Intelligently-driven educational resource sharing virtual cooperation method

Abstract

The invention relates to the technical field of intelligent education, in particular to an intelligent-driven educational resource sharing virtual cooperation method. The method aims to solve the technical problems of inaccurate education resource classification, single user contribution evaluation, fixed virtual scene and incomplete right and interest allocation in the prior art. According to the technical scheme, the method comprises the steps that educational resource data are acquired, knowledge entities are mapped through a domain knowledge graph, and subject feature vectors are generated; calculating a user teaching contribution entropy based on a multi-dimensional weighting algorithm; and establishing a dynamic mapping relationship between virtual space rendering parameters and subject feature vectors and between virtual space rendering parameters and teaching contribution entropies, and realizing adaptive evolution of the appearance, layout and system resource configuration of virtual space nodes. The method further comprises the functions of scene generation driven by an unstructured text, multi-party cooperative interaction and intelligent contract right and interest distribution, the resource matching efficiency can be improved, the user contribution can be accurately evaluated, the virtual cooperative experience is optimized, and the original rights and interests of resources are guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent education, in particular to an intelligent driving education resource sharing virtual collaboration method. BACKGROUND

[0002] With the deepening of the digital transformation of education, various technical application attempts have emerged in the field of professional education such as law, medicine, and engineering. On the one hand, immersive systems that use artificial intelligence technology to assist in domain knowledge reasoning and teaching have emerged; for example, some systems can perform semantic analysis on professional texts and reasoning based on domain knowledge graphs. On the other hand, virtual simulation technology is used to build practical teaching platforms, providing students with practical experience by simulating professional scenarios such as virtual courts, virtual operating rooms, and virtual laboratories. In addition, cloud platform-based education resource sharing methods have also been proposed to manage and evaluate course resources.

[0003] However, the existing technical solutions still have obvious limitations: First, most systems are essentially closed or isolated on-campus systems, and it is difficult to accurately identify, access, and effectively utilize high-quality education resources in a wider range; especially for highly specialized disciplines, the existing general tag classification method is difficult to accurately represent the complex sub-domain attributes of the discipline, resulting in barriers between resources of different institutions; Secondly, the existing solutions lack a multi-dimensional evaluation model that can accurately quantify users' continuous teaching contributions, and the resource providers' efforts and rewards are disconnected, often relying on simple download volume or click volume for evaluation, which cannot fully reflect the academic value and interactive value of resources, resulting in insufficient resource sharing motivation and an unsustainable virtuous cycle; Finally, even in a virtual environment, the rendering style of the scene and the system resource configuration are usually fixed and cannot adapt to the attribute differences of different subject content (for example, serious theoretical teaching and active case study require different visual atmospheres), and the virtual space rights and interests of users are static and fragmented, the virtual world cannot dynamically reflect and motivate academic contributions in the real world, and the technical means and teaching incentive goals have not been deeply integrated.

[0004] Therefore, an intelligent driving education resource sharing virtual collaboration method is proposed. SUMMARY

[0005] The purpose of the present application is to provide an intelligent driving education resource sharing virtual collaboration method to solve the technical problems of inaccurate classification of existing education resources, single evaluation of user contributions, fixed virtual scenes, and imperfect allocation of rights and interests.

[0006] To achieve the above technical problems, the purpose of the present application is to provide an intelligent driving education resource sharing virtual collaboration method, comprising the following steps: S1, obtaining the education resource data uploaded by the user node, establishing or calling the knowledge graph of the target field, and determining the subject sub-domain center node in the knowledge graph; mapping the knowledge entity in the education resource data to the node in the knowledge graph, calculating the semantic distance of the extracted knowledge entity in the knowledge graph and the preset subject sub-domain center node, and generating a subject feature vector representing the resource content tendency based on the inverse normalization of the semantic distance; S2, collecting the interaction behavior data of the education resource data within a preset time window, combining the content quality score and reference topology relationship of the resource, and calculating the teaching contribution entropy of the user node by using a multi-dimensional weighted algorithm based on content quality, reference depth, interaction activity derivative and academic index; S3, establishing a dynamic mapping relationship between the virtual space rendering parameters and the subject feature vector and the teaching contribution entropy; according to the dynamic mapping relationship, the appearance parameters of the three-dimensional model, the scene layout scale and the underlying system resource configuration of the virtual space node corresponding to the user node are adjusted in real time, so that the rendering state and system resource quota of the virtual space node are updated to the target level configuration corresponding to the current teaching contribution entropy value.

[0007] As a further improvement of the technical solution, in step S1, the education resource data includes video streaming media, document data and live data stream; When uploading the education resource data, a hash algorithm is used to generate a digital fingerprint of the education resource data, and the digital fingerprint is written into a distributed ledger to generate a non-fungible token, which records the digital signature of the user node to prove that the user node has ownership or related rights to the education resource data.

[0008] As a further improvement of the technical solution, in step S2, the calculation formula of teaching contribution entropy is as follows: , wherein, is the teaching contribution entropy in the preset time window , is the content quality score of the first resource, which is generated by the collaborative scoring of the key frame feature extracted by the pre-trained visual quality evaluation model and the metadata specification analysis model; is the total number of education resource data uploaded by the user node within the preset time window; is the reference depth weight, which is determined according to the in-degree centrality value of the resource in the knowledge graph reference relationship network; is the reference depth weight, which is determined according to the in-degree centrality value of the resource in the knowledge graph reference relationship network; The interaction activity level is a weighted sum of the number of valid resource visits, average dwell time, and number of interactive comments within a preset time window. This is the time derivative of the interaction activity. As an academic index, it is calculated by weighting and normalizing the total number of downloads and citations of the user node's historical uploaded resources according to preset weights. The configurable weight coefficients associated with resource types are determined by looking up the default configuration table based on the media type to which the resource belongs.

[0009] As a further improvement to this technical solution, step S3, establishing a dynamic mapping relationship between virtual space rendering parameters and the subject feature vector, specifically includes: A preset style material library, which contains multiple sets of texture materials, lighting models and architectural component templates corresponding to different discipline subdomains; Calculate the weight values ​​of each dimension in the subject feature vector to identify the dominant subject attributes; If the weight value of a single dimension exceeds the preset threshold, the style material corresponding to the dominant discipline attribute will be directly called for rendering. If there is no single dimension exceeding the preset threshold, then the rendering parameters of different style materials are linearly interpolated based on the weight values ​​of each dimension to generate mixed style rendering parameters.

[0010] As a further improvement to this technical solution, the specific process of adjusting the underlying system resource configuration in step S3 includes: Multiple system resource allocation levels are preset, and each level corresponds to a threshold range of teaching contribution entropy; The teaching contribution entropy of the user node is monitored in real time. When the teaching contribution entropy jumps to a higher threshold range, an expansion command is triggered. The expansion command controls the container orchestration system to increase the number of container clusters allocated to the user node, increase the bandwidth weight of the content delivery network, and unlock the data loading permissions of advanced simulation components or large conference components in the virtual space node.

[0011] As a further improvement to this technical solution, the method also includes a scene generation step based on unstructured scene description text: Obtain the user-inputted scenario description text data; By using entity recognition and relationship extraction models, comprehensive information such as scene entities, spatial relationships between entities, environmental description information, and action timing information is extracted from the scene description text data. Based on the extracted comprehensive information, a 3D asset library is retrieved, and matching 3D model assets are called up. The rule engine is used to automatically lay out the three-dimensional model assets into the virtual coordinate system according to the spatial position relationship, generating an initialized three-dimensional virtual simulation scene.

[0012] As a further improvement to this technical solution, after generating the three-dimensional virtual simulation scene, a collaborative interaction step is also included: A multi-party communication channel is established through the WebRTC protocol to receive connection requests initiated by clients from different geographical locations and to connect the clients to the three-dimensional virtual simulation scene; The system collects the displacement, rotation, and scaling operation commands of the current client on the 3D model asset, and broadcasts the operation commands to other clients using frame synchronization or state synchronization technology, driving the rendering engines of other clients to execute the same operation commands.

[0013] As a further improvement to this technical solution, the 3D asset library includes a preset domain scene component library, which contains facility models, object models, and character models modeled based on the features of different discipline subdomains. When searching the 3D asset library, the dominant subject subdomain attribute is identified based on the subject feature vector generated in step S1. The dominant subject subdomain attribute is used to perform tag matching on the domain scene component library to filter out a subset of candidate components, and a search operation is performed on the subset of candidate components.

[0014] As a further improvement to this technical solution, the method also includes steps for the allocation and use of rights based on smart contracts: Deploy smart contracts to monitor other user nodes' access to and interaction with the resource data; When the interactive behavior meets the preset triggering conditions, the smart contract automatically calculates the allocation amount of equity tokens based on the teaching contribution entropy and transfers the equity tokens to the user node's digital wallet. The preset triggering conditions include: other user nodes paying to purchase the resource data, watching more than a preset proportion of video streaming content in its entirety, and generating and uploading new derivative resources based on the resource data; When a request for expansion or style customization of a virtual space node is received, the system API interface is called to verify the amount of equity tokens in the digital wallet. If the verification is successful, the operation request is executed.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. In this intelligent-driven virtual collaboration method for sharing educational resources, the entity mapping and semantic distance calculation of the domain knowledge graph achieve accurate representation of the subdomain attributes of the resource discipline, solve the problem of ambiguous resource classification in existing technologies, and improve the efficiency of resource matching and retrieval. By calculating the teaching contribution entropy through a multi-dimensional weighted algorithm, the method comprehensively considers resource quality, citation depth, interaction activity and academic index to fully evaluate the user's teaching contribution and effectively incentivize the production of high-quality resources.

[0016] 2. In this intelligent-driven virtual collaboration method for sharing educational resources, the appearance of the three-dimensional model of virtual space nodes, scene layout, and system resource configuration dynamically evolve with the subject feature vector and teaching contribution entropy. This ensures the adaptability of virtual scenes and resource attributes (e.g., science and engineering are adapted to a precise style, and humanities are adapted to a discussion style), and also realizes the on-demand allocation of system resources, avoiding resource waste, while meeting the complex collaboration needs of high-contribution users.

[0017] 3. In this intelligent-driven virtual collaboration method for sharing educational resources, the digital fingerprint of the hash algorithm and non-fungible rights certificate realize the accurate authentication of the ownership of educational resources and solve the problem of resource infringement; the smart contract automatically monitors the resource interaction behavior and allocates rights tokens, which simplifies the rights allocation process and protects the legitimate rights and interests of the original users. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the overall process of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] It should be noted that the intelligent-driven virtual collaboration method for sharing educational resources proposed in this invention is universal and can be applied to multiple professional education fields such as law, medicine, history, and engineering. For ease of understanding and explanation, the following embodiments will use "legal education resources" as an example to illustrate the implementation process of this invention in a specific application scenario, but this does not constitute a limitation on the scope of protection of this invention.

[0021] Example 1 With the advancement of digital education, various technological applications have emerged in professional education fields (such as law, medicine, and engineering). However, existing technologies suffer from problems in areas such as inaccurate classification of educational resources, limited user contribution assessment, fixed virtual scenarios, and imperfect rights distribution. Therefore, as Figure 1 As shown, this embodiment discloses an intelligent-driven virtual collaboration method for sharing educational resources, which specifically includes the following steps: Step S1: Obtain educational resource data uploaded by user nodes, establish or call a knowledge graph of the target domain, and determine the center nodes of each subject subdomain in the knowledge graph; map the knowledge entities in the educational resource data to the nodes in the knowledge graph, calculate the semantic distance between the extracted knowledge entities and the preset center nodes of each subject subdomain in the knowledge graph, and generate subject feature vectors representing the tendency of resource content based on the inverse normalization of the semantic distance. In this embodiment, taking the legal education scenario as an example, the specific implementation process is as follows: For acquiring educational resource data, user nodes upload resources via a web interface or client. Resource types include video streaming (such as recorded law courses), document data (such as case analysis reports and compilations of legal provisions), and live data streams (such as virtual court hearings). During upload, the SHA-256 hash algorithm is used to calculate a 128-bit digital fingerprint, which uniquely corresponds to the resource data, preventing tampering. The digital fingerprint, the user node's digital signature (generated using the RSA-2048 algorithm), and resource metadata (such as title and upload time) are written into the consortium blockchain's distributed ledger to generate a non-fungible token (NFT). This token serves as proof of the user's ownership or related rights to the resource. For the construction and invocation of the domain knowledge graph, this example uses a legal knowledge graph. The ontology is built based on the "Directory of Legal Disciplines and Majors" and contains 10 subdomain central nodes (i.e., second-level legal disciplines, such as jurisprudence, constitutional and administrative law, criminal law, civil law, commercial law, intellectual property law, economic law, environmental and resource protection law, international law, and procedural law). Each central node is associated with a corresponding knowledge entity node (such as legal provisions, case names, and legal concepts). The knowledge graph is stored in the Neo4j graph database. If the system has pre-stored the graph, it is directly invoked. If it is not pre-stored or needs to be updated, public data is crawled through web crawlers, and entity recognition and relation extraction are performed using the BERT-BiLSTM-CRF model to supplement the knowledge graph. For knowledge entity mapping and semantic distance calculation, jieba word segmentation is used to segment the resource data into text. Knowledge entities (in this example, legal entities such as "property rights change" and "administrative reconsideration") are extracted using the BERT-BiLSTM-CRF model. The extracted entities are matched with entity nodes in the graph. If a node with a perfect match exists, it is directly associated. If not, the cosine similarity between the entity vector and similar entity nodes in the graph is calculated using Word2Vec. The node with the highest similarity (threshold ≥ 0.85) is taken as the mapping target. Semantic distance is calculated using the shortest path length, which is the minimum number of edges between the knowledge entity mapping node and the subject subdomain center node. For generating subject feature vectors, the reciprocal of the semantic distance corresponding to the center node of each subject subdomain is taken to obtain the distance weight (the shorter the semantic distance, the greater the weight). Then, the distance weights of all subject subdomains are normalized (the normalization formula is: ,in For the first The weight of each second-level discipline, (For semantic distance), generate a 10-dimensional subject feature vector (e.g., [0.05, 0.03, 0.04, 0.72, 0.06, 0.02, 0.03, 0.01, 0.02, 0.01], with civil law having the highest weight). Through the above steps, the digital fingerprint generated by the SHA-256 hash algorithm uniquely corresponds to the resource data. Combined with the NFT notarization of the consortium blockchain distributed ledger, the resource is prevented from being tampered with. As a rights certificate with user digital signature, the NFT solves the problem of difficult definition of resource ownership and easy infringement in the existing technology, and protects the legitimate rights and interests of the original user. Based on the knowledge graph constructed by the second-level discipline of law, the resource's discipline attributes are accurately represented. Semantic distance calculation can quantify the correlation between the resource and each discipline subdomain. The generated discipline feature vector can accurately represent the discipline attributes of the resource, solve the problem of vague classification by relying solely on text labels in the existing technology, and improve the accuracy of resource matching as well as the efficiency of resource retrieval and reuse. Step S2: Collect the interactive behavior data of the educational resource data within a preset time window, combine the content quality score and citation topology of the resource, and use a multi-dimensional weighted algorithm based on content quality, citation depth, derivative of interaction activity and academic index to calculate the teaching contribution entropy of the user node; The specific implementation process is as follows: For data collection, a 30-day time window is preset. Interaction data for resources within this window is collected, including the number of valid visits (filtering bot visits and deduplicating by IP address and device fingerprint), average dwell time (video resources require a viewing time of ≥30 seconds, and document resources require browsing ≥3 pages), and the number of interactive comments (filtering meaningless comments using a TextCNN model for semantic filtering). Simultaneously, the number of times a resource is cited (i.e., in-degree) is obtained through the knowledge graph's citation network, and in-degree centrality is calculated (the formula is: ,in, For the in-degree of resources, (The total number of resource nodes in the knowledge graph). The formula for calculating the entropy of teaching contribution is as follows: ,in, Preset time window Internal teaching contribution entropy; For the first The content quality score for each resource is generated through a collaborative scoring process: a pre-trained visual quality assessment model extracts keyframe features, while a text structure integrity analysis model analyzes metadata for standardization. Specifically, a pre-trained ResNet50 visual quality assessment model extracts keyframes from video resources (1 frame every 10 seconds), calculates features such as frame sharpness and color consistency, and outputs a visual quality score of 0-10. A BERT text structure integrity analysis model analyzes the metadata (title, abstract, keywords, references, and text structure) of document resources, assesses metadata integrity and standardization, and outputs a text quality score of 0-10. The quality score for the live stream is calculated by combining the real-time frame rate (≥30fps), the number of stutters (≤3 times / hour), and content relevance (matching the keywords extracted through speech recognition with the resource's theme). The final Qi is a weighted sum of the visual quality score (0 for non-video resources) and the text quality score (weights of 0.4 and 0.6 respectively), ranging from 0 to 10. The total number of educational resource data uploaded by the user node within the preset time window; The reference depth weight is determined based on the in-degree centrality value of the resource in the knowledge graph reference relationship network. Specifically: based on the in-degree centrality value, a mapping relationship is set: when the in-degree centrality ≥ 0.3, =1.0; when 0.1 ≤ in-degree centrality < 0.3, =0.7; when in-degree centrality <0.1, =0.3, for example, if the in-degree centrality of a case document is 0.25, then =0.7; The interaction activity level is calculated as a weighted sum of the number of valid resource visits, average dwell time, and number of interactive comments within a preset time window. Specifically, it is calculated using a weighted sum, as shown in the formula: ,in, For valid access counts, Average stay duration (in minutes). This refers to the number of interactive comments. For example, if a video resource has 1000 valid visits, an average viewing time of 25 minutes, and 80 interactive comments, then... ; The time derivative of this interaction activity is calculated using linear regression over 30 days. The rate of change, i.e. Where t_end - t_beginning = 30 days; if It is on the rise. If it is positive; if it shows a downward trend, It is a negative value; This is an academic index, calculated by weighting and normalizing the total number of downloads and citations of historically uploaded resources of the user node. Specifically, it is calculated based on the total number of downloads of historically uploaded resources of the user node (…). ) and total citations ( The calculation is as follows: ,in, This represents the maximum total number of downloads by all users on the platform. The maximum total number of references for all users on the platform (maximum value parameter) and The platform data is updated regularly, for example, daily, to reflect the relative ranking of user contributions across the entire platform. The value range is 0-1; The configurable weight coefficients associated with resource types are determined by looking up the values ​​in a pre-defined configuration table based on the media type of the resource. It should be noted that because the first term in the formula is an accumulated value with a large magnitude, while the third term is a normalized value with a smaller magnitude, the coefficients will vary in actual configuration. The value of is usually much smaller than Alternatively, the first term can be normalized during calculation to balance the impact of the numerical magnitudes of each dimension on the final result.

[0022] The preset configuration table for this embodiment is as follows: Video streaming media ( =0.5, =0.3, =0.2); Document data ( =0.6, =0.2, =0.2); Live data stream ( =0.4, =0.4, =0.2), for example, document resources =0.6, =0.2, =0.2; Substitute into the formula For example, a user uploads one document resource. =8.5, =0.7, =8.5 × 0.7 = 5.95; =2.3; =0.65; then ; Through the above steps, the content quality score combines visual and textual evaluation dimensions, avoiding the one-sidedness of single-dimensional evaluation and ensuring the objectivity of resource quality assessment. The citation depth weight is linked to in-degree centrality, reflecting the academic influence of the resource and making up for the deficiency of existing technologies in ignoring the academic value of resources. The derivative of interaction activity reflects the real-time dissemination effect of the resource, and the academic index considers the user's historical contribution. The teaching contribution entropy calculated by multi-dimensional weighting can comprehensively and dynamically evaluate user value, solve the problem of single evaluation dimensions in existing technologies, increase the output of high-quality resources with high academic value and high interactivity, and effectively incentivize users to upload high-quality resources.

[0023] Step S3: Establish a dynamic mapping relationship between virtual space rendering parameters and the subject feature vector and the teaching contribution entropy; based on the dynamic mapping relationship, adjust the 3D model appearance parameters, scene layout scale and underlying system resource configuration of the virtual space node corresponding to the user node in real time, so as to update the rendering status and system resource quota of the virtual space node to the target level configuration corresponding to the current teaching contribution entropy value. The specific implementation process is as follows: Establishing a dynamic mapping relationship between virtual space rendering parameters and the subject feature vectors specifically includes: a preset style material library containing multiple sets of materials corresponding to different subject subdomain attributes; in the legal scenario of this embodiment: The pre-defined style material library, stored in the distributed file system (HDFS), contains 10 sets of style materials corresponding to the second-level disciplines of law. Each set of materials includes textures, lighting models, and architectural component templates. For example, the texture material for civil law is beige fabric, the lighting model is warm-toned diffuse reflection (color temperature 5000K), symbolizing mediation and peace, and the architectural component templates are mediation tables and chairs, and contract display racks; the texture material for criminal law is a metallic texture with red accents, the lighting model is focused lighting (light spot diameter 1.5m), symbolizing majesty and vigilance, and the architectural component templates are the judge's bench and gavel model; the texture material for procedural law is gray stone, the lighting model is cool-toned uniform lighting (color temperature 6500K), and the architectural component templates are the courtroom gallery and evidence display stand. Calculate the weight values ​​of each dimension in the subject feature vector, identify the dominant subject attribute, and if the weight value of a single dimension exceeds a preset threshold, directly call the style material corresponding to the dominant subject attribute for rendering; if there is no single dimension exceeding the preset threshold, perform linear interpolation on the rendering parameters of different style materials based on the weight values ​​of each dimension to generate mixed-style rendering parameters, specifically: The system calculates the weight values ​​of each dimension in the subject feature vector, with a preset threshold of 0.6 for the weight of a single dimension. If the weight value of a certain dimension exceeds 0.6 (e.g., 0.72 for Civil Law), the system directly calls the style material corresponding to that subject for rendering. If there are no dimensions exceeding the threshold (e.g., the weights of each dimension are all between 0.1 and 0.5), the system performs a parameter blending operation, that is, it performs linear interpolation on the rendering parameters of different style materials based on the weights of each dimension. For example, if the subject feature vector is... (Jurisprudence 0.35, Civil Law 0.55), then the interpolation formula for texture material is: Material parameter = 0.35 × Jurisprudence material parameter + 0.55 × Civil Law material parameter, and the color temperature of the lighting model = 0.35 × 6000K + 0.55 × 5000K = 5350K; Scene layout scale adjustment: The scene layout scale is positively correlated with the sum of the weights of the subject feature vectors (the sum after normalization is 1) and the teaching contribution entropy, and a layout scale coefficient is set. ,in, This is the basic guarantee coefficient (with a value of 0.5, used to guarantee the allocation of basic scenario resources). The maximum teaching contribution entropy preset for the platform (value 10). The value ranges from 0.1 to 1.0, corresponding to different scene areas. When S < 0.3, the scene area is 100㎡ (containing only the basic resource display area); when 0.3 ≤ S < 0.6, the scene area is 300㎡ (adding a small discussion area); when S ≥ 0.6, the scene area is 800㎡ (adding a collaborative discussion area). For example, =4.16, =0.5×1.0+0.5×(4.16 / 10)=0.5+0.208=0.708, the scene area is 800㎡; The specific process of adjusting the underlying system resource configuration includes: pre-setting multiple system resource allocation levels, each level corresponding to a threshold range of teaching contribution entropy; real-time monitoring of the teaching contribution entropy of the user node, and triggering an expansion command when the teaching contribution entropy jumps to the threshold range of a higher level; the expansion command controls the container orchestration system to increase the number of container clusters allocated to the user node, increase the bandwidth weight of the content delivery network, and unlock the data loading permissions of the mock court component or large conference component in the virtual space node, specifically: In this embodiment, the preset system resource allocation hierarchy has three levels, and the corresponding threshold ranges for teaching contribution entropy are: Level 1 ( Level 2 Level 3 ); The resource configuration for each level is as follows: Tier 1 (Basic Configuration Status): Number of container clusters 2 (based on Kubernetes orchestration), Content Delivery Network (CDN) bandwidth weight 0.2, only basic resource display components are open; Tier 2 (Advanced Configuration Status): 4 container clusters, CDN bandwidth weight 0.5, open small discussion component (supports ≤10 people online at the same time); Tier 3 (Advanced Configuration): 8 container clusters, CDN bandwidth weight 0.8, unlock data loading permissions for the mock court component (supports ≤50 people online simultaneously) and the large conference component (supports ≤100 people online simultaneously).

[0024] The dynamic adjustment process of resource allocation is as follows: The system uses Prometheus to monitor the teaching contribution entropy of user nodes in real time, for example, The transition from 2.8 to 4.16 (entering tier 2) begins with the system comparing and finding that 4.16 falls within the range [3.0, 6.0), determining the target tier configuration as "tier 2". Then, the system detects that the current configuration is "tier 1", inconsistent with the target "tier 2", and triggers a state update command. Finally, based on the state update command, scaling and loading are performed, specifically including: first, a computing power update, where Kubernetes receives the command and adjusts the number of Pod replicas from 2 to 4; second, a network update, where the CDN scheduling center overwrites the node's bandwidth weight parameter to 0.5; and third, a scene update, where the front-end rendering engine receives the new... The coefficient dynamically loads the AssetBundle of the "small discussion component," expanding the virtual space scene boundary to 300㎡. Through the above steps, the system forcibly synchronizes the visual appearance and service capabilities of the virtual space nodes to the target level configuration corresponding to the current TCE value; at the same time, during the process of performing the above resource configuration adjustment, the system utilizes Kubernetes' rolling update mechanism and session persistence strategy to ensure that the interactive sessions of online users are not interrupted, achieving a smooth upgrade without any noticeable impact. Through the above steps, the dynamic mapping between subject feature vectors and style material libraries enables precise matching between the appearance of the virtual space and the subject attributes of the resources (e.g., criminal law resources corresponding to trial scene styles), enhancing the immersion and relevance of the teaching scenario and solving the problem of the disconnect between existing virtual spaces and resource attributes. The scale of the scene layout is positively correlated with the entropy of teaching contribution, realizing on-demand adaptation of scene size and avoiding low-contribution users occupying redundant scene resources. The hierarchical allocation of system resources allows high-contribution users to obtain more container clusters and bandwidth resources, unlocking complex collaborative components, solving the problem of poor user experience for high-demand users caused by the equal allocation of existing technology resources, while reducing the overall resource waste rate of the system.

[0025] Example 2 Considering the differences in resource update frequency, resource type distribution, and user collaboration needs across different platforms in practical applications—some platforms do not require high-frequency knowledge graph updates, some platforms experience dynamic changes in the proportion of resource types, and some users have higher demands for the smoothness of virtual scene style transitions and the customization of system resources—Example 2, as a variant of Example 1, aims to adapt to more application scenarios through localized technical adjustments, further optimize the technical effect, and enhance the flexibility and practicality of the solution. Specific modifications are as follows: Considering that the knowledge graph construction method in Example 1, which involves crawling data, is suitable for scenarios with frequent resource updates and the need for real-time entity supplementation, but suffers from high data crawling costs and long construction cycles, and that for platforms with low resource update frequencies (such as once every six months) and only basic subject classification is required, repeated data crawling is unnecessary, therefore, in step S1, the knowledge graph construction uses an open-source ontology library (such as LegalKB). After supplementing the subject subdomain center nodes (i.e., secondary subject center nodes), it can be directly called without re-crawling data, which is suitable for scenarios with low resource update frequencies. Through the above steps, the open-source LegalKB ontology library can be directly reused, and it can be called after supplementing the secondary subject center nodes, without the need for additional data crawling and annotation, thus reducing the construction cost of the knowledge graph and shortening the system deployment cycle. At the same time, the LegalKB ontology library has been publicly verified, and the accuracy of legal entities and relationships is higher, ensuring the reliability of subject feature vector generation, and adapting to scenarios with low resource update frequencies and a pursuit of low-cost deployment. Considering that the shortest path length calculation in Example 1 is suitable for scenarios with simple knowledge graph topology and clear entity relationships, but when the number of entity nodes in the knowledge graph is large (≥100,000) and the relationships are complex, the time complexity of the shortest path calculation is high (O(n...). 2 Furthermore, since this method cannot reflect the semantic association strength at the entity vector level (considering only topological structure), in step S1, the semantic distance calculation uses cosine similarity (rather than the shortest path length). Taking the legal profession as an example, the cosine similarity is calculated between the vectors of the legal entity mapping node and the second-level discipline center node. The higher the similarity, the closer the semantic distance, and the greater the weight of the discipline feature vector. For example, the cosine similarity between the "Administrative Review" mapping node and the "Constitutional Law and Administrative Law" center node is 0.92, and the similarity with other center nodes is all <0.3. Therefore, the weight of this dimension in the discipline feature vector is... Through the above steps, cosine similarity directly calculates the vector angle between the mapping node of the legal entity and the center node of the secondary discipline, reducing the time complexity to O(n), thus improving computational efficiency in large-scale knowledge graphs. At the same time, vector representation can capture the semantic connotation of entities (such as the semantic difference between "change of property rights" and "transfer of debt"), making the quantification of semantic distance more accurate and the dimensional weight distribution of discipline feature vectors more reasonable (for example, avoiding the misjudgment of entities with similar topological structures but unrelated semantics as highly related), thereby improving the accuracy of resource matching. Considering that the fixed configuration table in Example 1 is suitable for scenarios where the distribution of platform resource types is stable, but in reality the proportion of platform resource types may change dynamically (e.g., the proportion of video teaching resources increases from 40% to 70% in a semester), the fixed coefficients cannot adapt to this change, causing the teaching contribution entropy assessment to deviate from the actual resource value (e.g., the interactive activity value of video resources is not fully considered). Therefore, in step S2, the weighted coefficients... A dynamic adjustment mechanism is adopted, rather than a fixed configuration table. The coefficients are updated quarterly based on the distribution ratio of platform resource types: if video streaming accounts for ≥60%, then video resources... The coefficient (weight of the derivative of interaction activity) has been increased to 0.4. The coefficient drops to 0.4; if document data accounts for ≥50%, then the document-type resources... The coefficient increased to 0.7. The coefficient is reduced to 0.1. Through the above steps, the coefficient is dynamically adjusted quarterly based on the distribution ratio of resource types, making the evaluation model more closely reflect the actual operation of the platform: when the proportion of video streaming increases, the coefficient is increased. The coefficient (weighted derivative of interaction activity) can more accurately reflect the real-time dissemination value of video resources; when the proportion of document data increases, it improves... The coefficients (weighted by content quality and citation depth) highlight the academic value of document resources; after adjustment, the correlation between teaching contribution entropy and actual resource value is improved, the evaluation bias caused by fixed coefficients is avoided, and the platform scenarios with dynamic changes in resource types are adapted. Considering that the academic index in Example 1 only considers the total number of downloads and citations, and does not cover the teaching application value of resources (such as being included in university teaching syllabi or used as practical training cases), some high-quality resources with strong teaching applicability but low download and citation numbers are underestimated, failing to fully incentivize users to produce resources with strong teaching applicability. Therefore, in step S2, the academic index... The calculation incorporates "the number of times the resource has been adopted" (such as the number of times it has been included in the teaching syllabus of universities), with a weight of 0.2, and the adjusted formula is as follows: ,in, The number of times a resource has been adopted. This is the platform's maximum value. Through the above steps, the "number of times a resource is adopted" directly reflects the teaching application value of the resource. After its inclusion, the evaluation dimensions of the academic index are more comprehensive, and it can accurately identify high-quality "teaching-implementation" resources (such as a case document with a moderate number of downloads, but which has been included in the practical training courses of 3 universities, resulting in an increase of 0.2 in the academic index). At the same time, this dimension incentivizes users to upload resources that are more in line with actual teaching needs, increasing the proportion of teaching-implementation resources on the platform and adapting to scenarios that emphasize teaching applications. Considering that linear interpolation in Example 1 is suitable for scenarios with low requirements for style transition, but when the subject attributes of user resources are mixed (such as involving both civil law and commercial law), the style transition of linear interpolation is relatively abrupt (such as obvious texture splicing marks), affecting the immersiveness of the virtual scene. Therefore, in step S3, the interpolation algorithm for virtual space rendering parameters adopts non-linear interpolation (such as Bézier curve interpolation). Non-linear interpolation is suitable for scenarios that require a smoother style transition. For example, when interpolating style materials of jurisprudence and civil law, the weight allocation is adjusted by Bézier curve to make the transition area smoother. The textures and lighting effects are more natural. Through the above steps, Bézier curve interpolation adjusts the weight distribution through control points, enabling a smooth transition of rendering parameters (texture, lighting, component layout) for different disciplines (such as the natural blending of beige fabric in civil law and metallic texture in commercial law, without obvious splicing marks), thus improving the user's visual experience. At the same time, nonlinear interpolation supports custom control points, which can adjust the transition effect according to the platform's teaching needs (such as favoring a certain discipline style), enhancing the scene customization capability and adapting to collaborative scenarios with high requirements for immersion and customization in virtual scenes (such as high-end legal practice seminars). Considering that the three levels of Implementation Example 1 are suitable for most users, but the collaboration needs of some high-contribution users (such as renowned university professors and senior legal practice experts) are more complex (such as organizing mock courts with more than 100 participants across regions and customizing virtual space building layouts), the resource configuration of the existing levels (such as 8 container clusters and no customized permissions) cannot meet their needs, resulting in a low retention rate for high-contribution users. Therefore, in step S3, the system resource allocation levels are increased to 4, adding level 4 ( ≥8.0), corresponding to a container cluster size of 12, CDN bandwidth weight of 1.0, unlocking customized scene components (supporting user-defined building layouts of virtual space nodes), through the above steps, a new level 4 is added ( With a system version of ≥8.0, it provides more abundant system resources (12 container clusters, CDN bandwidth weight of 1.0) and customized functions (custom building layout), which can meet the complex collaboration needs of high-contribution users (such as low-latency operation of a 100-person mock court and the construction of personalized virtual teaching spaces). At the same time, higher-level rights and incentives further stimulate users' enthusiasm for contribution, improve the retention rate of high-contribution users on the platform, and form a positive cycle of "high contribution - high rights - higher contribution", which is suitable for high-end legal teaching collaboration, large-scale virtual training and other scenarios.

[0026] Example 3 Existing technologies cannot transform abstract legal case files into visual scenes, and poor synchronization during multi-party collaboration affects teaching effectiveness. Therefore, entity extraction and 3D modeling techniques are needed to achieve scene visualization and synchronized collaboration. The specific implementation process is as follows: Scene generation steps based on unstructured scene description text (embedded after step S3): The system obtains the scenario description text data input by the user. In this embodiment, the text data is legal case file text. The process of obtaining the case file text data includes: the user node uploads unstructured legal case file text (such as civil dispute case file, criminal investigation case file). The text format supports docx, pdf, and txt. The system extracts the text content through ApacheTika. Using an entity recognition and relation extraction model, comprehensive information including scene entities, spatial relationships between entities, environmental description information, and action sequence information is extracted from the scene description text data. Specifically, the ERNIE 3.0 entity recognition and relation extraction model is used to extract scene entities (in this example, case entities, such as "suspect" and "weapon"), spatial relationships (in this example, "doorway" in "A injured B at the entrance of A community"), environmental description information (in this example, such as "night" and "rainy day"), and action sequence information (in this example, such as "first attacked, then fled the scene") from the text. For example, from the criminal case file, the entities extracted are: "suspect Zhang", "victim Li", "fruit knife", and "XX Street Park"; spatial relationship: "Zhang attacked Li next to a park bench"; environmental description: "21:00" and "no lighting"; and action sequence: "Zhang took out a fruit knife → stabbed Li → fled the scene". Based on the extracted comprehensive information retrieval 3D asset library, matching 3D model assets are invoked. This 3D asset library includes a domain-specific scene component library. In this example, the component library contains components modeled based on legal discipline characteristics: such as criminal investigation scene facilities (caution tape, investigation tools), evidence models (murder weapon, fingerprint samples), and character models (suspect, victim, police officer) corresponding to criminal law; and civil dispute scene facilities (shops, residences), evidence models (contracts, transfer records), and character models (plaintiff, defendant, mediator) corresponding to civil law. The general 3D model library includes models of natural environments (trees, streetlights), weather effects (rainy days, nighttime), etc. Based on the subject feature vector generated in step S1, the dominant subject subdomain attribute is identified. The dominant subject subdomain attribute is used to perform tag matching on the domain scene component library to filter out a subset of candidate components. A retrieval operation is then performed on the subset of candidate components. For example, if the criminal law weight in the subject feature vector of the case file text is 0.78, then the subset of candidate components corresponding to the "criminal law" tag in the legal scene component library is filtered out. Then, the 3D model matching the entity is retrieved and extracted from the subset (such as "fruit knife" corresponding to the murder weapon model, and "Zhang" corresponding to the criminal suspect role model). The rule engine automatically lays out the 3D model assets into a virtual coordinate system according to the spatial relationship, generating an initialized 3D virtual simulation scene. In this example, the generated scene is a 3D crime scene. Specifically, the rule engine (based on Drools) lays out the 3D model assets into a virtual coordinate system (using a right-handed coordinate system, with the X-axis for the horizontal direction, the Y-axis for the vertical direction, and the Z-axis for the depth direction) according to the spatial relationship. For example, a park bench model is placed at position (10,0,5), a suspect model at position (11,0,5), a victim model at position (10.5,0,5.5), and a fruit knife model at position (9,0,5). Based on the environmental description "21:00" and "no lighting," the nighttime lighting model (low brightness, color temperature 3000K) and the no-lighting effect component are called to generate the initialized 3D crime scene scene. After generating the 3D virtual simulation scene, a collaborative interaction step is also included: for example: The system establishes a multi-party communication channel through the WebRTC protocol, receives connection requests initiated by clients from different geographical locations, and connects the clients to the three-dimensional crime scene scene. Specifically, the system establishes a multi-party communication channel through the WebRTC protocol, and clients from different geographical locations (such as students from university A and teachers from university B) initiate connection requests by inputting scene IDs. After verifying the user's identity (based on NFT rights certificates), the system connects the clients to the three-dimensional crime scene scene. The system collects the displacement, rotation, and scaling operation commands of the current client on the 3D model asset. Using frame synchronization or state synchronization technology, these operation commands are broadcast to other clients, driving their rendering engines to execute the same commands. Specifically: State synchronization technology (suitable for low-latency scenarios) is used. The client collects the user's operation commands on the 3D model asset (e.g., displacement "move the fruit knife model to coordinates (8,0,5)", rotation "rotate the suspect model by 90 degrees", scaling "enlarge the survey tool model by 2 times"). The client sends the model's state (coordinates, angle, size) after the operation to the server. After the server verifies the consistency of the state, it broadcasts it to all connected clients, driving their Unity3D rendering engines to perform the same state update, ensuring that all users see the same scene. If the number of collaborators is ≥50, frame synchronization technology is switched to, with the server broadcasting operation frames at a frequency of 15fps, and the clients synchronously executing the frame commands. Through the above steps, the ERNIE 3.0 model accurately extracts scene description text information. Combined with the automatic layout of the 3D asset library, it transforms abstract text into a visualized 3D virtual simulation scene, solving the problem of abstract cases being difficult to visualize in the teaching of certain subjects and improving students' understanding efficiency of cases. The label matching function of subject feature vectors narrows the 3D model retrieval range, improves retrieval efficiency, and avoids the resource consumption of full-database retrieval. The WebRTC protocol ensures low latency in multi-party communication, and state synchronization / frame synchronization technology ensures the consistency of multi-client operations, solving the problems of poor synchronization and interaction lag in existing collaborative systems and improving the collaborative experience of cross-regional virtual teaching.

[0027] Example 4 Given the lack of an automated resource rights allocation mechanism in existing technologies, user rights cannot be guaranteed in a timely manner. Therefore, it is necessary to use smart contracts to achieve transparency and automation in rights allocation. The specific implementation process is as follows: Steps for the allocation and use of rights based on smart contracts (embedded after step S2): Deploy smart contracts to monitor the access and interaction behavior of other user nodes to the resource data; specifically: deploy smart contracts based on the Ethereum blockchain, with contract code written in Solidity language, and the contract pre-sets the allocation rules, triggering conditions and usage rules of the equity token (named "LawResourceToken", abbreviated as LRT), the smart contract is associated with the distributed ledger of the consortium blockchain, and monitors the access and interaction behavior of other user nodes to the resource data in real time; When the interactive behavior meets the preset triggering conditions, the smart contract automatically calculates the allocation amount of equity tokens based on the teaching contribution entropy and transfers the equity tokens to the user node's digital wallet; specifically: First, the preset trigger conditions include: First, other user nodes purchase resource data (payment amount ≥ 10 RMB, converted at 1 RMB = 1 LRT); Second, the proportion of users who watch the complete video streaming content is ≥ 80% (calculated based on viewing time, 10 minutes = 0.1 LRT); Third, new derivative resources are generated based on this resource data and uploaded (must pass content association detection, association degree ≥ 0.7, calculated based on derivative resources). ×0.5 allocation LRT); Secondly, when the triggering conditions are met, the smart contract automatically calculates the allocation amount. For example, if user A uploads a video resource that user B watches completely (92% viewing rate, 60 minutes duration), then the allocation amount = 60 / 10 × 0.1 = 0.6 LRT; if user C purchases the resource for 50 yuan, the allocation amount = 50 × 1 = 50 LRT; if user D generates derivative resources based on the resource and uploads them (…), the allocation amount is calculated as follows: =3.0), the allocated amount = 3.0 × 0.5 = 1.5 LRT; the smart contract cumulatively allocated 52.1 LRT to user A's digital wallet (the wallet address is bound to the user node's digital signature). When a request for expansion or style customization of a virtual space node is received, the system API interface is invoked to verify the amount of equity tokens in the digital wallet. If the verification is successful, the operation request is executed; specifically: When user A initiates a virtual space node expansion request (e.g., expanding the scene area from 800㎡ to 1200㎡) or a style customization request (e.g., merging two academic styles), the system calls the API interface to verify the number of LRTs in user A's digital wallet. The default requirement is 100 LRTs for expansion and 50 LRTs for style customization. If user A's wallet has ≥100 LRTs, the verification passes, and the system executes the expansion operation, adjusting the scene layout scale coefficient. In version 1.0, the scene area has been expanded to 1200㎡; if the number of LRTs is insufficient, a verification failure message will be returned, and users can obtain LRTs by uploading more high-quality resources; Through the above steps, the smart contract monitors resource interaction behavior in real time and automatically triggers the allocation of rights tokens without manual calculation, solving the problems of low efficiency and error-proneness in rights allocation in existing technologies. The allocation cycle is shortened from the traditional monthly manual calculation to real-time allocation. Multi-dimensional triggering conditions (payment, full viewing, derivative creation) cover the main value conversion scenarios of resources, ensuring the diverse rights of original users. Rights tokens are linked to the unlocking of virtual space node functions, forming a positive cycle of "high-quality resource production - rights acquisition - function upgrade", further incentivizing users to upload high-quality resources and increasing the proportion of high-quality resources on the platform. The rights allocation records stored on the blockchain are tamper-proof, solving the problem of difficulty in obtaining evidence in rights disputes and ensuring the fairness and traceability of rights allocation.

[0028] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A smart-driven virtual collaborative method for sharing educational resources, characterized in that: Includes the following steps: S1. Obtain educational resource data uploaded by user nodes, establish or call a knowledge graph of the target domain, and determine the central nodes of each subject subdomain in the knowledge graph; The knowledge entities in the educational resource data are mapped to nodes in the knowledge graph. The semantic distance between the extracted knowledge entities and the preset central nodes of each subject subdomain in the knowledge graph is calculated. Based on the inverse normalization of the semantic distance, a subject feature vector representing the tendency of resource content is generated. S2. Collect the interactive behavior data of the educational resource data within a preset time window, combine the content quality score and citation topology of the resource, and use a multi-dimensional weighted algorithm based on content quality, citation depth, derivative of interaction activity and academic index to calculate the teaching contribution entropy of the user node. S3. Establish a dynamic mapping relationship between virtual space rendering parameters and the subject feature vector and the teaching contribution entropy; based on the dynamic mapping relationship, adjust the 3D model appearance parameters, scene layout scale and underlying system resource configuration of the virtual space node corresponding to the user node in real time, so as to update the rendering status and system resource quota of the virtual space node to the target level configuration corresponding to the current teaching contribution entropy value.

2. The intelligent-driven virtual collaboration method for sharing educational resources according to claim 1, characterized in that: In step S1, the educational resource data includes video streaming media, document data, and live streaming data streams; When uploading the educational resource data, a hash algorithm is used to generate a digital fingerprint of the educational resource data, and the digital fingerprint is written into a distributed ledger to generate a non-fungible rights certificate. The non-fungible rights certificate records the digital signature of the user node to prove that the user node has ownership or related rights to the educational resource data.

3. The intelligent-driven virtual collaboration method for sharing educational resources according to claim 1, characterized in that, In step S2, the formula for calculating the teaching contribution entropy is as follows: ,in, Preset time window Internal teaching contribution entropy, For the first The content quality score of each resource is generated by a pre-trained visual quality assessment model extracting keyframe features and a text structure integrity analysis model parsing metadata normativity for collaborative scoring. The total number of educational resource data uploaded by the user node within the preset time window; The reference depth weight is determined based on the in-degree centrality value of the resource in the knowledge graph reference relationship network; The interaction activity level is a weighted sum of the number of valid resource visits, average dwell time, and number of interactive comments within a preset time window. This is the time derivative of the interaction activity. As an academic index, it is calculated by weighting and normalizing the total number of downloads and citations of the user node's historical uploaded resources according to preset weights. The configurable weight coefficients associated with resource types are determined by looking up the default configuration table based on the media type to which the resource belongs.

4. The intelligent-driven virtual collaboration method for sharing educational resources according to claim 1, characterized in that, In step S3, establishing a dynamic mapping relationship between virtual space rendering parameters and the subject feature vector specifically includes: A preset style material library, which contains multiple sets of texture materials, lighting models and architectural component templates corresponding to different discipline subdomains; Calculate the weight values ​​of each dimension in the subject feature vector to identify the dominant subject attributes; If the weight value of a single dimension exceeds the preset threshold, the style material corresponding to the dominant discipline attribute will be directly called for rendering. If there is no single dimension exceeding the preset threshold, then the rendering parameters of different style materials are linearly interpolated based on the weight values ​​of each dimension to generate mixed style rendering parameters.

5. The intelligent-driven virtual collaboration method for sharing educational resources according to claim 1, characterized in that, In step S3, the specific process of adjusting the underlying system resource configuration includes: Multiple system resource allocation levels are preset, and each level corresponds to a threshold range of teaching contribution entropy; The teaching contribution entropy of the user node is monitored in real time. When the teaching contribution entropy jumps to a higher threshold range, an expansion command is triggered. The expansion command controls the container orchestration system to increase the number of container clusters allocated to the user node, increase the bandwidth weight of the content delivery network, and unlock the data loading permissions of advanced simulation components or large conference components in the virtual space node.

6. The intelligent-driven virtual collaboration method for sharing educational resources according to claim 1, characterized in that, The method also includes a scene generation step based on unstructured scene description text: Obtain the user-inputted scenario description text data; By using entity recognition and relationship extraction models, comprehensive information such as scene entities, spatial relationships between entities, environmental description information, and action timing information is extracted from the scene description text data. Based on the extracted comprehensive information, a 3D asset library is retrieved, and matching 3D model assets are called up. The rule engine is used to automatically lay out the three-dimensional model assets into the virtual coordinate system according to the spatial position relationship, generating an initialized three-dimensional virtual simulation scene.

7. The intelligent-driven virtual collaboration method for sharing educational resources according to claim 6, characterized in that, After generating the three-dimensional virtual simulation scene, a collaborative interaction step is also included: A multi-party communication channel is established through the WebRTC protocol to receive connection requests initiated by clients from different geographical locations and to connect the clients to the three-dimensional virtual simulation scene; The system collects the displacement, rotation, and scaling operation commands of the current client on the 3D model asset, and broadcasts the operation commands to other clients using frame synchronization or state synchronization technology, driving the rendering engines of other clients to execute the same operation commands.

8. The intelligent-driven virtual collaboration method for sharing educational resources according to claim 6, characterized in that: The 3D asset library includes a pre-defined domain scene component library, which contains facility models, object models, and character models modeled based on the features of different discipline subdomains. When searching the 3D asset library, the dominant subject subdomain attribute is identified based on the subject feature vector generated in step S1. The dominant subject subdomain attribute is used to perform tag matching on the domain scene component library to filter out a subset of candidate components, and a search operation is performed on the subset of candidate components.

9. The intelligent-driven virtual collaboration method for sharing educational resources according to claim 1, characterized in that, The method also includes steps for the allocation and use of rights based on smart contracts: Deploy smart contracts to monitor other user nodes' access to and interaction with the resource data; When the interactive behavior meets the preset triggering conditions, the smart contract automatically calculates the allocation amount of equity tokens based on the teaching contribution entropy and transfers the equity tokens to the user node's digital wallet. The preset triggering conditions include: other user nodes paying to purchase the resource data, watching more than a preset proportion of video streaming content in its entirety, and generating and uploading new derivative resources based on the resource data; When a request for expansion or style customization of a virtual space node is received, the system API interface is called to verify the amount of equity tokens in the digital wallet. If the verification is successful, the operation request is executed.

Images