Dynamic vr teaching scene construction system and method of adaptive content generation

Abstract

The application discloses a dynamic VR teaching scene construction system and method of adaptive content generation, and belongs to the technical field of virtual reality, which comprises the following steps: extracting scene characteristic variables affecting teaching targets, combining a multi-dimensional characteristic extraction model to form a primitive mode; expanding the dimension of the primitive mode based on a deep learning model and correlating a regulation protocol to generate a strategy scheme; selecting a strategy scheme type to generate a target scheme according to teaching task requirements; injecting learner coordinates into the target scheme to generate a scene parameter set and integrate a dynamic narrative script in combination with a learner portrait; selecting content, interaction and feedback three types of teaching elements from a resource library to generate a component scheme according to the scene parameter set, the dynamic narrative script and the regulation protocol; compiling the component scheme through a scene compiler, integrating and optimizing the three types of elements to generate a dynamic teaching unit, and deploying the dynamic teaching unit in a VR engine to respond to learner behavior, so that adaptive dynamic VR teaching scene construction is realized.

Description

Technical Field

[0001] This invention belongs to the field of virtual reality technology, specifically a dynamic VR teaching scene construction system and method with adaptive content generation. Background Technology

[0002] With the rapid development of virtual reality technology, VR teaching has gradually become an emerging trend in the education field. Teaching scenarios built using VR technology can provide learners with an immersive learning experience, greatly enhancing the fun and effectiveness of learning. However, in practical applications, VR teaching scenarios face numerous technical challenges. On the one hand, in order to simulate realistic physical environments and interactive effects in VR teaching scenarios, complex physics engines are usually required. However, when these complex physics engines run in real-time in VR, they consume a large amount of computing resources, causing a sharp drop in device frame rate and dynamic latency, severely impacting learners' immersion and interactive experience. On the other hand, existing VR teaching systems have deficiencies in behavior recognition, failing to accurately identify learners' complex behaviors and actions, resulting in insufficient precision in teaching interactions. Multi-user synchronization barriers prevent multiple learners from achieving real-time, smooth interaction and collaboration in the same virtual teaching scenario. Furthermore, the lack of a single environmental response means that the teaching scenario and content cannot be dynamically adjusted according to learners' behavior and needs, limiting the flexibility and adaptability of VR teaching. Currently, although there is some research on VR technology performance optimization and interaction improvement, there is still no comprehensive and effective method that can simultaneously solve the problems of complex physics engine operation performance, behavior recognition, multi-user synchronization and environmental response, etc., to meet the needs of building high-quality VR teaching scenarios. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention proposes a dynamic VR teaching scene construction system and method with adaptive content generation. It extracts scene feature variables that influence teaching objectives and combines them with a multi-dimensional feature extraction model to form primitive patterns. Based on a deep learning model, it expands the dimensions of the primitive patterns and associates them with control protocols to generate strategy schemes. According to the teaching task requirements, it selects the strategy scheme type to generate a target scheme. Combining learner profiles, it injects learner coordinates into the target scheme to generate a scene parameter set and integrates it with a dynamic narrative script. Based on the scene parameter set, the dynamic narrative script, and the control protocol, it selects three types of teaching elements—content, interaction, and feedback—from a resource library to generate component schemes. Through a scene compiler, it compiles the component schemes, integrates and optimizes the three types of teaching elements to generate dynamic teaching units, and deploys them in a VR engine to respond to learner behavior, thus realizing the construction of adaptive dynamic VR teaching scenes.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A method for constructing dynamic VR teaching scenarios with adaptive content generation includes:

[0006] Extract scenario feature variables that affect teaching objectives from a pre-set teaching prototype library, and combine them with a multi-dimensional feature extraction model to form a primitive pattern;

[0007] Based on a deep learning model, the dimensions of the primitive patterns are expanded and associated with control protocols to generate strategy schemes;

[0008] Select the strategy option type according to the teaching task requirements, and generate the target option through the solution decision model;

[0009] Learner coordinates are injected into the target solution based on learner profiles to generate a set of scene parameters, and a dynamic narrative script based on a generative model is integrated.

[0010] Based on the scene parameter set, dynamic narrative script, and control protocol, three types of teaching elements are selected from the resource library through a component matching model to generate a component scheme; the three types of teaching elements include content elements, interactive elements, and feedback elements.

[0011] By using a scenario compiler to compile component solutions, the three types of teaching elements are integrated and optimized according to dynamic narrative scripts and control protocols to generate dynamic teaching units. These dynamic teaching units are then deployed in the VR engine to respond to learner behaviors.

[0012] Specifically, the step of extracting scenario feature variables that affect teaching objectives from a pre-set teaching prototype library and combining them with a multi-dimensional feature extraction model to form primitive patterns includes:

[0013] Semantic analysis is performed on the pre-set teaching prototype library. Based on the knowledge point map, skill dimension matrix, and emotional trigger factors decomposed from the teaching objectives, a scenario feature variable type system is established. The scenario feature variable type system includes physical environment, interactive experience, and cognitive guidance.

[0014] A multi-dimensional feature extraction model based on convolutional neural networks is used to extract actual variable values ​​related to the type of scene feature variables from a teaching prototype library; the multi-dimensional feature extraction model based on convolutional neural networks extracts features through multiple convolutional layers and pooling layers, and outputs standardized variable values;

[0015] Based on the extracted actual variable values ​​related to the types of scene feature variables, the weights of each scene feature variable are determined using the analytic hierarchy process (AHP).

[0016] All extracted scene feature variables are organized according to a preset logical structure to form a primitive pattern; the primitive pattern is the sum of the products of all scene feature variables and their corresponding weights, that is, different combinations of values ​​of each scene feature variable constitute different instances of the primitive pattern.

[0017] Specifically, the semantic analysis of the preset teaching prototype library, based on the knowledge point map, skill dimension matrix, and emotional triggering factors decomposed from the teaching objectives, establishes a system of scene feature variable types, including:

[0018] Natural language processing technology is used to perform semantic analysis on the text descriptions in the pre-set teaching prototype library to identify concepts, relationships and attributes related to the teaching objectives;

[0019] Based on the teaching objectives, a knowledge graph construction model is used to decompose knowledge points into knowledge point graphs, and to determine the associations and dependencies between each knowledge point. The knowledge graph construction model uses a graph neural network to model the associations between knowledge points and generate a weighted knowledge association graph.

[0020] Based on the requirements for skills training, a skills dimension matrix is ​​constructed to determine different skills dimensions and their corresponding evaluation criteria and difficulty levels;

[0021] Analyze the emotional factors triggered during the teaching process to identify the emotional triggers;

[0022] Based on semantic analysis results, knowledge point graphs, skill dimension matrices, and emotional triggering factors, scenario feature variables are divided into three categories: physical environment, interactive experience, and cognitive guidance, and a scenario feature variable type system is established.

[0023] Specifically, the step of organizing all extracted scene feature variables according to a preset logical structure to form a primitive pattern includes:

[0024] The preset logical structure is a tree structure, with the teaching objective as the root node, the scene feature variable type as the intermediate node, and the specific scene feature variable as the leaf node.

[0025] Each scene feature variable is multiplied by its corresponding weight to obtain a weighted value for the scene feature variable.

[0026] The weighted values ​​of all scene feature variables are summed to obtain the value of the primitive pattern;

[0027] Different primitive pattern instances are generated for different values ​​of the feature variables of each scenario; each primitive pattern instance corresponds to a combination of teaching scenario features.

[0028] Specifically, the step of expanding the dimension of the primitive pattern based on a deep learning model and associating it with a control protocol to generate a strategy scheme includes:

[0029] The dynamic factors in the teaching process are analyzed, and the extended dimensions are determined through a factor identification model. The factor identification model uses a support vector machine algorithm to classify and identify the dynamic factors.

[0030] The extended dimension is associated with the scene feature variables in the primitive pattern, and a multi-dimensional primitive pattern extended model is constructed using the association model; the association model learns the non-linear relationship between the extended dimension and the scene feature variables through a neural network.

[0031] For each extended dimension, a corresponding control protocol is formulated; the control protocol includes control objectives, control rules, and control strategies.

[0032] The extended primitive pattern is integrated with the control protocol to generate a strategy scheme.

[0033] Specifically, the step of associating the extended dimension with scene feature variables in the primitive pattern and constructing a multi-dimensional primitive pattern extended model using the association model includes:

[0034] An association rule mining algorithm was used to analyze the association between extended dimensions and scene feature variables in primitive patterns.

[0035] Based on the correlation, the extended dimensions are bound to the corresponding scene feature variables to construct a multi-dimensional primitive pattern extension model;

[0036] Set the value range and variation rules for each extended dimension and scene feature variable in the multi-dimensional primitive pattern extension model.

[0037] Specifically, the step of injecting learner coordinates into the target solution based on learner profiles, generating a scene parameter set, and integrating a dynamic narrative script based on a generative model includes:

[0038] Learner profiles are constructed using a learner profile model based on learner attribute information. The learner attribute information includes basic information and learning history. The learner profile model adopts a deep learning model based on an attention mechanism, takes the learner's historical learning data as input, extracts features through a multi-layer neural network, and outputs a learner profile vector. The learner's historical learning data includes answer accuracy, knowledge point mastery time, and interaction frequency.

[0039] Based on the output of the learner profile model, learner coordinates are generated through a coordinate mapping model; the coordinate mapping model adopts an autoencoder structure to compress the learner profile vector into coordinate values.

[0040] Inject learner coordinates into the target solution, and adjust the values ​​of scene feature variables in the target solution based on the learner coordinates;

[0041] A dynamic narrative script is designed by combining the teaching objectives and the adjusted scene characteristic variables; the dynamic narrative script includes the storyline, character settings, and scene transition content.

[0042] Specifically, the step of selecting three types of teaching elements from the resource library based on the scene parameter set, dynamic narrative script, and control protocol, and generating a component scheme, includes:

[0043] Based on the scene feature variable values ​​in the scene parameter set and the requirements of the dynamic narrative script, determine the teaching content type, interaction method, and feedback form;

[0044] The Transformer-based component matching model encodes the features of content components in the resource library, transforms the scene parameter set and dynamic narrative script requirements into query vectors, retrieves and filters content components by calculating the similarity of query vectors, and optimizes the sorting of the filtered content components using a ranking model to obtain optimized content components.

[0045] Select interactive elements from the resource library, and set the functions and behaviors of the interactive elements according to the dynamic narrative script and control protocol; the interactive elements include operation buttons, menus, gesture recognition, and voice interaction;

[0046] Feedback elements are selected from the resource library, and the triggering time and content of the feedback elements are determined based on the learner's learning performance and control protocol; the feedback elements include text prompts, sound prompts, and animation effects;

[0047] The optimized content elements, interactive elements, and feedback elements are combined and arranged to generate a component solution.

[0048] Specifically, the steps of integrating and optimizing the three types of teaching elements according to dynamic narrative scripts and control protocols to generate dynamic teaching units through the scenario compiler component compilation scheme include:

[0049] The scenario compiler parses the component scheme and extracts information for each teaching element; the teaching element information includes element type, attribute parameters, and functional description.

[0050] Based on the dynamic narrative script, determine the triggering timing, execution order, and interaction logic of the teaching elements;

[0051] Based on the control agreement, the teaching components were optimized.

[0052] The three types of integrated and optimized teaching elements are encapsulated according to the requirements of dynamic narrative scripts and control protocols to generate dynamic teaching units; the dynamic teaching units include teaching elements and logic.

[0053] The adaptive content generation dynamic VR teaching scene construction system includes: primitive pattern construction module, strategy scheme generation module, target scheme generation module, parameter set generation module, component scheme generation module, and teaching unit generation module;

[0054] The primitive pattern construction module is used to extract scene feature variables that affect teaching objectives from a preset teaching prototype library and construct primitive patterns.

[0055] The strategy scheme generation module is used to expand the dimensions of the primitive pattern and associate it with the control protocol to generate a strategy scheme.

[0056] The target solution generation module is used to select the strategy solution type according to the teaching task requirements and generate the target solution through the solution decision model;

[0057] The parameter set generation module is used to inject learner coordinates into the target solution by combining learner profiles, generate scene parameter sets, and integrate them into dynamic narrative scripts.

[0058] The component scheme generation module is used to select three types of teaching elements from the resource library and generate a component scheme based on the scene parameter set, dynamic narrative script and control protocol;

[0059] The teaching unit generation module is used to integrate and optimize the three types of teaching elements according to the dynamic narrative script and control protocol through the scene compiler to generate dynamic teaching units, and deploy the dynamic teaching units in the VR engine.

[0060] Compared with the prior art, the beneficial effects of the present invention are:

[0061] 1. This invention proposes a dynamic VR teaching scene construction system with adaptive content generation, and optimizes and improves its architecture, operation steps and processes. The system has the advantages of simple process, low investment and operation costs and low production costs.

[0062] 2. This invention proposes a method for constructing dynamic VR teaching scenarios with adaptive content generation. On the one hand, by extracting scenario feature variables from a teaching prototype library to construct primitive patterns and expanding dimensional correlation control protocols to generate strategy schemes, it can accurately grasp teaching needs and flexibly adapt to dynamic changes in the teaching process, making the teaching scenario more aligned with teaching objectives and actual teaching situations. On the other hand, by combining learner profiles with coordinate injection to generate scenario parameter sets and integrating dynamic narrative scripts, and then rationally selecting three types of teaching elements from a resource library to generate component schemes, and finally compiling and integrating to generate dynamic teaching units and deploying them on a VR engine, this series of operations can fully consider individual differences among learners, provide a highly personalized learning experience, enhance learner participation and learning outcomes, and effectively improve teaching quality and efficiency. Attached Figure Description

[0063] Figure 1 This is a schematic diagram of the adaptive content generation dynamic VR teaching scene construction method of the present invention;

[0064] Figure 2 This is a flowchart illustrating the principle of the adaptive content generation method for constructing dynamic VR teaching scenarios according to the present invention.

[0065] Figure 3 The system architecture diagram for constructing a dynamic VR teaching scene with adaptive content generation according to the present invention is shown. Detailed Implementation

[0066] Example 1:

[0067] Please see Figure 1 and Figure 2 The present invention provides an embodiment of a method for constructing a dynamic VR teaching scene with adaptive content generation, the method comprising steps S1 to S6, including the following steps:

[0068] S1: Extract scenario feature variables that affect teaching objectives from the pre-set teaching prototype library, and combine them with a multi-dimensional feature extraction model to form a primitive pattern;

[0069] S2: Expand the dimensions of the primitive patterns based on the deep learning model and associate them with the control protocol to generate a strategy scheme;

[0070] S3: Select the strategy option type according to the teaching task requirements, and generate the target option through the option decision model;

[0071] Furthermore, the specific steps of S3 include:

[0072] (1) Conduct teaching task analysis to clarify the core elements of teaching tasks, including teaching objectives, teaching content and teaching objects. Among them, teaching objectives refer to the expected results of teaching activities; teaching content refers to the knowledge, skills and other information transmitted in the teaching process; and teaching objects refer to the group of learners participating in teaching activities.

[0073] (2) Based on the results of the teaching task analysis, determine the type of strategy plan, such as lecture-based strategy plan, inquiry-based strategy plan, and collaborative strategy plan;

[0074] If the teaching objective is to impart basic knowledge and skills, the teaching content is logical and systematic, and the target audience is beginners, then the strategy plan type should be a lecture-based strategy plan.

[0075] If the teaching objective is to cultivate students' inquiry ability and innovative thinking, the teaching content is open and challenging, and the target audience is learners with a foundation, then the strategy solution type should be determined as an inquiry-based strategy solution.

[0076] If the teaching objective is to improve students' teamwork and communication skills, the teaching content requires learners to complete it together, and the teaching objects are teams composed of multiple learners, then the strategy solution type should be determined as a collaborative strategy solution.

[0077] (3) Select a strategy solution that matches the teaching task requirements from the pre-generated strategy solution library according to the determined strategy solution type; the strategy solution library contains basic solution templates of different types such as lecture-type strategy solution, inquiry-type strategy solution, and collaborative strategy solution;

[0078] (4) Adjust and optimize the parameters of the selected strategy based on the specific details of the teaching task, such as adjusting the presentation order of the teaching content, setting appropriate interactive difficulty, and determining the frequency of feedback information, to generate the target plan.

[0079] Furthermore, based on the specific details of the teaching task, the selected strategy is adjusted and optimized in terms of parameters, including:

[0080] For lecture-based strategies, the presentation speed and depth of the teaching content should be adjusted according to the difficulty level of the teaching content and the learners' cognitive level; appropriate interactive elements should be set up, such as questions and answers, to improve learners' participation; and the types and frequency of feedback information should be determined, such as providing correct answers and error prompts in a timely manner.

[0081] For inquiry-based strategies, appropriate problem situations and inquiry tasks should be set up based on the openness and challenge of the teaching content; the guidance methods and prompts during the inquiry process should be adjusted to help learners gradually delve deeper into the inquiry; and evaluation criteria and feedback methods for the inquiry results should be determined, such as scoring and commenting on learners' inquiry reports.

[0082] For collaborative strategies, divide the teaching tasks into collaborative groups based on their complexity and team size; set roles and collaboration rules within the groups; and determine communication methods and feedback mechanisms during the collaboration process, such as online discussions and group presentations.

[0083] S4: Combine learner profiles to inject learner coordinates into the target solution, generate a scene parameter set, and integrate a dynamic narrative script based on a generative model;

[0084] S5: Based on the scene parameter set, dynamic narrative script, and control protocol, select three types of teaching elements from the resource library through the element matching model to generate a component scheme; the three types of teaching elements include content elements, interactive elements, and feedback elements;

[0085] S6: By compiling component solutions through a scene compiler, the three types of teaching elements are integrated and optimized according to dynamic narrative scripts and control protocols to generate dynamic teaching units, which are then deployed in the VR engine to respond to learner behavior.

[0086] The step of extracting scenario feature variables that affect teaching objectives from a pre-set teaching prototype library and combining them with a multi-dimensional feature extraction model to form primitive patterns includes:

[0087] S1.1: Perform semantic analysis on the pre-set teaching prototype library, and establish a scenario feature variable type system based on the knowledge point map, skill dimension matrix and emotional trigger factors decomposed from the teaching objectives; the scenario feature variable type system includes physical environment, interactive experience and cognitive guidance.

[0088] S1.2: Use a multi-dimensional feature extraction model based on convolutional neural networks to extract actual variable values ​​related to the type of scene feature variables from the teaching prototype library; the multi-dimensional feature extraction model based on convolutional neural networks extracts features through multiple convolutional layers and pooling layers, and outputs standardized variable values;

[0089] Furthermore, the specific steps in S1.2 include:

[0090] (1) Clarify the types of scene characteristic variables, including physical environment scene characteristic variables, interactive experience scene characteristic variables, and cognitive guidance scene characteristic variables. Among them, physical environment scene characteristic variables cover classroom layout, lighting conditions, sound effects, etc.; interactive experience scene characteristic variables cover the frequency of use of different operation methods, the degree of acceptance of feedback forms, and the degree of participation in collaborative modes, etc.; cognitive guidance scene characteristic variables cover the difficulty of problem guidance, the typicality of case analysis, and the effectiveness of prompt information, etc.

[0091] (2) Construct a multi-dimensional feature extraction model based on a convolutional neural network; the multi-dimensional feature extraction model includes multiple convolutional layers and pooling layers, used to extract deep features from the input data and output standardized variable values. The teaching prototype library is used as the data input source for the multi-dimensional feature model. The teaching prototype library contains various types of data related to scene features, such as scene images, user interaction data, and teaching content. Among them, the convolutional neural network is the existing technology in this field and is not an inventive solution of this application, so it will not be described in detail here.

[0092] (3) For the physical environment scene feature variables, a multi-dimensional feature extraction model based on convolutional neural network is used. The image recognition module in the multi-dimensional feature extraction model is used to analyze the scene images in the teaching prototype library. Deep features in the scene images are extracted through multiple convolutional layers and pooling layers, and then the actual variable values ​​such as classroom layout and lighting conditions are extracted. At the same time, the audio data in the teaching prototype library is processed using the multi-dimensional feature extraction model to extract the actual variable values ​​related to sound effects such as volume and audio type. Among them, image recognition is the existing technology in this field and is not the inventive solution of this application. It will not be described in detail here.

[0093] (4) For interactive experience scenario feature variables, relying on the multi-dimensional feature extraction model based on convolutional neural network, log analysis and mining of user interaction data in the teaching prototype library are performed, and deep features in the interaction data are extracted through multiple convolutional layers and pooling layers. The actual variable values ​​such as the frequency of use of different operation methods, the degree of acceptance of feedback forms, and the degree of participation in collaboration modes are statistically obtained.

[0094] (5) For the characteristic variables of cognitive guidance scenarios, the teaching content in the teaching prototype library is processed by using the text mining module in the multi-dimensional feature extraction model based on convolutional neural network, and deep features in the teaching text are extracted through multiple convolutional layers and pooling layers to complete keyword extraction and theme analysis, and determine the actual variable values ​​such as the difficulty of problem guidance, the typicality of case analysis, and the effectiveness of prompt information.

[0095] (6) By integrating the multi-dimensional data mining results of the multi-dimensional feature extraction model based on convolutional neural networks in the process of extracting feature variables in physical environment, interactive experience and cognitive guidance scenarios, the actual variable values ​​corresponding to the feature variables of each type of scenario are integrated to obtain a set of actual variable values ​​related to the feature variables of each type of scenario, and the variable values ​​in this set are the standardized variable values ​​output by the model.

[0096] S1.3: Based on the extracted actual variable values ​​related to the type of scene feature variables, the weight of each scene feature variable is determined using the analytic hierarchy process (AHP).

[0097] Furthermore, the specific steps in S1.3 include:

[0098] (1) Construct a hierarchical structure model, with teaching objectives as the target layer, scenario feature variable types as the criteria layer, and each specific scenario feature variable as the solution layer;

[0099] (2) An evaluation group composed of experts and teachers was invited to compare the relative importance of each element in the criteria layer and the scheme layer in pairs using the 1-9 scale method to construct a judgment matrix. The 1-9 scale method is the existing technology in this field and is not an inventive solution of this application, so it will not be described in detail here.

[0100] (3) Perform a consistency check on the judgment matrix. If the check passes, calculate the largest eigenvalue of the judgment matrix and its corresponding eigenvector. After normalizing the eigenvector, obtain the weights of the feature variables of each scene at the corresponding level.

[0101] (4) The weighted average method is used to combine the weights of the criteria layer and the scheme layer to determine the final weight of each scenario characteristic variable relative to the teaching objective. The weighted average method is existing technology in this field and is not an inventive solution of this application. It will not be elaborated here.

[0102] S1.4: Organize all extracted scene feature variables according to a preset logical structure to form a primitive pattern; the primitive pattern is the sum of the products of all scene feature variables and their corresponding weights, that is, different combinations of values ​​of each scene feature variable constitute different instances of the primitive pattern.

[0103] The semantic analysis of the pre-set teaching prototype library, based on the knowledge point map, skill dimension matrix, and emotional trigger factors decomposed from the teaching objectives, establishes a system of scene feature variable types, including:

[0104] S1.1.1: Natural language processing technology is used to perform semantic analysis on the text descriptions in the pre-set teaching prototype library to identify the concepts, relationships and attributes related to the teaching objectives. Natural language processing technology is existing technology in this field and is not an inventive solution of this application, so it will not be described in detail here.

[0105] S1.1.2: Based on the teaching objectives, a knowledge graph construction model is used to decompose knowledge points into knowledge point graphs, and to determine the associations and dependencies between each knowledge point; the knowledge graph construction model uses a graph neural network to model the associations between knowledge points and generate a weighted knowledge association graph.

[0106] S1.1.3: Based on the skill training requirements, construct a skill dimension matrix to determine different skill dimensions and their corresponding evaluation standards and difficulty levels; each row in the skill dimension matrix represents a different skill, and each column represents the evaluation score for different skills;

[0107] S1.1.4: Analyze the emotional factors triggered during the teaching process and identify the emotional triggering factors; the emotional triggering factors include learning interest, learning motivation, and learning pressure;

[0108] S1.1.5: Based on the semantic analysis results, knowledge point graph, skill dimension matrix and emotional triggering factors, the scene feature variables are divided into three categories: physical environment, interactive experience and cognitive guidance, and a scene feature variable type system is established.

[0109] The categories include physical environment (classroom layout, lighting conditions, sound effects), interactive experience (operation methods, feedback formats, collaboration modes), and cognitive guidance (question guidance, case analysis, prompts).

[0110] All extracted scene feature variables are organized according to a preset logical structure to form primitive patterns, including:

[0111] S1.4.1: The preset logical structure is a tree structure, with the teaching objective as the root node, the scene feature variable type as the intermediate node, and the specific scene feature variable as the leaf node;

[0112] S1.4.2: Multiply each scene feature variable by its corresponding weight to obtain the weighted value of that scene feature variable;

[0113] S1.4.3: Sum the weighted values ​​of all scene feature variables to obtain the value of the primitive pattern;

[0114] S1.4.4: Generate different primitive pattern instances for different values ​​of each scene feature variable; each primitive pattern instance corresponds to a teaching scene feature combination.

[0115] The process of extending the dimension of the primitive pattern based on a deep learning model and associating it with a control protocol to generate a strategy scheme includes:

[0116] S2.1: Analyze the dynamic changing factors in the teaching process and determine the extended dimensions through a factor identification model; the factor identification model uses the support vector machine algorithm to classify and identify the dynamic changing factors.

[0117] The factor identification model, as a component of the deep learning model, uses the support vector machine algorithm to classify and identify dynamically changing factors, outputs the evaluation results of the degree of influence of different dynamically changing factors on the teaching process, and then determines the dimensions that need to be expanded, such as the time dimension and the learning state dimension. The support vector machine algorithm is existing technology in this field and is not an inventive solution of this application, so it will not be described in detail here.

[0118] Furthermore, the specific steps of S2.1 include:

[0119] (1) Identify the dynamic factors in the teaching process, including learner factors and teaching factors; learner factors include the learning progress change pattern, learning effect fluctuation, and learning preference differences; teaching factors are determined based on teaching objectives and teaching plans, and involve factors that may affect the teaching process, such as adjustments to the difficulty of teaching content and changes in teaching pace.

[0120] (2) Constructing a factor identification model; the factor identification model adopts the support vector machine algorithm to classify and identify dynamic changing factors in the teaching process. The support vector machine algorithm is the existing technology in this field and is not an inventive solution of this application. It will not be described in detail here.

[0121] (3) Analyze the learner’s historical learning data stored in the teaching prototype library for learner factors, including: extracting key indicators from the historical learning data, such as the mastery of knowledge points at each stage, changes in test scores, and records of learning resource selection, sorting out the learning progress change pattern, learning effect fluctuation and learning preference differences of learners, and inputting these analysis results into the factor identification model as specific data of learner-related dynamic change factors.

[0122] (4) Combining the pre-set teaching objectives and teaching plans, sort out the teaching factors. Among them, the teaching objectives are the expected results of the teaching activities, and the teaching plan is the specific arrangement of the teaching process. Based on this, evaluate the degree of influence of different teaching factors on the teaching process, and input the influence degree evaluation data into the factor identification model.

[0123] (5) The factor identification model classifies and identifies the input learner factor data and teaching factor data to obtain a quantitative assessment of the degree of influence of different dynamic factors on the teaching process;

[0124] (6) Based on the quantitative evaluation results output by the factor identification model, determine the extended dimensions; when the evaluation results show that the learners' learning progress varies greatly and has a significant impact on the teaching process, determine the time dimension as the dimension that needs to be extended; when the evaluation results show that the learners' learning effect is closely related to their learning preferences and that this association has a prominent impact on the teaching process, determine the learning status dimension as the dimension that needs to be extended, thus forming a complete closed loop from the analysis of dynamic changing factors to the determination of extended dimensions.

[0125] S2.2: Associate the extended dimension with the scene feature variables in the primitive pattern, and use the association model to construct a multi-dimensional primitive pattern extended model; the association model learns the non-linear relationship between the extended dimension and the scene feature variables through a neural network;

[0126] S2.3: For each extended dimension, formulate a corresponding control protocol, which includes control objectives, control rules, and control strategies;

[0127] Among them, the control target is set based on the teaching objective, the control rules are formulated according to the changing patterns of scenario characteristic variables, and the control strategy is determined according to the characteristics of the extended dimension, so as to achieve precise control of the teaching scenario.

[0128] For example, regarding the time dimension, based on the time requirements of the teaching task and the learners' average learning progress, time control goals are set, such as ensuring that most learners can complete the learning of the teaching content within the specified teaching time; time control rules are set, such as triggering the simplification or acceleration of the teaching content when the learners' learning progress lags behind the average progress by more than a certain percentage; and providing extended learning content or increasing learning challenges when the learners' learning progress is ahead of the average progress by more than a certain percentage; and time control strategies are designed, such as using methods such as dynamically adjusting the playback speed of the teaching content, skipping some non-critical content, and providing fast learning channels to achieve time control.

[0129] For example, regarding the learning state dimension, based on learners' learning effectiveness evaluation results and learning preference analysis, learning state regulation goals are formulated, such as improving learners' learning interest and participation; learning state regulation rules are formulated, such as adjusting the difficulty and presentation of teaching content when learners' learning effectiveness is poor; and adding visual elements such as pictures and videos when learners' learning preferences are visual; and learning state regulation strategies are designed, such as using personalized recommendations of learning content, providing diverse learning activities, and giving timely feedback and encouragement to achieve learning state regulation.

[0130] S2.4: Integrate the extended primitive pattern with the control protocol to generate a strategy scheme.

[0131] Furthermore, the integration process is optimized using deep learning models to ensure that the expanded primitive patterns and control protocols are logically consistent and can work synergistically according to the dynamic changes in the teaching process. After integration, a strategy plan that meets the needs of the teaching task is generated, forming a complete closed loop from dynamic factor analysis, dimension expansion, protocol formulation to plan generation.

[0132] The step of associating the extended dimension with scene feature variables in the primitive pattern and constructing a multi-dimensional primitive pattern extended model using the association model includes:

[0133] S2.2.1: An association rule mining algorithm is used to analyze the association between the extended dimension and the scene feature variables in the primitive pattern, and to determine which scene feature variables will be affected by the extended dimension. In this invention, the association rule mining algorithm adopts the Apriori algorithm. The Apriori algorithm is the prior art in this field and is not an inventive solution of this application, so it will not be described in detail here.

[0134] S2.2.2: Based on the correlation, the extended dimensions are bound to the corresponding scene feature variables to construct a multi-dimensional primitive pattern extension model. For example, the time dimension is associated with scene feature variables such as the presentation speed of teaching content and the response time of interactive operations; the learning state dimension is associated with scene feature variables such as the difficulty of question guidance and the level of detail of feedback information.

[0135] Furthermore, the process of constructing a multi-dimensional primitive pattern extension model includes:

[0136] (1) Analyze dynamic factors in teaching, such as changes in learners’ learning progress, learning style, and learning environment, updates to teaching resources, and adjustments to teaching objectives and tasks. For example, in online learning scenarios, learners’ network conditions may affect their learning experience, which are dynamic factors that need to be considered.

[0137] (2) Select key extended dimensions that have a significant impact on the teaching scenario from dynamic factors;

[0138] (3) Collect and expand data related to dimensions and scenario characteristic variables, including learners’ historical learning data and real-time data during the teaching process;

[0139] (4) Use the Apriori algorithm to analyze the relationship between the extended dimension and scene feature variables;

[0140] (5) Based on the discovered relationships, establish binding rules between extended dimensions and scene feature variables. The rules can clearly specify how the corresponding scene feature variables should be adjusted when a certain extended dimension takes a specific value.

[0141] (6) Implement binding rules in the system to bind the extended dimensions with the corresponding scene feature variables. The logical judgment of the rules and the assignment of variables can be implemented through programming to ensure that the scene feature variables can be automatically adjusted when the extended dimensions change. For example, when the system detects that the learner's programming basic level has improved from intermediate to advanced, it will automatically adjust the difficulty of the teaching content from medium to difficult.

[0142] (7) Design the overall structure of the multi-dimensional primitive pattern extension model, determine the hierarchy and components of the model, for example, take the teaching objective as the root node, and the extended dimensions and scene feature variables as child nodes to construct a tree-structured extended model;

[0143] (8) Clarify the meaning and attributes of each element in the multi-dimensional primitive pattern extension model, including extension dimensions, scene feature variables, binding rules, etc., and define clear names, types, value ranges and other attributes for each element to facilitate the management and use of the model. For example, define the type of learning time arrangement as discrete, with a value range of 1-2 hours per day, 3-4 hours per day, and more than 5 hours per day.

[0144] (9) Set reasonable value ranges and change rules for each dimension and scene feature variable in the multi-dimensional primitive pattern extension model. The value range should be determined according to the actual situation and teaching objectives. For example, set the difficulty value range of the teaching content as simple, medium, difficult, and very difficult.

[0145] (10) Use actual teaching data to verify the constructed multi-dimensional primitive pattern extension model, check whether the model can accurately reflect the relationship between the extended dimensions and the scene feature variables, and whether it can meet the dynamic generation requirements of the teaching scene. Optimize and adjust the model according to the verification results to obtain the constructed multi-dimensional primitive pattern extension model.

[0146] S2.2.3: Set the value range and change rules for each extended dimension and scene feature variable in the multi-dimensional primitive pattern extension model.

[0147] The process of injecting learner coordinates into the target solution based on learner profiles, generating a scene parameter set, and integrating it into a dynamic narrative script based on a generative model includes:

[0148] S4.1: Based on learner attribute information, a learner profile is constructed using a learner profile model. The learner attribute information includes basic information and learning history. The learner profile model employs a deep learning model based on an attention mechanism, taking learner historical learning data as input, extracting features through a multi-layer neural network, and outputting a learner profile vector containing dimensions of learning style, knowledge weaknesses, and learning pace. The learner historical learning data includes answer accuracy, knowledge point mastery time, and interaction frequency. The multi-layer neural network is existing technology in this field and is not an inventive solution of this application; therefore, it will not be elaborated upon here.

[0149] The construction of learner profiles is based on machine learning models, which are existing technologies in this field and are not the inventive solutions of this application, and will not be described in detail here.

[0150] S4.2: Based on the output of the learner profile model, learner coordinates are generated through a coordinate mapping model; the coordinate mapping model adopts an autoencoder structure to compress the high-dimensional learner profile vector into low-dimensional coordinate values.

[0151] The learner coordinates include multiple dimensions, such as cognitive style dimension, knowledge mastery dimension, and preference dimension.

[0152] S4.3: Inject learner coordinates into the target solution and adjust the scene feature variable values ​​in the target solution according to the learner coordinates;

[0153] For example, the presentation of teaching content can be adjusted according to learners' cognitive styles, the difficulty of teaching content can be adjusted according to learners' knowledge levels, and the theme of teaching content can be adjusted according to learners' interests and preferences.

[0154] S4.4: Design a dynamic narrative script by combining the teaching objectives and the adjusted scene characteristic variable values; the dynamic narrative script includes the storyline, character settings, and scene transition content.

[0155] The step involves selecting three types of teaching elements from the resource library based on the scene parameter set, dynamic narrative script, and control protocol, and generating a component scheme, including:

[0156] S5.1: Based on the scene feature variable values ​​in the scene parameter set and the requirements of the dynamic narrative script, determine the required teaching content type, interaction method, and feedback form;

[0157] S5.2: The Transformer-based component matching model encodes the features of content components in the resource library, transforms the scene parameter set and dynamic narrative script requirements into query vectors, retrieves and filters content components by calculating the similarity of the query vectors, and optimizes and sorts the filtered content components using a ranking model to obtain optimized content components; the content components include text, images, videos, audio, 3D models, etc.; the filtering and optimization of content components includes adjusting the difficulty, length, and style of the content;

[0158] S5.3: Select interactive elements from the resource library, and set the functions and behaviors of the interactive elements according to the dynamic narrative script and control protocol; the interactive elements include operation buttons, menus, gesture recognition, and voice interaction; the functions and behaviors of the interactive elements include the triggering conditions, response methods, and operation processes of the interaction;

[0159] S5.4: Select feedback elements from the resource library, and determine the triggering time and content of the feedback elements based on the learner's learning performance and control protocol; the feedback elements include text prompts, sound prompts, animation effects, and reward mechanisms; the feedback content includes giving praise and rewards when the learner answers correctly, and giving prompts and corrections when the learner answers incorrectly.

[0160] S5.5: Combine and arrange the optimized content elements, interactive elements, and feedback elements to generate a component scheme, ensuring that the elements can work together to achieve the teaching objectives.

[0161] The specific steps of the scenario compiler-based component compilation scheme, which integrates and optimizes the three types of teaching elements according to dynamic narrative scripts and control protocols to generate dynamic teaching units, include:

[0162] S6.1: The scene compiler parses the component scheme and extracts the information of each teaching element; the teaching element information includes element type, attribute parameters, and functional description;

[0163] S6.2: Based on the dynamic narrative script, determine the triggering time, execution order and interaction logic of the teaching elements. For example, when a specific plot occurs, trigger the corresponding interactive elements and display the content elements in a preset order.

[0164] S6.3: Optimize the teaching elements according to the control protocol; the optimization includes, but is not limited to, format conversion of content elements, response speed optimization of interactive elements, and personalized settings of feedback elements.

[0165] S6.4: The three types of integrated and optimized teaching elements are encapsulated according to the requirements of dynamic narrative scripts and control protocols to generate dynamic teaching units; the dynamic teaching units contain all the elements and logic required for teaching.

[0166] Example 2:

[0167] Please see Figure 3 Another embodiment of the present invention provides: a dynamic VR teaching scene construction system with adaptive content generation, comprising:

[0168] The module includes: primitive pattern construction module, strategy scheme generation module, target scheme generation module, parameter set generation module, component scheme generation module, and teaching unit generation module.

[0169] The primitive pattern construction module is used to extract scenario feature variables that affect teaching objectives from a pre-set teaching prototype library, construct primitive patterns, and provide a foundation for subsequent strategy and solution generation.

[0170] The strategy scheme generation module is used to expand the dimensions of the primitive pattern and associate it with the control protocol to generate strategy schemes to adapt to dynamic changes in the teaching process.

[0171] The target solution generation module is used to select the strategy solution type according to the teaching task requirements, generate the target solution through the solution decision model, and prepare for the subsequent generation of scenario parameter sets.

[0172] The parameter set generation module is used to inject learner coordinates into the target solution by combining learner profiles, generate scene parameter sets, and integrate dynamic narrative scripts to make the teaching scene more in line with the characteristics of learners.

[0173] The component scheme generation module is used to select three types of teaching elements from the resource library based on the scene parameter set, dynamic narrative script and control protocol, and generate component schemes to provide materials for the generation of dynamic teaching units.

[0174] The teaching unit generation module is used to compile component schemes through a scene compiler, integrate and optimize the three types of teaching elements according to dynamic narrative scripts and control protocols, generate dynamic teaching units, and deploy dynamic teaching units in the VR engine to respond to learner behavior.

[0175] The primitive pattern construction module includes: a semantic analysis unit, a weight determination unit, and a primitive pattern organization unit;

[0176] The semantic analysis unit is used to perform semantic parsing on the text descriptions in the pre-set teaching prototype library using natural language processing technology.

[0177] The weight determination unit is used to determine the weight of each scene feature variable based on the extracted actual variable values ​​related to the type of scene feature variables using the analytic hierarchy process.

[0178] The primitive pattern organization unit is based on a tree structure with a pre-defined logical structure. The teaching objective is the root node, the scene feature variable type is the intermediate node, and the specific scene feature variable is the leaf node. The primitive pattern value is calculated to generate different primitive pattern instances.

[0179] The strategy solution generation module includes: dimensional analysis unit, model building unit, protocol formulation unit, and strategy solution integration unit;

[0180] The dimensional analysis unit is used to analyze dynamic and changing factors in the teaching process and determine the dimensions that need to be expanded.

[0181] The model building unit is used to analyze the relationship between the extended dimensions and scene feature variables in the primitive patterns using association rule mining algorithms, and to build a multi-dimensional primitive pattern extended model.

[0182] The protocol formulation unit is used to formulate corresponding control protocols for each extended dimension, including control objectives, control rules, and control strategies.

[0183] The strategy scheme integration unit is used to integrate the extended primitive pattern with the control protocol to generate a strategy scheme.

[0184] The target solution generation module includes: a requirements analysis unit and a solution selection unit;

[0185] The requirements analysis unit is used to analyze the requirements of teaching tasks, such as specific teaching objectives and teaching time constraints.

[0186] The scheme selection unit is used to select the appropriate strategy scheme type according to the needs of the teaching task and generate the target scheme.

[0187] The parameter set generation module includes: a profile construction unit, a coordinate transformation unit, a scheme adjustment unit, and a script design unit;

[0188] The profile building unit is used to build learner profiles based on learner attribute information and machine learning models.

[0189] The coordinate transformation unit is used to analyze learner profile features and transform learner profile features into learner coordinates.

[0190] The scheme adjustment unit is used to inject learner coordinates into the target scheme and adjust the scene feature variable values ​​in the target scheme according to the learner coordinates.

[0191] The script design unit is used to design dynamic narrative scripts, including storylines, character settings, and scene transition content, by combining teaching objectives and adjusted scene characteristic variable values.

[0192] The component solution generation module includes: a requirements determination unit, a component selection unit, and a component arrangement unit;

[0193] The requirement determination unit is used to determine the required teaching content type, interaction method, and feedback form based on the scene feature variable values ​​in the scene parameter set and the requirements of the dynamic narrative script;

[0194] The component selection unit is used to select content components, interactive components, and feedback components from the resource library;

[0195] The component orchestration unit is used to combine and orchestrate selected content elements, interactive elements, and feedback elements to generate component solutions.

[0196] The teaching unit generation module includes: solution analysis unit, logic determination unit, component optimization unit, packaging unit, and deployment response unit;

[0197] The scheme parsing unit is used by the scenario compiler to parse the component scheme and extract information for each teaching element, including element type, attribute parameters, and functional description.

[0198] The logic determination unit is used to determine the triggering timing, execution order, and interaction logic of teaching elements based on the dynamic narrative script;

[0199] The component optimization unit is used to optimize teaching components according to the control protocol.

[0200] The encapsulation unit is used to encapsulate the integrated and optimized three types of teaching elements according to the requirements of dynamic narrative scripts and control protocols to generate dynamic teaching units;

[0201] Deploy response units are used to deploy dynamic instructional units within the VR engine, enabling them to respond to learner behavior.

[0202] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments under the guidance of the present invention without departing from the spirit and scope of the present invention. All of these variations are within the protection scope of the present invention.

[0203] If the technical solution disclosed herein involves personal information, the product using this technical solution has clearly informed the user of the personal information processing rules and obtained the user's voluntary consent before processing the personal information. If the technical solution disclosed herein involves sensitive personal information, the product using this technical solution has obtained the user's separate consent before processing the sensitive personal information, and also meets the requirement of "express consent". For example, at personal information collection devices such as cameras, clear and prominent signs are set up to inform users that they have entered the scope of personal information collection and that personal information will be collected. If an individual voluntarily enters the collection scope, it is deemed that they have agreed to the collection of their personal information; or on the personal information processing device, with clear signs / information informing users of the personal information processing rules, authorization is obtained from the individual through pop-up information or by asking the individual to upload their personal information; wherein, the personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the types of personal information processed.

Claims

1. A dynamic VR teaching scenario construction method of adaptive content generation, characterized in that, The method comprises the following steps: Extracting scene characteristic variables affecting the teaching target from a preset teaching prototype library, combining a multi-dimensional feature extraction model to form a primitive mode; Expanding the dimension of the primitive mode based on a deep learning model and associating a regulation and control protocol to generate a strategy scheme; Selecting a strategy scheme type according to the teaching task requirement, and generating a target scheme through a scheme decision model; Injecting learner coordinates into the target scheme in combination with a learner portrait to generate a scene parameter set, and integrating a dynamic narrative script based on a generative model; According to the scene parameter set, the dynamic narrative script and the regulation and control protocol, selecting three types of teaching elements from a resource library through an element matching model to generate a component scheme; the three types of teaching elements include content elements, interactive elements and feedback elements; Compiling the component scheme through a scene compiler, integrating and optimizing the three types of teaching elements according to the dynamic narrative script and the regulation and control protocol to generate a dynamic teaching unit, and deploying the dynamic teaching unit in a VR engine to respond to learner behavior; The method comprises the following steps: Performing semantic analysis on the preset teaching prototype library, establishing a scene characteristic variable type system according to the knowledge point graph, skill dimension matrix and emotional trigger factor decomposed from the teaching target; the scene characteristic variable type system includes physical environment, interactive experience and cognitive guidance; Using a multi-dimensional feature extraction model based on a convolutional neural network to extract actual variable values related to the scene characteristic variable type from the teaching prototype library; the multi-dimensional feature extraction model based on the convolutional neural network extracts features through multiple convolutional layers and pooling layers, and outputs standardized variable values; Determining the weight of each scene characteristic variable by using the analytic hierarchy process according to the extracted actual variable values related to the scene characteristic variable type; Organizing all the extracted scene characteristic variables according to a preset logical structure to form a primitive mode; the primitive mode is the sum of the product of all scene characteristic variables and their corresponding weights, that is, different value combinations of each scene characteristic variable constitute different instances of the primitive mode; The method comprises the following steps: Analyzing the dynamic change factors in the teaching process, and determining the expansion dimension through a factor identification model; the factor identification model uses a support vector machine algorithm to classify and identify the dynamic change factors; Correlating the expansion dimension with the scene characteristic variables in the primitive mode, and constructing a multi-dimensional primitive mode expansion model by using a correlation model; the correlation model learns the nonlinear relationship between the expansion dimension and the scene characteristic variables through a neural network; Formulating a corresponding regulation and control protocol for each expansion dimension; the regulation and control protocol includes a regulation and control target, a regulation and control rule and a regulation and control strategy; Integrating the expanded primitive mode with the regulation and control protocol to generate a strategy scheme.

2. The dynamic VR teaching scenario construction method of adaptive content generation of claim 1, wherein, The method comprises the following steps: Performing semantic analysis on the preset teaching prototype library, establishing a scene characteristic variable type system according to the knowledge point graph, skill dimension matrix and emotional trigger factor decomposed from the teaching target; the scene characteristic variable type system includes physical environment, interactive experience and cognitive guidance; The natural language processing technology is used for semantic analysis on text descriptions in the preset teaching prototype library, and concepts, relations and attributes related to the teaching target are identified; Based on the teaching target, a knowledge graph construction model is used to decompose knowledge points into a knowledge point graph, and the association and dependency relationship between the knowledge points are determined; the knowledge graph construction model uses a graph neural network to model the knowledge point association relationship, and generates a weighted knowledge association graph; According to the skill training requirements, a skill dimension matrix is constructed to determine different skill dimensions and their corresponding evaluation standards and difficulty levels; Analyze the emotional factors triggered in the teaching process to determine the emotional trigger factors; According to the semantic analysis results, the knowledge point graph, the skill dimension matrix and the emotional trigger factors, the scene characteristic variables are divided into three categories: physical environment, interactive experience and cognitive guidance, and a scene characteristic variable type system is established.

3. The dynamic VR teaching scenario construction method of adaptive content generation of claim 2, wherein, The extracted all scene characteristic variables are organized according to the preset logical structure to form the primitive mode, including: The preset logical structure is a tree structure, taking the teaching target as the root node, the scene characteristic variable type as the intermediate node, and the specific scene characteristic variable as the leaf node; Multiply each scene characteristic variable by its corresponding weight to get the weighted value of the scene characteristic variable; Add the weighted values of all scene characteristic variables to get the value of the primitive mode; Different primitive mode instances are generated for different values of each scene characteristic variable; each primitive mode instance corresponds to a teaching scene characteristic combination.

4. The dynamic VR teaching scenario construction method of adaptive content generation of claim 3, wherein, The extension dimension is associated with the scene characteristic variable in the primitive mode, and a multi-dimensional primitive mode extension model is constructed by using an association model, including: An association rule mining algorithm is used to analyze the association relationship between the extension dimension and the scene characteristic variable in the primitive mode; According to the association relationship, the extension dimension is bound to the corresponding scene characteristic variable to construct a multi-dimensional primitive mode extension model; Set the value range and change rule for each extension dimension and scene characteristic variable in the multi-dimensional primitive mode extension model.

5. The dynamic VR teaching scenario construction method of adaptive content generation as claimed in claim 1, wherein, The learner coordinates are injected into the target scheme by combining the learner portrait, a scene parameter set is generated, and a dynamic narrative script based on the generated model is integrated, including: According to the learner attribute information, a learner portrait is constructed by a learner portrait model; the learner attribute information includes basic information and learning history; the learner portrait model uses a deep learning model based on an attention mechanism, takes the learner historical learning data as input, extracts features through a multi-layer neural network, and outputs a learner portrait vector; the learner historical learning data includes correct answer rate, knowledge point mastery time, and interactive operation frequency; Based on the output result of the learner portrait model, a learner coordinate is generated by a coordinate mapping model; the coordinate mapping model uses an autoencoder structure to compress the learner portrait vector into a coordinate value; The learner coordinates are injected into the target scheme, and the scene characteristic variable values in the target scheme are adjusted according to the learner coordinates; Combine the teaching target and the adjusted scene characteristic variable values to design a dynamic narrative script; the dynamic narrative script includes plot, character setting and scene transition content.

6. The dynamic VR teaching scenario construction method of adaptive content generation as claimed in claim 1, wherein, The component scheme is generated by selecting three types of teaching elements from a resource library according to the scene parameter set, the dynamic narrative script, and a regulation protocol, including: According to the scene characteristic variable values in the scene parameter set and the requirements of the dynamic narrative script, the teaching content type, the interaction mode, and the feedback form are determined; The content elements in the resource library are feature-encoded based on the element matching model of the Transformer, the scene parameter set and the dynamic narrative script requirements are converted into query vectors, the content elements are retrieved and screened by calculating the similarity of the query vectors, and the screened content elements are optimized by using a ranking model to obtain optimized content elements; Interactive elements are selected from the resource library, and the functions and behaviors of the interactive elements are set according to the dynamic narrative script and the regulation protocol; the interactive elements include operation buttons, menus, gesture recognition, and voice interaction; Feedback elements are selected from the resource library, and the triggering time and feedback content of the feedback elements are determined according to the learning performance of the learners and the regulation protocol; the feedback elements include text prompts, sound prompts, and animation effects; The optimized content elements, interactive elements, and feedback elements are combined and arranged to generate a component scheme.

7. The dynamic VR teaching scenario construction method of adaptive content generation as claimed in claim 1, wherein, The specific steps of compiling the component scheme by the scene compiler to integrate and optimize the three types of teaching elements according to the dynamic narrative script and the regulation protocol to generate a dynamic teaching unit include: The scene compiler analyzes the component scheme and extracts the information of each teaching element; the teaching element information includes element type, attribute parameter, and function description; According to the dynamic narrative script, the triggering time, execution order, and interaction logic of the teaching elements are determined; According to the regulation protocol, the teaching elements are optimized; The integrated and optimized three types of teaching elements are encapsulated according to the requirements of the dynamic narrative script and the regulation protocol to generate a dynamic teaching unit; the dynamic teaching unit includes teaching elements and logic.

8. A dynamic VR teaching scenario construction system for adaptive content generation, for implementing the method of dynamic VR teaching scenario construction for adaptive content generation according to any one of claims 1-7, characterized in that, It includes: Primitive mode construction module, strategy scheme generation module, target scheme generation module, parameter set generation module, component scheme generation module, and teaching unit generation module; The primitive mode construction module is used to extract the scene characteristic variables that affect the teaching target from the preset teaching prototype library to construct a primitive mode; The strategy scheme generation module is used to expand the dimension of the primitive mode and associate the regulation protocol to generate a strategy scheme; The target scheme generation module is used to select the strategy scheme type according to the teaching task requirements and generate a target scheme through a scheme decision model; The parameter set generation module is used to combine the learner portrait with the learner coordinates to the target scheme to generate a scene parameter set and integrate a dynamic narrative script; The component scheme generation module is used to select three types of teaching elements from a resource library according to the scene parameter set, the dynamic narrative script, and a regulation protocol to generate a component scheme; The teaching unit generation module is used to compile the component scheme by the scene compiler to integrate and optimize the three types of teaching elements according to the dynamic narrative script and the regulation protocol to generate a dynamic teaching unit, and deploy the dynamic teaching unit in a VR engine.

Images