teaching plan to generate a multi-dimensional learning behavior dataset; Performing spatiotemporal alignment and feature annotation processing on the learning behavior dataset to construct a dynamic learning process map; Performing a time series analysis on the learning process graph, and combining it with a preset learning ability assessment index library to identify learning progress deviations and abnormal knowledge mastery events; Based on the learning progress deviation and knowledge mastery abnormal events, the preset feedback template library is called, and a structured teaching plan feedback report is generated in combination with the student cognitive model reasoning.

6. The method according to claim 5, characterized in that The method further comprises: Based on the learning progress deviation and knowledge acquisition abnormal events, the loss value is optimized by the following calculation model: Where L is the loss function, α is the learning progress deviation penalty coefficient, Γ(·) is the learning progress deviation function, Δ t is the learning accuracy deviation data, n is the total number of abnormal types, k is the knowledge abnormal event type index, λ k is the dynamic teaching weight coefficient, ψ(·) is the abnormal quantization function, Δ k is the set of k-th type of knowledge anomaly events, β is the cognitive alignment strength coefficient, D KL is the KL divergence operator, C m is the probability distribution of students’ cognitive states, M ideal It is the target cognitive model; Based on the optimization loss value, the updated parameters of the generative AI model are calculated using the following formula: Among them, θ new is the updated generative AI model parameter vector, θ old is the parameter vector of the generative AI model before updating, η is the adaptive learning rate, is the gradient of the loss function; The updated parameters are used to update the parameters of the generative AI model, and the project-based learning scenario is regenerated based on the updated generative AI model to update the initial teaching plan.

7. The method according to claim 1, characterized in that The said parsing of curriculum standard documents and extraction of core subject concepts include: Access to standardized curriculum standards data; Structuring the curriculum standard data and identifying key knowledge points; Based on the preset subject label system, the key knowledge points are mapped to core concepts of the subject.

8. A generative AI-assisted teaching plan automatic construction device, characterized in that: The device comprises: Teaching concept extraction module, used to parse curriculum standard documents and extract core concepts of the subject; A learning context generation module is used to generate customizable project-based learning contexts based on the core concepts of the subject, combined with a preset context database and interdisciplinary knowledge graph, using a generative AI model; The teaching process generation module is used to generate corresponding structured teaching activity process templates based on customized project-based learning scenarios using the preset teaching design model; A resource tool matching module is used to obtain corresponding teaching resources, cognitive tools and evaluation templates from a preset resource and tool library based on the teaching activity process template; The lesson plan generation module is used to combine the teaching resources, cognitive tools, evaluation templates and teaching activity process templates to form an initial lesson plan.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

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

Images