A case generation management method and system based on artificial intelligence education

Abstract

The application discloses a case generation management method and system based on artificial intelligence education, and belongs to the technical field of computers.The case generation management method and system based on artificial intelligence education are provided, a dynamic candidate case set is generated, environment adaptability analysis and a data-driven closed-loop optimization mechanism are used, the matching of teaching cases and teaching targets and teaching environments is realized, and the quality of cases is continuously optimized through multidimensional data feedback.The case generation management method and system based on artificial intelligence education have the advantages that the dynamic teaching environment adaptation and the closed-loop optimization mechanism can solve the problems of low matching degree of case generation and teaching targets, poor environment adaptability and lack of data-driven optimization to a certain extent, and the teaching cases are more targeted and the teaching effect is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer technology, in particular to a case generation management method and system based on artificial intelligence education. BACKGROUND

[0002] In the field of artificial intelligence education technology, using generative models to automatically create teaching cases has become an important means to improve teaching efficiency. Existing technologies usually focus on the generation algorithm of the case itself, such as automatically generating corresponding scenarios or questions by inputting knowledge points. However, this generation process is often one-way and static, that is, once the case is generated, it is considered as the final product, and the subsequent actual teaching effect cannot be effectively fed back to the system to guide the optimization of the initial generation process.

[0003] The current method has a significant defect: the generation, screening, deployment and optimization of cases are disconnected. Specifically, the system lacks a closed-loop management mechanism and cannot use the massive learning interaction data generated after the application of the case in the real teaching environment (such as student behavior, cognitive state changes, etc.) as optimization basis to iteratively and improve the generated case. This leads to the fact that the generated case may be divorced from the actual teaching scene requirements and it is difficult to continuously guarantee its teaching effectiveness and adaptability.

[0004] Therefore, the core problem to be solved by the existing technology is how to realize the whole life cycle closed-loop management of teaching cases from generation to optimization, so that the case can evolve and continuously improve based on real teaching feedback data, thereby dynamically improving its teaching value. SUMMARY

[0005] The embodiments of the present application provide a case generation management method and system based on artificial intelligence education, which can solve the problems of low matching degree between case generation and teaching target, poor environmental adaptability and lack of data-driven optimization to a certain extent, improve the pertinence and teaching effect of teaching cases, and the technical solutions are as follows:

[0006] On the one hand, a case generation management method based on artificial intelligence education is provided, and the method comprises:

[0007] In response to a teaching case generation request for target teaching content, target knowledge point description information, teaching target description information and a constraint condition set corresponding to the target teaching content are determined, the constraint condition set being used to constrain the difficulty level, interactive form and evaluation standard of the generated teaching case; based on the target knowledge point description information, the teaching target description information and the constraint condition set, a candidate teaching case set of the target teaching content is generated; an initial teaching case is determined from the candidate teaching case set, and the initial teaching case is deployed to a target learning environment; the initial teaching case is optimized based on learning interaction data flow obtained from the target learning environment to obtain a target teaching case of the target teaching content, the learning interaction data flow including a behavior log sequence, cognitive state change data and environmental context information.

[0008] Further, the application also proposes that based on the target knowledge point description information, a pre-constructed domain adaptive knowledge graph is queried, and a semantic network subgraph corresponding to the target knowledge point description information is extracted from the domain adaptive knowledge graph, the semantic network subgraph including a core knowledge point node, an associated knowledge point node, a cognitive path sequence and a common misconception concept set, the domain adaptive knowledge graph being an educational domain knowledge representation structure composed of a knowledge point ontology layer, a teaching relationship layer and a cognitive model layer; based on the teaching target description information and the cognitive path sequence in the semantic network subgraph, a target generation strategy template is obtained from a generation strategy rule library; based on the semantic network subgraph, the target generation strategy template and the constraint condition set, a candidate teaching case set is generated.

[0009] Further, the application also proposes that the teaching target description information is subjected to teaching strategy intention recognition to obtain a teaching strategy vector representation, the teaching strategy vector representation being used to indicate the relative weight proportion of knowledge imparting, ability cultivation and literacy improvement; based on the teaching strategy vector representation and the cognitive path sequence in the semantic network subgraph, a matching degree score of each strategy template in the generation strategy rule library is determined, the matching degree score being a quantitative indicator of teaching strategy fitness and cognitive path consistency; the strategy templates in the generation strategy rule library are sorted according to the matching degree score, the top N strategy templates with the highest matching degree score are selected as a candidate strategy template set, N being a positive integer; each strategy template in the candidate strategy template set is subjected to teaching adaptability analysis to generate an adaptability evaluation result, the adaptability evaluation result being used to represent the compatibility of the strategy template with the difficulty level and the interactive form in the constraint condition set; based on the matching degree score and the adaptability evaluation result, a target generation strategy template is determined from the candidate strategy template set.

[0010] Further, the application also proposes that, based on the core knowledge point nodes and the associated knowledge point nodes in the semantic network subgraph, a basic case content set is generated through a neural network generation model, the basic case content set including multiple initial case texts meeting knowledge point coverage requirements; based on the content structure templates and the interactive design templates in the target generation strategy templates, the basic case content set is structurally reorganized to generate a standardized case framework set, each case framework in the standardized case framework set including content partitions and interactive node designs; based on the difficulty levels and the evaluation standards in the constraint condition set, teaching adaptability verification is performed on the standardized case framework set through a rule engine to obtain a verified case framework set, the teaching adaptability verification including difficulty consistency checking and evaluation feasibility analysis; multi-dimensional quality evaluation is performed on the verified case framework set to obtain a case quality score matrix, the case quality score matrix including content accuracy scores, teaching effectiveness scores, and technical feasibility scores; based on the case quality score matrix, a first multi-objective optimization algorithm is used to screen a candidate teaching case set meeting preset quality conditions from the verified case framework set.

[0011] Further, the application also proposes that, the environment parameter configuration of the target learning environment is obtained, the environment parameter configuration including hardware device parameters, network environment parameters, and learning terminal types; based on the environment parameter configuration, teaching environment adaptability analysis is performed on each candidate case in the candidate teaching case set to generate an environment adaptability score, the environment adaptability score being used to represent the running compatibility and interactive fluency of the candidate case in the target learning environment; multi-dimensional teaching effect prediction evaluation is performed on the candidate teaching case set to obtain a teaching effect prediction score, the teaching effect prediction evaluation including knowledge mastery prediction, ability improvement prediction, and participation prediction; based on the environment adaptability score and the teaching effect prediction score, a second multi-objective decision algorithm is used to determine the comprehensive evaluation score of each candidate case, the second multi-objective decision algorithm dynamically adjusting the weight distribution of different score dimensions according to the teaching target description information; based on the comprehensive evaluation score, the candidate teaching case set is sorted to generate a candidate case sorting list, and an initial teaching case is determined from the sorted candidate cases based on a preset selection strategy.

[0012] Further, the application also proposes to construct a virtual test environment based on the environmental parameter configuration, and deploy a digital twin teaching scenario in the virtual test environment, the digital twin teaching scenario including hardware configuration, network topology and terminal device simulator consistent with the target learning environment; run each candidate teaching case in the digital twin teaching scenario, and collect case running data in real time, the case running data including rendering frame rate time sequence data, interaction response delay sequence and resource occupation fluctuation curve; perform multi-scale feature extraction on the case running data of each candidate teaching case to obtain a time domain feature set, a frequency domain feature set and a nonlinear dynamics feature set of each candidate teaching case, the nonlinear dynamics feature set including Lyapunov exponent, fractal dimension and entropy value index; based on the time domain feature set, the frequency domain feature set and the nonlinear dynamics feature set of each candidate teaching case, generate an environmental adaptability score of each candidate teaching case.

[0013] Further, the application also proposes to determine a response stability index of each candidate teaching case under different load conditions based on the interaction response delay sequence in the time domain feature set of each candidate teaching case, the response stability index being used to represent the predictability and consistency of case interaction response; determine a resource occupation mode when each candidate teaching case runs based on the energy spectrum distribution feature in the frequency domain feature set of each candidate teaching case, and generate a resource use efficiency score, the resource use efficiency score reflecting the optimized utilization degree of hardware resources by the case; extract the fractal dimension and entropy value index in the nonlinear dynamics feature set to construct a system complexity descriptor, the system complexity descriptor being used to quantify the adaptive ability and robustness of each candidate teaching case in a dynamic environment; based on the hardware performance benchmark and network quality requirement in the environmental parameter configuration, establish an environmental adaptability evaluation benchmark model; input the response stability index, the resource use efficiency score and the system complexity descriptor into the environmental adaptability evaluation benchmark model, and generate an environmental adaptability score of each candidate teaching case through a weighted fusion algorithm.

[0014] Further, the application also proposes to perform spatio-temporal alignment and feature fusion processing on the learning interaction data stream to obtain a multi-dimensional feature tensor, the multi-dimensional feature tensor including time sequence behavior feature, spatial attention feature and cognitive state feature; process the multi-dimensional feature tensor through a causal inference algorithm to obtain a case effect evaluation report and an optimization suggestion set, the case effect evaluation report being a structured evaluation result; based on the case effect evaluation report and the optimization suggestion set, generate a case optimization instruction sequence, the case optimization instruction sequence including content adjustment instruction, interaction improvement instruction and difficulty adjustment instruction; execute the case optimization instruction sequence to optimize the initial teaching case to obtain a target teaching case.

[0015] Further, the application also proposes that the multi-dimensional feature tensor is subjected to causal structure learning to obtain a teaching effect causal network diagram, the teaching effect causal network diagram comprising behavior feature nodes, cognitive feature nodes, environment feature nodes, and causal relationship edges between the nodes; based on the teaching effect causal network diagram, a counterfactual reasoning algorithm is used to determine the causal contribution of each feature node to the teaching effect index, and the causal contribution is used to quantify the influence intensity of each feature factor on the teaching effect; based on the causal contribution, key feature factors and bottleneck feature factors affecting the teaching effect are identified, the key feature factors being features having significant positive influence on the teaching effect, and the bottleneck feature factors being features having constraints on the teaching effect; based on the analysis results of the key feature factors and the bottleneck feature factors, a case effect evaluation report and an optimization suggestion set comprising a teaching effect diagnosis conclusion and targeted improvement suggestions are generated in combination with the teaching target description information.

[0016] Further, the application also proposes that the case effect evaluation report is subjected to structured analysis, and teaching effect bottleneck features and optimization priority indexes are extracted, the optimization priority indexes being determined based on the causal contribution and the teaching effect influence degree; based on the teaching effect bottleneck features, a corresponding optimization operation template set is matched from a case optimization rule library, the optimization operation template set comprising executable modification operation sequences for different bottleneck types; according to the optimization priority indexes, the optimization operation template set is sorted and filtered to obtain a reference operation template in the optimization operation template set, and the template parameters of the reference operation template are initialized based on the teaching target description information; the initialized reference operation template is subjected to sequence optimization by a dynamic programming algorithm to obtain an optimization instruction execution sequence, the dynamic programming algorithm taking maximizing the bottleneck feature improvement effect and minimizing the content modification range as the optimization objective; the optimization instruction execution sequence is subjected to dual verification of teaching logic consistency and technical implementation feasibility, and the optimization instruction execution sequence is fine-tuned according to the verification results to obtain a case optimization instruction sequence.

[0017] In one aspect, a case generation management system based on artificial intelligence education is provided, and the system comprises:

[0018] A determination module is configured to determine target knowledge point description information, teaching target description information, and a constraint condition set corresponding to target teaching content in response to a teaching case generation request for the target teaching content, the constraint condition set being used to constrain the difficulty level, interactive form, and evaluation standard of the generated teaching case;

[0019] A case set generation module is configured to generate a candidate teaching case set of the target teaching content based on the target knowledge point description information, the teaching target description information, and the constraint condition set;

[0020] a deployment module configured to determine an initial teaching case from the candidate teaching case set and deploy the initial teaching case to a target learning environment;

[0021] an optimization module configured to optimize the initial teaching case based on a learning interaction data stream obtained from the target learning environment to obtain a target teaching case of the target teaching content, the learning interaction data stream including a behavior log sequence, cognitive state change data, and environmental context information.

[0022] In one aspect, a computer device is provided, which includes one or more processors and one or more memories, and at least one computer program is stored in the one or more memories, and the computer program is loaded and executed by the one or more processors to implement the case generation management method based on artificial intelligence education.

[0023] In one aspect, a computer readable storage medium is provided, which stores at least one computer program, and the computer program is loaded and executed by a processor to implement the case generation management method based on artificial intelligence education.

[0024] In one aspect, a computer program product or computer program is provided, which includes program code stored in a computer readable storage medium, and a processor of a computer device reads the program code from the computer readable storage medium, and the processor executes the program code to make the computer device execute the case generation management method based on artificial intelligence education.

[0025] As can be seen from the above, the case generation management method and system based on artificial intelligence education provided by the present application realize the matching of teaching cases, teaching targets and teaching environments through a dynamic candidate case set generation, environmental adaptability analysis and data-driven closed-loop optimization mechanism, and continuously optimize the quality of cases through multi-dimensional data feedback, and have the advantages of dynamic teaching environment adaptation and closed-loop optimization mechanism, which to some extent solve the problems of low matching degree of case generation and teaching targets, poor environmental adaptability and lack of data-driven optimization, and improve the pertinence of teaching cases and teaching effect. BRIEF DESCRIPTION OF DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0027] Figure 1This is a schematic diagram of the implementation environment of a case generation and management method for artificial intelligence-based education provided in an embodiment of this application;

[0028] Figure 2 This is a flowchart of a case generation and management method for artificial intelligence-based education provided in an embodiment of this application;

[0029] Figure 3 This is a partial flowchart of a case generation and management method for artificial intelligence-based education provided in an embodiment of this application;

[0030] Figure 4 This is a partial flowchart of another case generation and management method for artificial intelligence-based education provided in an embodiment of this application;

[0031] Figure 5 This is a partial flowchart of another case generation and management method based on artificial intelligence education provided in the embodiments of this application;

[0032] Figure 6 This is a schematic diagram of the structure of a case generation and management system for artificial intelligence-based education provided in an embodiment of this application;

[0033] Figure 7 This is a schematic diagram of the structure of a server provided in an embodiment of this application. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0035] In this application, the terms "first," "second," etc., are used to distinguish identical or similar items with essentially the same function. It should be understood that there is no logical or temporal dependency between "first," "second," and "nth," nor are there any restrictions on quantity or execution order.

[0036] Artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain better results.

[0037] Machine Learning (ML) is a multidisciplinary field that involves probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory, and many other disciplines. It specifically studies how computers can simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge sub-models to continuously improve their performance.

[0038] Normalization: mapping a series of values with different ranges to the interval (0, 1) to facilitate data processing. In some cases, the normalized values can be directly implemented as probabilities.

[0039] Attention weight: can represent the importance of certain data in the training or prediction process, indicating the size of the input data's impact on the output data. Data with high importance has a higher value of attention weight, and data with low importance has a lower value of attention weight. In different scenarios, the importance of data is not the same, and the process of training attention weight of the model is also the process of determining the importance of data.

[0040] Target knowledge point description information: the target knowledge point description information is a structured data object used to uniquely identify and describe the teaching content unit targeted by the generated teaching case, which at least includes a knowledge point identifier, a knowledge point name and a knowledge point content abstract.

[0041] Teaching goal description information: the teaching goal description information is a data set used to indicate the expected teaching results achieved through the teaching case, which is used to represent the target focus degree in the dimensions of knowledge transmission, ability cultivation and literacy improvement.

[0042] Target learning environment: the target learning environment is the hardware and software integrated environment in which the initial teaching case is deployed and run, which is defined by specific hardware devices, network configurations, operating system platforms and teaching support software.

[0043] Behavior log sequence: a timestamped record set used to describe the sequence of operation events in the learner's interaction with the teaching case.

[0044] Cognitive state change data: quantitative data sequence representing the learner's knowledge mastery or thinking state change inferred by analyzing interaction behavior.

[0045] Environmental context information: a parameter set used to describe the external environment conditions when the learning process occurs, including device state, network condition and physical environment parameter.

[0046] Domain adaptive knowledge graph: the domain adaptive knowledge graph is a semantic network knowledge representation structure specific to the education field, which defines concepts and attributes through the knowledge point ontology layer, defines teaching logic associations through the teaching relationship layer, and describes the cognitive laws of learners through the cognitive model layer.

[0047] Core knowledge point node: the central node in the semantic network subgraph that directly corresponds to the target knowledge point description information.

[0048] Associated knowledge point nodes: Related knowledge point nodes connected to the core knowledge point node through predefined relationships in the semantic network subgraph.

[0049] Cognitive path sequence: A learning path composed of a series of associated knowledge point nodes, ordered based on teaching logic and cognitive laws.

[0050] Common misconception concept set: A set of identifiers of common misconceptions or concept confusions commonly associated with a specific knowledge point, induced from teaching experience data.

[0051] Teaching strategy intent recognition: A natural language processing process that analyzes teaching goal description information to determine its implied preferred teaching method.

[0052] Teaching strategy fit degree: A numerical indicator used to quantify the degree of match between the teaching method implied by the generated strategy template and the recognized teaching strategy intent.

[0053] Cognitive path consistency: A numerical indicator used to quantify the degree of agreement between the activity flow suggested by the generated strategy template and the cognitive path sequence in the knowledge graph.

[0054] Structured reorganization: A data processing process that organizes, segments, and embeds interactive elements into the generated initial case text according to predefined content structure templates and interaction design templates.

[0055] Rule engine: A software component used to execute predefined business rule logic, specifically used in this method to verify whether the case framework meets the difficulty and evaluation criteria.

[0056] First multi-objective optimization algorithm: An optimization algorithm used to balance multiple conflicting objectives such as content accuracy, teaching effectiveness, and technical feasibility, and to select candidate cases from the Pareto optimal solution set.

[0057] Teaching environment adaptability analysis: A process used to evaluate the running compatibility, performance, and interaction fluency of candidate teaching cases in a specific target learning environment.

[0058] Multi-dimensional teaching effectiveness prediction evaluation: A predictive calculation based on case characteristics and historical data, evaluating the potential teaching effectiveness of a case in multiple dimensions such as knowledge mastery, ability improvement, and participation.

[0059] Second multi-objective decision algorithm: A decision algorithm used to integrate environment adaptability scores and teaching effectiveness prediction scores, and dynamically adjust weights based on teaching goals to calculate a comprehensive evaluation score.

[0060] Virtual testing environment: an isolated testing platform that is computationally and interactively equivalent to the target learning environment, simulated by software.

[0061] Digital twin teaching scenario: a dynamic simulation model that is functionally and behaviorally consistent with the target learning environment, constructed in the virtual testing environment.

[0062] Multi-scale feature extraction: an analysis technique that extracts features reflecting short-term, medium-term, and long-term patterns from time series data simultaneously.

[0063] Time-domain feature set: a set of features directly obtained from time series data, such as mean, variance, and delay.

[0064] Frequency-domain feature set: a set of features obtained after converting time series data to the frequency domain, such as energy spectrum distribution and dominant frequency.

[0065] Nonlinear dynamics feature set: a set of features used to characterize the intrinsic randomness and complexity of a system, including Lyapunov exponent, fractal dimension, and entropy index.

[0066] Lyapunov exponent: a dynamic index used to quantify the dependence of a system on initial conditions and judge the chaotic properties of a system.

[0067] Fractal dimension: a fractional dimension used to measure the geometric complexity of a time series trajectory or system attractor.

[0068] Entropy index: a statistical index used to measure the degree of disorder or information uncertainty of a system.

[0069] Interaction response delay sequence: a time series data recording the time required for a case to respond to the learner's continuous operations.

[0070] Energy spectrum distribution feature: the distribution of signal energy at different frequency components after Fourier transform from time domain to frequency domain of the interaction response delay sequence.

[0071] System complexity descriptor: an index based on the nonlinear dynamics feature set, used to comprehensively evaluate the complexity and robustness of the case's running behavior.

[0072] Environment adaptability evaluation benchmark model: a mathematical model or rule set based on environment parameter configuration, used to map multiple performance indicators into a unified adaptability score.

[0073] Causal structure learning: a machine learning method that infers the causal relationship direction and graph structure between variables from observed data.

[0074] Counterfactual reasoning algorithm: a counterfactual reasoning algorithm for estimating how the teaching effectiveness index will change when a certain feature changes.

[0075] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data for analysis, stored data, displayed data, etc.) and signals involved in the present application are authorized by the user or fully authorized by all parties, and the collection, use and processing of related data need to comply with relevant laws, regulations and standards of relevant countries and regions.

[0076] Figure 1 is an implementation environment schematic diagram of a case generation management method based on artificial intelligence education provided by an embodiment of the present application, referring to Figure 1 The implementation environment can include a node 110 and a server 140.

[0077] The node 110 is connected to the server 140 through a wireless network or a wired network. Optionally, the node 110 is a smart phone, a tablet computer, a notebook computer, a desktop computer, etc., but is not limited thereto. The node 110 installs and runs an application program supporting case generation management based on artificial intelligence education.

[0078] The server 140 is a stand-alone physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery network (CDN) and big data and artificial intelligence platforms, etc. Basic cloud computing services. The server 140 can provide background services for the application program running on the node 110.

[0079] In the related art, the use of generative models to automatically create teaching cases has become an important means to improve teaching efficiency in the field of artificial intelligence education technology. Existing methods usually focus on the case generation algorithm itself, such as automatically generating scenarios or questions by inputting knowledge points. However, this generation process is often one-way and static, and once the case is generated, it is considered as the final product, and the subsequent actual teaching effect cannot be effectively fed back to the system to guide optimization. The generation, screening, deployment and optimization of cases are disconnected from each other, lack of closed-loop management mechanism, and cannot use learning interaction data generated in real teaching environment as optimization basis, resulting in that the generated cases may deviate from the actual teaching scene demand, and it is difficult to continuously guarantee the teaching effectiveness and adaptability. For example, the cases generated by the online programming teaching platform may cause interaction delay or rendering lag due to not considering the performance difference of students' devices, affecting the learning experience.

[0080] To solve the above problems, the inventors find that the related art cannot integrate the whole life cycle of teaching cases into a unified management framework, leading to disconnection between case generation and optimization. By analyzing dynamic feedback data in teaching scenarios, such as student behavior logs, cognitive state changes and environmental context information, it is found that these data contain key clues for case optimization. Therefore, it is necessary to build a closed-loop management mechanism to integrate case generation, environmental adaptation deployment and dynamic optimization, and form iterative evolution capabilities. The specific idea is as follows: first, generate candidate cases based on knowledge points and teaching goals, and determine the initial case through environmental adaptability screening; then collect multi-dimensional interaction data in real teaching environment, and use data-driven methods to optimize the case, forming a closed-loop iterative path of generation-deployment-optimization.

[0081] Therefore, the present application proposes the following technical solutions, taking the server as an example for the execution subject, referring to Figure 2 , comprising the following steps.

[0082] 201, in response to a teaching case generation request for target teaching content, determining target knowledge point description information, teaching goal description information and constraint condition set corresponding to the target teaching content;

[0083] 202, generating a candidate teaching case set of the target teaching content based on the target knowledge point description information, the teaching goal description information and the constraint condition set;

[0084] 203, determining an initial teaching case from the candidate teaching case set, and deploying the initial teaching case to a target learning environment;

[0085] 204, optimizing the initial teaching case based on the learning interaction data stream obtained from the target learning environment to obtain a target teaching case of the target teaching content.

[0086] The constraint condition set is used to constrain the difficulty level, interaction form and evaluation standard of the generated teaching case, the learning interaction data stream includes behavior log sequence, cognitive state change data and environment context information, the target knowledge point description information refers to the structured description of the core knowledge points involved in the teaching content and their associated relationships, which can be realized by using the semantic network subgraph in the knowledge graph to ensure that the case content corresponds to the knowledge system. The teaching goal description information refers to the quantitative index definition of the teaching goal, which can be represented by a multi-dimensional vector, such as converting knowledge mastery, ability cultivation and literacy improvement goals into weight proportions to guide the case generation direction. The constraint condition set refers to the technical limitations and teaching requirements of case generation, which can be realized by using the preset condition set in the rule engine, such as limiting the case interaction form to graphical or text question and answer to ensure that the generated case meets the actual environmental carrying capacity. The candidate teaching case set refers to multiple candidate schemes generated by the generation model and strategy template, which can be realized by using a neural network generation model combined with a rule reorganization method to generate diversified initial case frameworks. The initial teaching case is deployed to the target learning environment, which means that the filtered case is adapted to the specific teaching scene and run, which can be realized by using digital twin technology to build a virtual test environment to verify the compatibility of the case on real devices. The learning interaction data stream optimization refers to adjusting the case content based on the real-time collected student behavior data and environmental parameters, which can be realized by using causal reasoning algorithm to analyze the key influencing factors in the data and generate targeted optimization instructions.

[0087] Specifically, when receiving a teaching case generation request, first parse the knowledge point structure corresponding to the target teaching content, extract the semantic network subgraph in the knowledge graph, and ensure that the case covers the core knowledge points and associated concepts. Combine the multi-dimensional vector representation of the teaching goal to match the adaptive strategy template from the generation strategy library, such as selecting an interactive case structure that focuses on practical ability cultivation. Through the difficulty level and evaluation standard in the constraint condition set, the technical feasibility of the generated initial case is verified, such as checking the running smoothness of the case on low-configuration terminals. After deploying the filtered initial case to the actual teaching environment, real-time collection of student operation sequences, cognitive evaluation results and device running state data during the case learning process is performed. Use spatiotemporal alignment technology to fuse multi-source heterogeneous data into a multi-dimensional feature tensor, identify the key factors affecting teaching effectiveness through causal reasoning, such as finding that the response delay of a specific interaction node causes student distraction. Based on the analysis results, optimization instructions are generated, such as adjusting the media resource resolution in the case or simplifying complex operation steps, forming the iterative target teaching case.

[0088] Compared with the related art, the existing method only focuses on the algorithm optimization of the case generation stage, and does not establish a data closed loop of generation and optimization, resulting in that the case cannot adapt to the dynamic teaching demand. For example, after the traditional system generates a mathematical application question, it cannot adjust the difficulty distribution of the question according to the error mode of the student's answering. While the scheme can locate the case design defects through real-time data collection and causal analysis after deployment, for example, identify the setting problem of the easily confused options in the question, and automatically optimize the option arrangement logic. In addition, the related art usually ignores the difference of the teaching environment, and the same case may have compatibility problems on different terminal devices, while the scheme can ensure the running stability after deployment through the digital twin test environment to verify the case adaptability.

[0089] Through the above technical solutions, the present application realizes the closed-loop management of the whole life cycle of the teaching case, and organically integrates the case generation, environment adaptation and dynamic optimization links. Based on the multi-dimensional feedback data in the real teaching environment, the case content and interactive design can be continuously optimized, solving the problem of the separation of the case generation and optimization links in the traditional method. Through the constraint condition set and environment adaptability analysis, it is ensured that the generated case not only meets the requirements of the teaching target, but also adapts to the technical conditions of the specific teaching scene. The spatio-temporal fusion and causal reasoning of the learning interaction data stream provide a data-driven decision basis for case optimization, and improve the adaptability and effectiveness of the case in the teaching practice.

[0090] The present application further proposes the following technical solutions, see Figure 3 Taking the server as an example, the execution subject includes the following steps.

[0091] 301, based on the target knowledge point description information, query the pre-constructed domain adaptive knowledge graph, and extract the semantic network subgraph corresponding to the target knowledge point description information from the domain adaptive knowledge graph;

[0092] 302, based on the teaching target description information and the cognitive path sequence in the semantic network subgraph, obtain the target generation strategy template from the generation strategy rule library;

[0093] 303, based on the semantic network subgraph, the target generation strategy template and the constraint condition set, generate a candidate teaching case set.

[0094] The semantic network subgraph includes a core knowledge point node, an associated knowledge point node, a cognitive path sequence, and a common misconception concept set. The domain adaptive knowledge graph is an educational domain knowledge representation structure composed of a knowledge point ontology layer, a teaching relationship layer, and a cognitive model layer. The domain adaptive knowledge graph is an educational domain knowledge representation structure composed of a knowledge point ontology layer, a teaching relationship layer, and a cognitive model layer. Specifically, the hierarchical relationship between knowledge points is constructed by using an ontology modeling tool. The teaching relationship layer is established by using a teaching logic reasoning engine. The cognitive model layer is constructed by combining a cognitive diagnosis model, which is used to realize the structured expression of the knowledge system and the fusion of cognitive laws. The semantic network subgraph is a substructure extracted from the knowledge graph, which includes a core knowledge point node, an associated knowledge point node, a cognitive path sequence, and a common misconception concept set. Specifically, a graph neural network algorithm is used for subgraph extraction to ensure the knowledge coverage integrity and cognitive path scientificity of the candidate cases. The generation strategy rule library is a database that stores various teaching strategy templates. Specifically, a template indexing mechanism based on a teaching strategy classification tree is used to dynamically match and generate strategies according to the teaching target.

[0095] Specifically, when receiving a teaching case generation request, first, the semantic network subgraph related to the target knowledge point is obtained by querying the knowledge graph. The subgraph includes the association relationship between knowledge points and cognitive path data. Then, according to the teaching target description information, the strategy intention is identified, and the abstract teaching target is converted into a strategy vector. The cognitive path sequence is combined to filter the generation strategy template with the highest matching degree from the rule library. Finally, the structured knowledge of the semantic subgraph, the teaching design rules of the strategy template, and the difficulty level and interactive form in the constraint condition are fused in multiple dimensions, and a candidate case set is generated in batches by the content generation engine. In this process, the hierarchical structure of the knowledge graph ensures the knowledge accuracy of the case content, and the dynamic matching mechanism of the strategy template realizes the docking of the teaching intention and the case design.

[0096] Compared with related technologies, the traditional method only relies on keyword matching to generate case content, lacks systematic modeling of the teaching logic relationship between knowledge points, and leads to fragmentation of case content and difficulty in adapting to dynamic teaching targets. The present scheme realizes the explicit expression of the association relationship between knowledge points and the cognitive path by extracting the semantic network subgraph of the domain knowledge graph, and combines the dynamic filtering mechanism of the strategy template to make the generated candidate cases meet the requirements of the integrity of the knowledge system and flexibly adapt to different teaching strategy needs.

[0097] By the technical solution, the technical problem of dynamic adaptation of the teaching case generation process to the teaching target and the knowledge structure is solved. Through the collaborative mechanism of the knowledge graph and the strategy template, it is ensured that the candidate case set covers the complete knowledge system and meets the preset teaching strategy requirements. For example, in the teaching of mathematical geometry, the system can automatically match the inquiry learning strategy template according to the teaching target of "cultivating spatial reasoning ability", and generate candidate cases containing interactive modeling tools in combination with the cognitive path of geometric knowledge points, thereby improving the case generation efficiency and teaching adaptability.

[0098] The application further proposes a method for obtaining a target generation strategy template from a generation strategy rule base based on teaching target description information and a cognitive path sequence in a semantic network subgraph, including performing teaching strategy intention recognition on the teaching target description information to obtain a teaching strategy vector representation, determining a matching degree score of the strategy template based on the vector representation and the cognitive path sequence, selecting the top N candidate strategy templates according to the score ranking, performing teaching adaptability analysis to generate an adaptability evaluation result, and finally determining the target generation strategy template based on the score and the evaluation result.

[0099] The teaching strategy vector representation refers to the decomposition of the teaching target into the weight proportions of the three dimensions of knowledge transmission, ability cultivation and literacy improvement, which can be specifically implemented by using an intention recognition model in natural language processing combined with a semantic analysis algorithm, and is used for quantifying the strategic orientation of the teaching target. The matching degree score refers to the cosine similarity between the strategy template and the teaching strategy vector, which is weighted and summed by combining the coverage index of the cognitive path sequence, and can be specifically implemented by using a multi-dimensional similarity evaluation algorithm, and is used for measuring the degree of fit between the strategy template and the teaching target. The candidate strategy template set refers to the top N strategy templates with the highest matching degree reserved by sorting and screening, which can be specifically implemented by using a Top-N sorting algorithm combined with a dynamic threshold adjustment mechanism, and is used for balancing the strategy diversity and selection efficiency. The teaching adaptability analysis refers to verifying the compatibility of the strategy template with the difficulty level and the interactive form in the constraint condition, which can be specifically implemented by using a rule engine combined with a constraint satisfaction degree evaluation model, and is used for ensuring the implementability of the generated strategy in the actual teaching environment.

[0100] Specifically, the method converts abstract teaching goals into three-dimensional weight indicators through a teaching strategy vector representation, establishing a quantifiable teaching intent model. In combination with the cognitive path sequence in the semantic network, the logical structure of the knowledge system is considered simultaneously in the strategy matching process, and a matching degree scoring mechanism is used to screen out preliminary candidate strategies. Subsequently, the compatibility of the candidate strategies with the teaching environment constraints is verified through teaching adaptability analysis, and finally a dual evaluation mechanism is used to comprehensively match the adaptability results and the matching degree, realizing the dynamic balance of the generated strategy between the teaching goal orientation and environmental adaptability. For example, when the teaching goal focuses on ability training, the weight of the ability dimension in the vector representation is increased, and strategies containing practical links are preferentially matched, while virtual simulation schemes that do not meet the current equipment conditions are excluded through adaptability analysis.

[0101] Compared with related technologies, the existing scheme usually uses keyword matching or fixed rule selection to generate strategies, which cannot dynamically adapt to changes in teaching goal weights and lacks consideration of the logic of cognitive paths. The present scheme realizes the matching of strategy selection and teaching goals through vector representation and dynamic scoring mechanism, ensures that the strategy conforms to the learning rules in combination with the cognitive path sequence, and solves the compatibility problem of the strategy and the teaching environment through adaptability analysis, improving the effectiveness and implementability of the generated strategy.

[0102] Through the above technical solutions, the present application solves the problem of insufficient dynamic matching of teaching case generation strategies and teaching goals, realizes the intelligence and adaptability of strategy selection through vectorized intent recognition and a dual evaluation mechanism. At the same time, the compatibility of the strategy and the constraint conditions is verified through teaching adaptability analysis, to a certain extent, solving the technical defects of poor adaptability of the generated strategy to the teaching environment, ensuring that the generated case conforms to the teaching goal and can be smoothly implemented in the actual environment.

[0103] The present application further proposes generating a candidate teaching case set based on a semantic network subgraph, a target generation strategy template, and a constraint condition set, including: generating a basic case content set through a neural network generation model based on the core knowledge point nodes and the associated knowledge point nodes in the semantic network subgraph; structuring and recombining the basic case content set based on the content structure template and the interaction design template in the target generation strategy template to generate a standardized case framework set; verifying the teaching adaptability of the standardized case framework set based on the difficulty level and the evaluation standard in the constraint condition set through a rule engine; performing multi-dimensional quality evaluation on the verified case framework set to obtain a case quality score matrix; and selecting a candidate teaching case set based on the case quality score matrix through a first multi-objective optimization algorithm.

[0104] The semantic network subgraph refers to a subgraph structure extracted from the field adaptive knowledge graph, containing core knowledge point nodes, associated knowledge point nodes and cognitive path sequences, which can be specifically implemented by using a graph database query technology, and is used to ensure the knowledge coverage completeness of the case content. The neural network generation model refers to a text generation model based on a deep learning architecture, which can be specifically implemented by using a Transformer model, and is used to generate diversified initial case texts according to the knowledge point association relationship. The target generation strategy template refers to a pre-defined strategy rule containing a content structure template and an interaction design template, which can be specifically implemented by using an XML structured template, and is used to standardize the interaction logic and content organization mode of the case. The rule engine refers to a logic processing module for performing teaching adaptability verification, which can be specifically implemented by using a Drools rule engine, and is used to convert teaching constraints into executable verification rules. The case quality scoring matrix refers to multi-dimensional scoring data containing content accuracy, teaching effectiveness and technical feasibility, which can be specifically generated by using an analytic hierarchy process, and is used to quantitatively evaluate the comprehensive quality of the case.

[0105] Specifically, the core knowledge point nodes and the associated knowledge point nodes are input to the neural network generation model through the topological relationship of the semantic network subgraph to generate a basic case text set covering the knowledge point association. The content structure template is used to divide the unstructured initial case text into standardized content partitions, and the interaction design template is used to insert interaction nodes in the case framework. The rule engine performs difficulty consistency checking on the case framework based on a difficulty level threshold, and verifies the feasibility of the evaluation indicators according to the evaluation standards. The case quality scoring matrix calculates the scoring indicators of different dimensions by weighting, and the first multi-objective optimization algorithm selects a candidate case set with balanced quality scores based on the Pareto frontier theory.

[0106] Compared with related technologies, the existing case generation method usually directly outputs an initial case without structured processing, lacks dynamic verification of interaction design and teaching constraints, and leads to unstable case quality. The present scheme generates a standardized framework through strategy template reorganization, and combines the adaptability verification of the rule engine to form a dynamic matching mechanism between the case generation process and the teaching demand, and balances the evaluation results of different quality dimensions by using a multi-objective optimization algorithm, thereby improving the overall adaptability of the candidate case set.

[0107] Through the above technical solutions, the present application realizes the collaborative control of knowledge point coverage completeness and teaching strategy specification in the teaching case generation process, ensures that the candidate cases meet the preset difficulty level and evaluation standards through the structured reorganization and dynamic verification mechanism, and selects a candidate case set with both teaching effectiveness and technical feasibility based on multi-dimensional quality evaluation, thereby solving the technical problem that the case generation and teaching scene demand are disconnected in the traditional method.

[0108] The present application further proposes the following technical solutions, please refer toFigure 4 Taking the server as an example, the execution subject is a server, and the method comprises the following steps.

[0109] 401, obtaining an environment parameter configuration of a target learning environment, the environment parameter configuration comprising hardware device parameters, network environment parameters and learning terminal types;

[0110] 402, based on the environment parameter configuration, performing teaching environment adaptability analysis on each candidate case in the candidate teaching case set to generate an environment adaptability score;

[0111] 403, performing multi-dimensional teaching effect prediction evaluation on the candidate teaching case set to obtain a teaching effect prediction score;

[0112] 404, based on the environment adaptability score and the teaching effect prediction score, determining a comprehensive evaluation score of each candidate case by using a second multi-objective decision algorithm;

[0113] 405, sorting the candidate teaching case set based on the comprehensive evaluation score to generate a candidate case sorting list, and determining an initial teaching case from the sorted candidate cases based on a preset selection strategy.

[0114] The environment adaptability score is used to represent the running compatibility and interaction fluency of the candidate case in the target learning environment, the second multi-objective decision algorithm dynamically adjusts the weight distribution of different score dimensions according to the teaching target description information, the teaching effect prediction evaluation includes knowledge mastery prediction, ability improvement prediction and participation prediction, the environment parameter configuration refers to the hardware device parameters, network environment parameters and learning terminal types of the target learning environment, which can be realized by collecting data such as device model, network bandwidth and terminal operating system version through a system configuration interface, to ensure that the case selection process meets the actual physical condition constraints. The teaching environment adaptability analysis refers to running the candidate case in a virtual test environment to simulate a real teaching scene, which is realized by creating a simulation environment containing the same hardware configuration and network topology using digital twin technology, to predict the running compatibility of the case in the real environment. The multi-dimensional teaching effect prediction evaluation refers to quantitatively predicting the knowledge mastery, ability improvement and participation, which is realized by using a machine learning model based on historical teaching data for prediction analysis, to break through the subjective limitations of traditional artificial experience judgment. The second multi-objective decision algorithm refers to an optimization algorithm that dynamically adjusts the weights of different score dimensions, which is realized by using a fuzzy logic weight distribution mechanism based on teaching target description, to balance the priority of teaching needs and environmental constraints.

[0115] Specifically, the device parameters and network parameters of the target environment are acquired by the environmental parameter configuration acquisition module to establish a digital twin teaching scene containing a terminal simulator. When the candidate cases run in the virtual environment, real-time running data such as rendering frame rate and response delay are collected, and time-domain features and frequency-domain features are generated through multi-scale feature extraction. At the same time, the teaching effect prediction model generates knowledge mastery prediction values and participation prediction values based on case content features and cognitive path data. The second multi-objective decision algorithm dynamically adjusts the fusion coefficient of the environment adaptability score and the teaching effect score according to the weight proportion of knowledge transmission and ability cultivation in the teaching target, and generates a comprehensive evaluation score ranking list. The selection strategy module selects the initial teaching case according to the preset rules, such as preferentially selecting the case with the lowest resource occupation in the top three comprehensive scores.

[0116] Compared with related technologies, the traditional case selection method only considers the matching degree of teaching content and does not establish a virtual test environment to verify technical feasibility, which may easily lead to running lag or device incompatibility problems after case deployment. Related technologies use fixed weights for case scoring, which cannot dynamically adjust the evaluation standard according to specific teaching targets, such as not being able to increase the scoring weight of interaction fluency in courses that emphasize interactive ability. The present scheme pre-generates environment adaptability data through digital twin technology, combined with a multi-objective dynamic decision mechanism, to a certain extent, solves the problem that technical feasibility and teaching effect are difficult to balance in the case selection process.

[0117] Through the above technical solutions, the present application realizes intelligent optimization and dynamic adaptation of teaching cases, which is specifically embodied in: environment adaptability analysis based on a virtual test environment can identify candidate cases with compatibility risks in advance, avoiding running failures after deployment; machine learning models are used to predict the teaching effects of different cases, providing objective and quantitative basis for the selection process; the dynamic weight allocation mechanism can automatically optimize the scoring strategy according to the priority of the teaching target, ensuring that the selected case meets the environmental running conditions and achieves the preset teaching purpose.

[0118] The application further proposes to configure a virtual test environment based on environmental parameters, and deploy a digital twin teaching scenario in the virtual test environment, which includes hardware configuration, network topology and terminal device simulator consistent with the target learning environment; run each candidate teaching case in the digital twin teaching scenario, and collect case running data in real time, including rendering frame rate timing data, interaction response delay sequence and resource occupation fluctuation curve; multi-scale feature extraction is performed on the case running data of each candidate teaching case to obtain the time domain feature set, frequency domain feature set and nonlinear dynamics feature set of each candidate teaching case, and the nonlinear dynamics feature set includes Lyapunov index, fractal dimension and entropy index; based on the time domain feature set, frequency domain feature set and nonlinear dynamics feature set of each candidate teaching case, the environmental adaptability score of each candidate teaching case is generated.

[0119] Among them, the digital twin teaching scenario refers to a simulation environment consistent with the physical properties of the target learning environment constructed by virtualization technology, which can be implemented by containerization deployment and hardware abstraction layer technology, and is used to accurately simulate the running conditions of the real teaching scenario. The case running data refers to the performance index set generated when the candidate case is executed in the simulation environment, which can be collected by embedded probes and system monitoring tools, and is used to reflect the running state of the case in a specific environment. Multi-scale feature extraction refers to signal analysis of running data from different dimensions, which can be implemented by wavelet transform and phase space reconstruction algorithm, and is used to fully capture the behavior characteristics of the case in different time scales and system dynamics. Lyapunov index is used to quantify the sensitivity of the system to initial conditions, fractal dimension is used to describe the complexity of data fluctuation pattern, and entropy index is used to measure the degree of system disorder, and the combination of the three can evaluate the robustness of the case in a dynamic environment.

[0120] Specifically, the virtual test environment constructs a high-fidelity digital twin scenario by mirroring the hardware configuration and network topology parameters of the target learning environment. When the candidate case runs in this scenario, the system monitoring module captures real-time indicators such as rendering frame rate, interaction delay and resource occupancy rate. The time domain feature set analyzes the stability of the interaction response by calculating the mean and variance of the delay sequence; the frequency domain feature set extracts the periodicity of resource occupation by Fourier transform; the nonlinear dynamics feature set judges the degree of system chaos by calculating the Lyapunov index, and evaluates the complexity of data fluctuation by combining the fractal dimension. The three types of features are fused by weighting to generate an environmental adaptability score, which quantifies the compatibility of the candidate case with the target environment.

[0121] Compared with the related art, the traditional method relies on manual testing or a simplified simulation environment, and cannot accurately reflect the heterogeneous hardware configuration and dynamic network conditions of the real teaching scene. The related art usually only monitors a single indicator such as average response time, and lacks the ability to analyze the nonlinear behavior of the system. The scheme realizes the full-factor modeling of the environmental parameters through the digital twin technology, and can more accurately predict the performance of the case in the actual deployment by combining the multi-scale feature extraction method.

[0122] Through the above technical scheme, the application solves the problem of insufficient compatibility verification of candidate cases running in a real teaching environment, and realizes quantitative evaluation of interaction fluency and resource utilization efficiency. By capturing the nonlinear characteristics of rendering frame rate fluctuation and interaction delay, the performance bottleneck that may occur in a high concurrency or network fluctuation scenario is effectively identified. This enables educational institutions to predict the running stability of the case before deployment, avoiding teaching interruptions caused by environmental adaptability problems.

[0123] The application further proposes a scheme for generating an environmental adaptability score based on a time-domain feature set, a frequency-domain feature set, and a nonlinear dynamics feature set of each candidate teaching case, including: determining a response stability indicator based on the interaction response delay sequence in the time-domain feature set; generating a resource use efficiency score based on the energy spectrum distribution characteristics in the frequency-domain feature set; extracting the fractal dimension and entropy value indicators in the nonlinear dynamics feature set to construct a system complexity descriptor; establishing an environmental adaptability evaluation benchmark model based on hardware performance benchmarks and network quality requirements; and generating an environmental adaptability score through a weighted fusion algorithm.

[0124] The time-domain feature set refers to the interaction response delay sequence generated when the candidate case runs, and the delay fluctuation variance and range can be calculated using a sliding window statistical method to analyze the response stability under different load conditions. The frequency-domain feature set refers to the energy spectrum distribution characteristics obtained by Fourier transform of the case running data, and the periodic characteristics of the resource occupation mode can be calculated by the main frequency band energy proportion to evaluate the resource use efficiency. The nonlinear dynamics feature set includes fractal dimension and sample entropy indicators, which can be calculated using the box counting method and approximate entropy algorithm to quantify the adaptive ability of the case in a dynamic environment. The environmental adaptability evaluation benchmark model refers to a scoring benchmark established according to the hardware configuration and network bandwidth of the target learning environment, which can be constructed by setting resource occupation thresholds and response time constraints. The weighted fusion algorithm refers to a calculation method for dynamically adjusting the weights of each feature according to the teaching target, which can determine the weight coefficients of different evaluation dimensions using the analytic hierarchy process.

[0125] Specifically, when running the candidate case in a virtual test environment, a time-domain feature set is constructed by collecting an interaction response delay sequence, and the fluctuation amplitude of the response delay is calculated as a stability indicator. The case running data is subjected to spectral analysis, and the main frequency band energy proportion feature is extracted to evaluate its utilization efficiency of hardware resources. Meanwhile, the fractal dimension and sample entropy value are calculated to construct a descriptor reflecting the complexity of the system. The above indicators are input into a pre-established evaluation benchmark model, and according to the hardware performance benchmark and network quality requirements of the target environment, the multi-dimensional features are converted into a unified environment adaptability score through a weighted fusion algorithm. For example, in a low-configuration terminal environment, the weight coefficient of the resource utilization efficiency score can be set to 0.6, the response stability weight is set to 0.3, and the system complexity weight is set to 0.1, so as to preferentially select the candidate case with low resource occupation.

[0126] Compared with the related art, the existing scheme usually only tests the running performance of the case in an ideal environment, while the present scheme can accurately predict the performance of the case under dynamic load conditions by constructing a virtual test environment to collect real running data and combining time-domain, frequency-domain and nonlinear dynamics multi-dimensional feature analysis. The traditional method relies on single resource peak monitoring, while the present scheme can discover potential performance bottlenecks by identifying the periodic pattern of resource occupation through energy spectrum distribution characteristics. In addition, the introduction of fractal dimension and entropy value indicators can effectively evaluate the adaptive ability of the case in a complex network environment, breaking through the limitations of linear evaluation models.

[0127] Through the above technical scheme, the present application can accurately quantify the running compatibility of the candidate teaching case in different teaching environments, and effectively identify the response stability, resource utilization efficiency and system robustness of the case through multi-dimensional feature fusion evaluation. The present scheme solves the problem that the traditional method cannot dynamically evaluate the environmental adaptability of the case, ensuring that the selected teaching case can run stably in the target learning environment, and at the same time has the adaptive ability to cope with network fluctuations and device performance differences.

[0128] The present application further proposes the following technical scheme, see Figure 5 Taking a server as an example, the execution subject, the following steps are included.

[0129] 501, spatiotemporal alignment and feature fusion processing are performed on the learning interaction data stream to obtain a multi-dimensional feature tensor, the multi-dimensional feature tensor including time-series behavior features, spatial attention features and cognitive state features;

[0130] 502, the multi-dimensional feature tensor is processed by a causal reasoning algorithm to obtain a case effect evaluation report and an optimization suggestion set, the case effect evaluation report being a structured evaluation result;

[0131] 503、based on the case effect evaluation report and the optimization suggestion set, a case optimization instruction sequence is generated, the case optimization instruction sequence includes content adjustment instructions, interaction improvement instructions and difficulty adjustment instructions;

[0132] 504、the case optimization instruction sequence is executed, and the initial teaching case is optimized to obtain a target teaching case.

[0133] Among them, the space-time alignment and feature fusion processing refers to synchronously integrating learning interaction data of different timestamps and spatial positions, which can be specifically implemented by using a time series interpolation algorithm combined with a spatial attention mechanism to solve the space-time dislocation problem of multi-source heterogeneous data. The causal reasoning algorithm refers to identifying the influencing factors of teaching effectiveness by constructing a causal relationship network, which can be specifically implemented by using a causal structure learning model combined with a counterfactual reasoning method to break through the limitations of traditional correlation analysis. The case optimization instruction sequence refers to a set of modification instructions for case content, interaction mode and difficulty level, which can be specifically implemented by using a dynamic programming algorithm to sequence the optimization operation template to ensure the logical coherence of the optimization process.

[0134] Specifically, after the learning interaction data stream is aligned in space and time, the time sequence features of the behavior log and the spatial distribution features of the cognitive state are fused into a multi-dimensional feature tensor. The causal reasoning algorithm constructs a teaching effectiveness causal network graph on the tensor, quantifies the causal contribution of each feature to the teaching goal through counterfactual reasoning, and identifies key positive factors and bottleneck constraints. Based on the optimization instruction sequence generated, the dynamic programming algorithm balances the modification range and the effect improvement amplitude, and preferentially executes adjustment operations that have a significant impact on teaching effectiveness. The optimized case retains the original teaching logic, corrects content bias, optimizes interaction nodes, and adjusts difficulty gradient, forming a dynamically matched teaching case for the teaching goal.

[0135] Compared with related technologies, the traditional method only relies on preset rules for case generation, lacking a feedback optimization mechanism based on real teaching data. The data analysis in related technologies mostly uses statistical correlation methods, which cannot distinguish between causal relationships and false associations, resulting in inaccurate optimization direction. The present scheme accurately identifies the key influencing factors of teaching effectiveness through the causal reasoning algorithm, and generates an optimization instruction sequence through dynamic programming, realizing data-driven closed-loop iterative optimization.

[0136] Through the above technical solutions, the present application solves the problem of disconnection between teaching case generation and teaching effect feedback, and realizes dynamic closed-loop optimization of the case. By positioning the bottleneck factors affecting the teaching effect through causal reasoning, invalid adjustment operations are avoided. Through multi-dimensional feature fusion and instruction sequence generation, the case optimization process ensures content accuracy, interaction fluency and difficulty adaptability, improving the consistency of the case and the teaching goal.

[0137] The application further proposes processing the multi-dimensional feature tensor through a causal reasoning algorithm to obtain a case effect evaluation report and an optimization suggestion set. Specifically, the multi-dimensional feature tensor is subjected to causal structure learning to obtain a teaching effect causal network graph; the causal contribution degree of each feature node is determined based on the network graph through a counterfactual reasoning algorithm; the key feature factors and bottleneck feature factors are identified according to the causal contribution degree; and the evaluation report and the optimization suggestion set are generated in combination with the teaching target.

[0138] Among them, the multi-dimensional feature tensor refers to a data structure processed through spatio-temporal alignment and feature fusion, containing time series behavior features, spatial attention features and cognitive state features, which can be specifically implemented by using a tensor decomposition algorithm combined with a feature splicing technology, and is used for unified representation learning of multi-modal information in interactive data. Causal structure learning refers to a method for constructing causal relationships between variables through a Bayesian network or a structural equation model, which can be specifically implemented by using a PC algorithm or an FCI algorithm, and is used for revealing the causal association between teaching effect and each feature. The counterfactual reasoning algorithm refers to a method for quantifying causal effects by constructing counterfactual scenarios, which can be specifically implemented by using a latent outcome framework or a double machine learning model, and is used for excluding confounding factors and accurately evaluating the influence of features. The causal contribution degree refers to a quantitative value obtained by calculating the change amplitude of the teaching effect index when the feature variable is intervened, which can be specifically implemented by using an average causal effect or a conditional average causal effect index, and is used for objectively measuring the real action strength of each feature on the teaching effect.

[0139] Specifically, the teaching case optimization process first converts the learning interactive data into a multi-dimensional feature tensor containing spatio-temporal correlation, constructs a teaching effect causal network graph through causal structure learning, and explicitly expresses the causal relationships between behavior features, cognitive features and environmental features. Based on the network graph, counterfactual reasoning is performed, for example, assuming that the teaching effect changes in amplitude when an environmental feature is changed, so as to calculate the causal contribution degree of each feature. By setting a contribution degree threshold, key feature factors and bottleneck feature factors are identified, for example, a positive impact feature with a contribution degree higher than a preset threshold is marked as a key factor, and a negative impact feature with a contribution degree lower than a preset threshold is marked as a bottleneck factor. In combination with the priority weight in the teaching target description information, an evaluation report containing specific improvement directions is generated, for example, when the teaching target focuses on knowledge mastery, the bottleneck factors associated with cognitive features are preferentially optimized.

[0140] In some embodiments, the causal structure learning can employ a constraint-based PC algorithm combined with domain knowledge constraints, such as pre-defining possible causal relationship edges between teaching environment features and cognitive features. The counterfactual reasoning can employ a two-stage machine learning framework, the first stage estimates the relationship between feature variables and confounding factors through a random forest regression model, and the second stage calculates the causal effect through a linear regression model. The causal contribution calculation can introduce standardization processing, such as normalizing each feature contribution to a percentage form, facilitating cross-case comparative analysis.

[0141] Compared with related technologies, the traditional method only identifies influencing factors through statistical correlation analysis, and cannot distinguish between real causal relationships and false associations. For example, the related technology may mistakenly determine that high participation behavior and teaching effect improvement are a causal relationship, while in fact both may be affected by cognitive state improvement. The present scheme removes the interference of confounding factors through causal reasoning, such as finding the real causal effect of interaction response delay on participation, thereby accurately locating the technical bottleneck that needs to be optimized.

[0142] Through the above technical solutions, the present application can identify the essential factors affecting the case effect based on real teaching data, avoiding the optimization decision being misled by false associations. Through causal contribution quantitative analysis, the optimization priority is clear, and the degree and efficiency of case improvement are improved. Combined with the dynamic adjustment of the optimization direction of the teaching target, the dual optimization mechanism of data-driven and teaching logic is realized, forming a closed-loop feedback system for continuous iteration of teaching cases.

[0143] The present application further proposes a method of generating case optimization instruction sequences based on case effect evaluation reports and optimization suggestion sets, including structured analysis of case effect evaluation reports to extract teaching effect bottleneck features and optimization priority indicators, matching optimization operation template sets based on bottleneck features and sorting and filtering according to priority indicators, initializing reference operation template parameters based on teaching targets, generating instruction execution sequences through dynamic programming algorithm optimization, and performing dual verification and fine-tuning of teaching logic and technical implementation on the instruction sequence.

[0144] The optimization priority index refers to an optimization order quantitative parameter determined according to the causal contribution degree and the teaching effect influence degree, and can be specifically implemented by using a weighted scoring algorithm, and is used to guide the execution order of the optimization operation. The case optimization rule library refers to a database storing optimization operation templates corresponding to different teaching bottleneck types, and can be specifically constructed by using a graph database, and is used to quickly match an optimization strategy consistent with the current bottleneck characteristics. The dynamic programming algorithm refers to a sequence decision method taking maximizing the bottleneck improvement effect and minimizing the modification range as optimization objectives, and can be specifically implemented by using Bellman equation, and is used to generate an optimal instruction execution order. The double verification mechanism refers to a verification process for checking the teaching logic correctness and the technical implementation feasibility of the optimization instruction, and can be specifically implemented by using a combination of formal verification and sandbox testing, and ensures the implementability of the optimization scheme.

[0145] Specifically, when the system receives a case effect evaluation report, first, structured data is extracted by using natural language processing technology, and main obstacle factors of teaching effect improvement and their priorities are identified. According to the identified bottleneck types, a set of applicable operation templates is retrieved from the pre-stored optimization rule library. Then, the templates are sorted according to the priority index, and variables in the templates are initialized and assigned values in combination with the teaching target parameters. The expected effect and modification cost of different operation combinations are calculated by using the dynamic programming algorithm, and an optimal instruction sequence is generated. Finally, the optimization instruction is simulated and executed in a virtual teaching environment, and the rationality of the teaching logic and the feasibility of the technical implementation are verified, and the instruction parameters are adjusted according to the verification results.

[0146] Compared with related technologies, the traditional method usually generates optimization suggestions by using fixed rules, lacks in-depth analysis of the causal relationship of the teaching effect, and causes deviation of the optimization direction from the actual problem. The related technical solution often ignores the influence of the execution order on the overall effect when generating the optimization instruction, and lacks constraint control on the modification range. In addition, the existing system generally lacks a verification link for the implementability of the optimization scheme, and is prone to produce optimization schemes that are effective in theory but cannot be deployed in practice.

[0147] Through the above technical solutions, the generation and reliable execution of the teaching case optimization instruction are realized. The optimization direction is accurately aligned with the teaching target by using structured analysis and causal analysis, the optimization effect and the modification cost are balanced by using the dynamic programming algorithm, and the double verification mechanism effectively prevents the generation of invalid optimization schemes. This enables the teaching case to be continuously iterated according to the real teaching feedback, and gradually improves the teaching effectiveness while keeping the core content of the case stable.

[0148] All the optional technical solutions described above can be combined in any way to form optional embodiments of the present application, and will not be repeated here.

[0149] Figure 6is a structural schematic diagram of a case generation management system based on artificial intelligence education provided by an embodiment of the present application, referring to Figure 6 , the system comprises:

[0150] The determination module 601 is configured to determine target knowledge point description information, teaching target description information and a constraint condition set corresponding to the target teaching content in response to a teaching case generation request of the target teaching content, and the constraint condition set is used to constrain the difficulty level, interactive form and evaluation standard of the generated teaching case.

[0151] The case set generation module 602 is configured to generate a candidate teaching case set of the target teaching content based on the target knowledge point description information, the teaching target description information and the constraint condition set.

[0152] The deployment module 603 is configured to determine an initial teaching case from the candidate teaching case set and deploy the initial teaching case to the target learning environment.

[0153] The optimization module 604 is configured to optimize the initial teaching case based on a learning interaction data stream obtained from the target learning environment to obtain a target teaching case of the target teaching content, and the learning interaction data stream comprises a behavior log sequence, cognitive state change data and environmental context information.

[0154] It should be noted that the case generation management system based on artificial intelligence education provided by the above embodiment is only exemplified by the division of the above functional modules when performing case generation management. In actual application, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the computer device is divided into different functional modules to complete all or part of the functions described above. In addition, the case generation management system based on artificial intelligence education provided by the above embodiment and the case generation management method based on artificial intelligence education embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0155] Figure 7 is a structural schematic diagram of a server provided by an embodiment of the present application. The server 700 can have great differences due to different configurations or performances, and can include one or more processors (Central Processing Units, CPUs) 701 and one or more memories 702. The one or more memories 702 store at least one computer program, which is loaded and executed by the one or more processors 701 to implement the methods provided by the above-mentioned various method embodiments. Of course, the server 700 can also have a wired or wireless network interface, a keyboard, an input and output interface, and other components for realizing the functions of the device, so as to perform input and output. The server 700 can also include other components for realizing the functions of the device, which will not be repeated here.

[0156] In an example embodiment, a computer readable storage medium, such as a memory including a computer program executable by a processor to perform the above-mentioned artificial intelligence education-based case generation management method is also provided. For example, the computer readable storage medium can be a Read-Only Memory (ROM), a Random Access Memory (RAM), a Compact Disc Read-Only Memory (CD-ROM), a magnetic tape, a floppy disk, and an optical data storage device, etc.

[0157] In an example embodiment, a computer program product or computer program is also provided, which includes program code stored in a computer readable storage medium, and a processor of a computer device reads the program code from the computer readable storage medium, and the processor executes the program code to cause the computer device to perform the above-mentioned artificial intelligence education-based case generation management method.

[0158] In some embodiments, the computer program related to the embodiments of the present application can be deployed to execute on one computer device, or on multiple computer devices located in one place, or on multiple computer devices distributed in multiple places and interconnected through a communication network, which can constitute a blockchain system.

[0159] Those of ordinary skill in the art can understand that all or part of the steps of the above-mentioned embodiments can be completed by hardware, or by a program instructing relevant hardware, and the program can be stored in a computer readable storage medium, such as a Read-Only Memory, a magnetic disk or an optical disk, etc.

[0160] The above is only an optional embodiment of the present application, and is not used to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A case generation and management method based on artificial intelligence education, characterized in that, The method includes: In response to a request to generate teaching cases for the target teaching content, the system determines the target knowledge point description information, teaching objective description information, and constraint set corresponding to the target teaching content. The constraint set is used to constrain the difficulty level, interaction format, and evaluation criteria of the generated teaching cases. Based on the target knowledge point description information, a pre-constructed domain-adaptive knowledge graph is queried, and a semantic network subgraph corresponding to the target knowledge point description information is extracted from the domain-adaptive knowledge graph. The semantic network subgraph includes core knowledge point nodes, related knowledge point nodes, cognitive path sequences, and a set of common misconceptions. The domain-adaptive knowledge graph is an educational domain knowledge representation structure composed of a knowledge point ontology layer, a teaching relationship layer, and a cognitive model layer. Based on the teaching objective description information and the cognitive path sequences in the semantic network subgraph, a target generation strategy template is obtained from the generation strategy rule base. Based on the semantic network subgraph, the target generation strategy template, and the constraint set, a set of candidate teaching cases is generated. Initial teaching cases are determined from the set of candidate teaching cases, and the initial teaching cases are deployed to the target learning environment; The initial teaching case is optimized based on the learning interaction data stream obtained from the target learning environment to obtain the target teaching case of the target teaching content. The learning interaction data stream includes behavior log sequences, cognitive state change data, and environmental context information.

2. The method according to claim 1, characterized in that, The step of obtaining a target generation strategy template from the generation strategy rule base based on the teaching objective description information and the cognitive path sequence in the semantic network subgraph includes: The teaching strategy intent is identified by analyzing the teaching objective description information to obtain a teaching strategy vector representation, which is used to indicate the relative weight ratio of knowledge transmission, ability cultivation and quality improvement. Based on the teaching strategy vector representation and the cognitive path sequence in the semantic network subgraph, the matching score of each strategy template in the generated strategy rule base is determined. The matching score is a quantitative indicator of the fit of the teaching strategy and the consistency of the cognitive path. The strategy templates in the generated strategy rule base are sorted according to the matching score, and the top N strategy templates with the highest matching score are selected as the candidate strategy template set, where N is a positive integer; A teaching adaptability analysis is performed on each strategy template in the candidate strategy template set to generate an adaptability assessment result. The adaptability assessment result is used to indicate the compatibility of the strategy template with the difficulty level and interaction form in the constraint set. Based on the matching score and the adaptability evaluation results, the target generation strategy template is determined from the candidate strategy template set.

3. The method according to claim 1, characterized in that, The generation of a candidate teaching case set based on the semantic network subgraph, the target generation strategy template, and the constraint set includes: Based on the core knowledge point nodes and related knowledge point nodes in the semantic network subgraph, a basic case content set is generated through a neural network generation model. The basic case content set includes multiple initial case texts that meet the knowledge point coverage requirements. Based on the content structure template and interaction design template in the target generation strategy template, the basic case content set is restructured to generate a standardized case framework set. Each case framework in the standardized case framework set includes content partitioning and interaction node design. Based on the difficulty level and evaluation criteria in the set of constraints, the standardized case framework set is tested for teaching adaptability through a rule engine to obtain a tested case framework set. The teaching adaptability test includes difficulty consistency check and evaluation feasibility analysis. A multi-dimensional quality assessment was performed on the verified case framework set to obtain a case quality scoring matrix, which includes content accuracy score, teaching effectiveness score, and technical feasibility score. Based on the case quality scoring matrix, a set of candidate teaching cases that meet the preset quality conditions is selected from the verified case framework set using a first multi-objective optimization algorithm.

4. The method according to claim 1, characterized in that, The step of determining initial teaching cases from the candidate teaching case set includes: Obtain the environment parameter configuration of the target learning environment, which includes hardware device parameters, network environment parameters, and learning terminal type; Based on the environmental parameter configuration, a teaching environment adaptability analysis is performed on each candidate case in the candidate teaching case set to generate an environment adaptability score. The environment adaptability score is used to represent the running compatibility and interactive fluency of the candidate case in the target learning environment. The candidate teaching case set is subjected to a multi-dimensional teaching effectiveness prediction and evaluation to obtain a teaching effectiveness prediction score. The teaching effectiveness prediction and evaluation includes knowledge mastery prediction, ability improvement prediction, and participation prediction. Based on the environmental adaptability score and the teaching effect prediction score, the comprehensive evaluation score of each candidate case is determined by the second multi-objective decision algorithm. The second multi-objective decision algorithm dynamically adjusts the weight allocation of different scoring dimensions according to the teaching objective description information. The candidate teaching case set is sorted based on the comprehensive evaluation score to generate a candidate case sorting list, and the initial teaching case is determined from the sorted candidate cases based on a preset selection strategy.

5. The method according to claim 4, characterized in that, Based on the configured environmental parameters, the process of performing a teaching environment suitability analysis on each candidate case in the candidate teaching case set and generating an environment suitability score includes: A virtual testing environment is constructed based on the environmental parameter configuration, and a digital twin teaching scenario is deployed in the virtual testing environment. The digital twin teaching scenario includes hardware configuration, network topology, and terminal device simulator consistent with the target learning environment. Each candidate teaching case is run in the digital twin teaching scenario, and case running data is collected in real time. The case running data includes rendering frame rate timing data, interaction response delay sequence, and resource usage fluctuation curve. Multi-scale feature extraction is performed on the case operation data of each candidate teaching case to obtain the time-domain feature set, frequency-domain feature set, and nonlinear dynamics feature set of each candidate teaching case. The nonlinear dynamics feature set includes Lyapunov exponent, fractal dimension, and entropy index. Based on the time-domain feature set, frequency-domain feature set, and nonlinear dynamics feature set of each candidate teaching case, an environmental adaptability score is generated for each candidate teaching case.

6. The method according to claim 5, characterized in that, The environmental adaptability score for each candidate teaching case is generated based on its time-domain feature set, frequency-domain feature set, and nonlinear dynamics feature set, including: Based on the interaction response delay sequence in the time-domain feature set of each candidate teaching case, the response stability index of each candidate teaching case under different load conditions is determined. The response stability index is used to characterize the predictability and consistency of the case interaction response. Based on the energy spectrum distribution characteristics in the frequency domain feature set of each candidate teaching case, the resource occupancy mode of each candidate teaching case during operation is determined, and a resource utilization efficiency score is generated. The resource utilization efficiency score reflects the degree of optimization of hardware resources by the case. The fractal dimension and entropy indices are extracted from the nonlinear dynamic feature set to construct a system complexity descriptor. The system complexity descriptor is used to quantify the adaptability and robustness of each candidate teaching case in a dynamic environment. Based on the hardware performance benchmark and network quality requirements in the environmental parameter configuration, an environmental adaptability evaluation benchmark model is established. The response stability index, resource utilization efficiency score, and system complexity descriptor are input into the environmental adaptability assessment benchmark model, and an environmental adaptability score for each candidate teaching case is generated through a weighted fusion algorithm.

7. The method according to claim 1, characterized in that, The optimization of the initial teaching case based on the learning interaction data stream obtained from the target learning environment to obtain the target teaching case for the target teaching content includes: The learning interaction data stream is subjected to spatiotemporal alignment and feature fusion processing to obtain a multidimensional feature tensor, which includes temporal behavioral features, spatial attention features and cognitive state features. The multidimensional feature tensor is processed by a causal reasoning algorithm to obtain a case effect evaluation report and a set of optimization suggestions. The case effect evaluation report is a structured evaluation result. Based on the case effect evaluation report and the set of optimization suggestions, a case optimization instruction sequence is generated, which includes content adjustment instructions, interaction improvement instructions, and difficulty adjustment instructions. The initial teaching case is optimized by executing the case optimization instruction sequence to obtain the target teaching case.

8. The method according to claim 7, characterized in that, The process of processing the multidimensional feature tensor using a causal reasoning algorithm yields a case performance evaluation report and a set of optimization suggestions, including: Causal structure learning is performed on the multidimensional feature tensor to obtain a causal network graph of teaching effect, which includes behavioral feature nodes, cognitive feature nodes, environmental feature nodes, and causal relationship edges between nodes. Based on the aforementioned causal network diagram of teaching effectiveness, the causal contribution of each feature node to the teaching effectiveness index is determined by the counterfactual reasoning algorithm. The causal contribution is used to quantify the influence intensity of each feature factor on the teaching effectiveness. Based on the causal contribution, key characteristic factors and bottleneck characteristic factors affecting teaching effectiveness are identified. The key characteristic factors are those that have a significant positive impact on teaching effectiveness, and the bottleneck characteristic factors are those that restrict teaching effectiveness. Based on the analysis results of the key and bottleneck characteristic factors, and combined with the teaching objective description information, a case effect evaluation report and a set of optimization suggestions are generated, which include diagnostic conclusions on teaching effectiveness and targeted improvement suggestions.

9. A case generation and management system based on artificial intelligence education, characterized in that, include: The determination module is used to respond to a request to generate teaching cases for the target teaching content, and determine the target knowledge point description information, teaching objective description information and constraint set corresponding to the target teaching content. The constraint set is used to constrain the difficulty level, interaction form and evaluation criteria of the generated teaching cases. The case set generation module is used to query a pre-built domain-adaptive knowledge graph based on the target knowledge point description information, and extract a semantic network subgraph corresponding to the target knowledge point description information from the domain-adaptive knowledge graph. The semantic network subgraph includes core knowledge point nodes, related knowledge point nodes, cognitive path sequences, and a set of common misconceptions. The domain-adaptive knowledge graph is an educational domain knowledge representation structure composed of a knowledge point ontology layer, a teaching relationship layer, and a cognitive model layer. Based on the teaching objective description information and the cognitive path sequences in the semantic network subgraph, a target generation strategy template is obtained from the generation strategy rule base. Based on the semantic network subgraph, the target generation strategy template, and the set of constraints, a set of candidate teaching cases is generated. The deployment module is used to determine an initial teaching case from the candidate teaching case set and deploy the initial teaching case to the target learning environment; An optimization module is used to optimize the initial teaching case based on the learning interaction data stream obtained from the target learning environment to obtain a target teaching case for the target teaching content. The learning interaction data stream includes behavior log sequences, cognitive state change data, and environmental context information.

Images