Software-defined science and education resource dynamic collaborative recommendation method and system based on knowledge graph

Abstract

The invention discloses a software-defined science and education resource dynamic collaborative recommendation method and system based on a knowledge graph. The method comprises the following steps: constructing an artificial intelligence knowledge graph; obtaining science and education resource data of a science and education institution, extracting knowledge entities in the science and education resource data, and connecting the knowledge entities to the artificial intelligence knowledge graph to obtain a science and education resource knowledge graph; constructing a user preference network model containing graph updating and time sequence feature fusion, and extracting local resource features / user features to obtain local resource features and user features of each client; constructing a local knowledge graph based on the science and education resource knowledge graph, and performing model training by taking the local resource features of each client as initial features until the model converges; and inputting the users and the resources into the graph neural network for prediction, and recommending the resources with high scores to each user. According to the invention, the data source and the coverage range of the recommendation system are greatly expanded, and multi-source science and education resource recommendation in a real sense is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of resource recommendation, and in particular to a software-defined scientific and educational resource dynamic collaborative recommendation method and system based on a knowledge graph. Background Art

[0002] In the process of education digitization, a huge amount of scientific and educational resources have been generated. These scientific and educational resources have greatly improved the efficiency of learners' resource acquisition. However, there are problems such as complex resources and difficulty in management. In response to this, a software-defined scientific and educational resource management model has been proposed. Through knowledge graph mining, the skill points and knowledge points contained in scientific and educational resources are included, and the knowledge graph is used as a medium to construct a knowledge semantic network of scientific and educational resources. This overcomes the problem of multi-source massive scientific and educational resource management from a semantic level. However, when dealing with massive scientific and educational resources, the software-defined scientific and educational resource model still faces a serious "knowledge wandering" problem, that is, it is difficult for users to obtain the resources they need from massive, multi-source distributed scientific and educational resources.

[0003] A common solution to the problem of knowledge wandering is to build a recommendation system based on collaborative filtering. However, in a multi-source distribution scenario, different scientific and educational resources and the interactive information between users and resources are stored in different institutions. For the sake of privacy protection and knowledge security, it is difficult to share data between different institutions, which brings difficulties to the recommendation of scientific and educational resources based on collaborative filtering.

[0004] In response to this, recommendation solutions based on federated learning have been proposed. Most of these solutions aggregate at the model level, which makes it difficult to effectively integrate user preference features in different institutions and improve recommendation accuracy. Summary of the invention

[0005] To achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a software-defined science and education resource dynamic collaborative recommendation method, comprising: Based on artificial intelligence knowledge entities and the relationships between entities, an artificial intelligence knowledge graph is constructed; the scientific and educational resource data of scientific and educational institutions are obtained, knowledge entities in the scientific and educational resource data are extracted, and connected to the artificial intelligence knowledge graph to obtain the scientific and educational resource knowledge graph; Build a user preference network model that includes graph updates and temporal feature fusion, extract local resource features / user features, and obtain local resource features and user features of each client; Build a local knowledge graph based on the knowledge graph of science and education resources, and use the local resource features of each client as the initial features for model training until the model converges; Users and resources are input into the graph neural network for prediction, and resources with high scores are recommended to each user.

[0006] Furthermore, based on the AI ​​knowledge entities and the relationships between them, an AI knowledge graph is constructed, including: Build a public knowledge corpus in the field of artificial intelligence, use entity recognition algorithms and relationship extraction algorithms to extract knowledge entities and relationships between entities in the field of artificial intelligence, and form a knowledge graph in the field of artificial intelligence; specifically: Corpus construction: Collect and annotate AI-related texts to build a corpus; Named Entity Recognition: On a given corpus, the BERT language processing model is used to extract text features, and then the model is trained with the help of , and the performance is evaluated on the validation set. The trained model is applied to the complete corpus for named entity recognition; Relationship extraction: extract relevant features, train using a recurrent neural network, evaluate model performance on a validation set, and construct a knowledge graph with the extracted relationships and identified entities.

[0007] Furthermore, the scientific and educational resource data of scientific and educational institutions are obtained, knowledge entities in the scientific and educational resource data are extracted, and connected to the artificial intelligence knowledge graph to obtain the scientific and educational resource knowledge graph, including: For each scientific and educational institution, different types of data, such as metadata, text, video, image, and file, contained in the scientific and educational resources it holds are processed, the knowledge entities and their associations contained therein are extracted, and they are aligned and connected with the constructed artificial intelligence knowledge graph to construct a software-defined scientific and educational resource knowledge graph that integrates artificial intelligence knowledge points and scientific and educational resources; specifically: Feature extraction of multimodal science and education resources: The science and education resources of different modalities including text, video, image, and audio are extracted separately. BERT is used to process text data, and convolutional networks are used to process image and video data. Transformer is used to perform weighted fusion on the processed results to generate a comprehensive semantic feature representation of science and education resources, and the feature representation is used to output the category of each entity. Entity matching from scientific and educational resources to knowledge graph: perform cosine similarity measurement on the learned features and knowledge point features in the knowledge graph, select the closest result for linking, and complete the knowledge link from scientific and educational resources to entities in KG.

[0008] Furthermore, a user preference network model including graph update and temporal feature fusion is constructed to extract local resource features / user features and obtain local resource features and user features of each client, including: On each client, a user preference network is constructed by graph modeling the "user-resource" preference relationship. The cross entropy loss function is constructed using the preference relationship to train the model. When the model converges, the local resource characteristics and user characteristics of each client are obtained.

[0009] Furthermore, a local knowledge graph is constructed based on the knowledge graph of science and education resources, and the local resource features of each client are used as initial features for model training until the model converges, including: Model building: Aiming at the scenario of recommending scientific and educational resources, a recommendation network is built based on graph convolutional neural network; Model training: Based on the constructed model, a loss function is constructed, and data is input into the model for loss calculation and the model parameters are updated through back propagation until the model converges.

[0010] Further, model building includes: Graph update result extraction: the current moment The picture and The picture Compare and delete The graph content at the moment, get the updated graph , and then Processing on Feature Embedding Module: At each layer, for each edge type Assign one Dimensional Embedding , and use both edge type embeddings and node embeddings to compute the attention score:

[0011] in Representation Node and nodes The type of edge between is a learnable matrix used to transform type embeddings; Residual link module: adds pre-activated residual connections for cross-layer node representation, assuming It is the edge The attention weight, is the activation function. Use learnable linear transformations when the feature dimension of a layer changes Transform the features and perform residual aggregation on the transformed features:

[0012] Similarly, the residual module is also introduced for the learning of edge features, as follows:

[0013] Multi-head attention module: introduces a multi-head attention mechanism, using The independent attention modules learn the features separately and concatenate the learned features to obtain the final feature representation. The corresponding update rules are as follows:

[0014]

[0015]

[0016] Final output layer The output of the layer is:

[0017] Time series feature fusion: Before the moment The features learned at each moment are , and then the recurrent neural network is used to process the features and fuse the features containing historical information as follows:

[0018] Preference information prediction: The output features are then input into the prediction module based on the multi-layer perceptron. For each moment, the feature user features learned and resource characteristics , after concatenating the features, input them into the multi-layer perceptron to get the predicted score as follows .

[0019] Further, model training includes: Construct cross entropy: During the training process, the dynamic changes of the simulated graph are simulated to enhance the fit between the learned model and the actual evolution. Specifically, for each moment involved in the training process , respectively construct the corresponding training sets ,for The true labels and corresponding predicted labels contained in , construct the cross entropy loss as follows:

[0020] Model update: Then use the back propagation of the neural network to calculate the gradient and optimize the parameters to evaluate whether the stopping condition is reached. If the model has not converged, continue training until the model converges.

[0021] In a second aspect, the present invention provides a dynamic collaborative recommendation system for scientific and educational resources, comprising: The science and education resource knowledge graph acquisition module is used to construct an artificial intelligence knowledge graph based on artificial intelligence knowledge entities and the relationships between entities; obtain the science and education resource data of scientific and educational institutions, extract the knowledge entities in the science and education resource data, connect them to the artificial intelligence knowledge graph, and obtain the science and education resource knowledge graph; The feature extraction module is used to build a user preference network model that includes graph updates and temporal feature fusion, extract local resource features / user features, and obtain local resource features and user features of each client; The model training module is used to build a local knowledge graph based on the knowledge graph of scientific and educational resources. The local resource features of each client are used as initial features for model training until the model converges. The prediction output module is used to input users and resources into the graph neural network for prediction and recommend high-scoring resources to each user.

[0022] In a third aspect, the present invention provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the software-defined knowledge graph-based dynamic collaborative recommendation method for scientific and educational resources when executing the computer program.

[0023] In a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the software-defined knowledge graph-based dynamic collaborative recommendation method for scientific and educational resources.

[0024] Compared with the prior art, the present invention has the following technical effects: The present invention uses an open knowledge graph as the core. This technical solution cleverly breaks the isolation of scientific and educational resources and user preference information between different educational institutions, laying a solid foundation for building a unified and efficient distributed recommendation model. As a structured knowledge representation method, the knowledge graph not only contains rich domain knowledge, but also provides the relationship between entities, which enables it to become a link connecting the resources of various educational institutions. Through this technology, various institutions can share and use each other's resource information while protecting user privacy, thereby greatly broadening the data source and coverage of the recommendation system, and realizing multi-source scientific and educational resource recommendations in a true sense.

[0025] Furthermore, the solution uses the knowledge graph as a medium for feature propagation, combined with the graph convolutional neural network, to build a powerful feature learning and propagation mechanism. The graph convolutional neural network can deeply mine the information in the graph structure, and pass the feature information of the node to its neighboring nodes through convolution operations, thereby realizing feature aggregation and propagation. In this process, the knowledge graph, as the underlying graph structure, provides rich semantic information for the characteristics of resources and user preference characteristics, making feature learning and propagation more accurate and efficient. This mechanism can not only effectively cope with the challenges brought by user preference heterogeneity and resource heterogeneity, but also improve the accuracy and robustness of the recommendation model through feature aggregation at the semantic level.

[0026] In addition, this technical solution also fully considers the dynamic nature of user preferences and the efficiency requirements in the scenario of recommending massive scientific and educational resources. By introducing dynamic graph neural networks, this solution realizes dynamic adjustment and optimization of the model structure. Dynamic graph neural networks can adjust the feature representation of nodes and edges in real time according to changes in user behavior and resource updates, thereby keeping the model synchronized with user preferences and resource status. This dynamic adjustment mechanism avoids the problem of model failure caused by changes in user preferences or resource updates in traditional recommendation systems, and also greatly improves the response speed and efficiency of the recommendation system. Since there is no need to retrain the model, this solution reduces the system's operation and maintenance costs and time costs while maintaining high efficiency.

[0027] In summary, this technical solution effectively solves the key problems in the recommendation of multi-source scientific and educational resources by utilizing open knowledge graphs, combining graph convolutional neural networks for feature propagation and aggregation, and introducing dynamic graph neural networks to achieve dynamic adjustment of the model. It provides strong technical support for building an efficient, accurate and dynamic recommendation system. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0029] The present invention is further described below in conjunction with the accompanying drawings: Example 1, please refer to Figure 1 The present invention provides a software-defined science and education resource dynamic collaborative recommendation method based on knowledge graph, comprising: Based on artificial intelligence knowledge entities and the relationships between entities, an artificial intelligence knowledge graph is constructed; the scientific and educational resource data of scientific and educational institutions are obtained, knowledge entities in the scientific and educational resource data are extracted, and connected to the artificial intelligence knowledge graph to obtain the scientific and educational resource knowledge graph; Build a user preference network model that includes graph updates and temporal feature fusion, extract local resource features / user features, and obtain local resource features and user features of each client; Build a local knowledge graph based on the knowledge graph of science and education resources, and use the local resource features of each client as the initial features for model training until the model converges; Users and resources are input into the graph neural network for prediction, and resources with high scores are recommended to each user.

[0030] By building an artificial intelligence knowledge graph and integrating knowledge entities from the scientific and educational resource data of scientific and educational institutions, a scientific and educational resource knowledge graph containing rich domain knowledge and resource information is formed. This graph not only provides a comprehensive knowledge background for the recommendation system, but also reveals the potential connections between resources through the relationships between entities, thereby improving the accuracy and relevance of recommendations.

[0031] The dynamic and temporal nature of the user preference network model enhances the personalization of recommendations: The constructed user preference network model not only considers the user's current preferences, but also captures the dynamic changes of user preferences and historical behavior patterns through graph update and time series feature fusion mechanisms. This enables the recommendation system to more accurately understand the user's personalized needs, provide resource recommendations that meet the user's current interests and historical preferences, and enhance the personalization and timeliness of recommendations.

[0032] This method allows each client to train the model based on the local knowledge graph and resource features until the model converges. This distributed training method not only reduces the dependence on the central server, but also improves the efficiency and scalability of model training. At the same time, users and resources are input into the graph neural network for prediction, and high-scoring resources are quickly recommended to each user, ensuring the real-time and responsiveness of the recommendation system.

[0033] As the core component of the recommendation system, the knowledge graph has a powerful semantic understanding capability, which enables the recommendation system to have a deeper understanding of the content of resources and the intention of users. This not only improves the accuracy of recommendations, but also enhances the interpretability of recommendation results and the credibility of users, making it easier for users to accept and trust the recommendations of the recommendation system.

[0034] Embodiment 2, the present invention provides a software-defined science and education resource dynamic collaborative recommendation method based on knowledge graph, specifically comprising: The system targeted by the software-defined science and education resource multi-source collaborative recommendation algorithm based on knowledge graph includes Different scientific and educational institutions For each institution ,Include Users, Scientific and educational resources and the preference relationship between users and scientific and educational resources ; In addition, there is a knowledge graph in the field of artificial intelligence in the system, which contains a knowledge graph composed of knowledge points related to the field of artificial intelligence The main purpose of the software-defined multi-source collaborative recommendation algorithm for scientific and educational resources based on knowledge graph is to learn the characteristics of users and resources from the user preference relationship network for scientific and educational resources maintained by different scientific and educational institutions and integrate them with the help of artificial intelligence knowledge graph, so as to extract more comprehensive user and resource characteristics and achieve more accurate resource recommendation.

[0035]

[0036] 1. Public knowledge graph construction: Build a public knowledge corpus in the field of artificial intelligence, use entity recognition algorithms and relationship extraction algorithms to extract knowledge entities and relationships between entities in the field of artificial intelligence, and form a knowledge graph in the field of artificial intelligence. ; Local science and education resource links: For each science and education institution , process different types of data such as metadata, text, video, image, and file contained in the scientific and educational resources it holds, extract the knowledge entities and their associations, align and connect them with the constructed artificial intelligence knowledge graph, and build a software-defined scientific and educational resource knowledge graph that integrates artificial intelligence knowledge points and scientific and educational resources ; Local resource feature / user feature extraction based on graph neural network: On each client, a user preference network is constructed by graph modeling the preference relationship between “user-resource”. ; Use the preference relationship to build a cross entropy loss function and train the model. When the model converges, the local resource characteristics and user characteristics of each client are obtained; Cross-entity feature propagation based on knowledge graph: each entity Utilize local science and education resource collection and shared AI knowledge graph Build a local knowledge graph , using the scientific and educational resource features learned in step (3) as the initial features, input them into the convolutional neural network for feature learning, each for Each entity in , Learn local features ; Subsequently, all institutions share ; Then, each client uses the attention mechanism to fuse the knowledge point features to obtain comprehensive knowledge point features ; Local feature update: Given Update local resource characteristics and associate the updated resource characteristics with Input the graph neural network to complete the local user feature and resource feature update; Model training: Repeatedly run until the model converges; Recommendation of scientific and educational resources based on federated collaborative filtering: users and resources are input into the graph neural network for prediction, and resources with higher scores are recommended to each user.

[0037] The knowledge graph construction described in 1 mainly includes the following processes: Corpus construction: Collect and annotate AI-related texts to build a corpus.

[0038] Named entity recognition: On a given corpus, the BERT language processing model is used to extract text features, and then the model is trained with the help of , and the performance is evaluated on the validation set, and the parameters are adjusted to improve the effect. The trained model is applied to the complete corpus for named entity recognition.

[0039] Relationship extraction: extract relevant features, train using a recurrent neural network, evaluate model performance on a validation set, and construct a knowledge graph with the extracted relationships and identified entities.

[0040] The scientific and educational resource link shown in 1 mainly includes the following steps: Feature extraction of multimodal science and education resources: Science and education resources of different modalities including text, video, image, audio, etc. are extracted separately, such as using BERT to process text data, using convolutional networks to process image and video data, etc., and using Transformer to perform weighted fusion on the processed results to generate a comprehensive semantic feature representation of science and education resources, and use this feature representation to output the category of each entity.

[0041] Entity matching from scientific and educational resources to knowledge graph: The learned features are measured by cosine similarity with the knowledge point features in the knowledge graph, and the closest results are selected for linking to complete the matching from scientific and educational resources to knowledge graph. Knowledge links of entities in .

[0042] As shown in 1, the resource feature / user feature extraction based on graph neural network includes the following steps: Model building: Aiming at the scenario of recommending scientific and educational resources, a recommendation network is built based on graph convolutional neural network; Model training: Based on the constructed model, a loss function is constructed, and data is input into the model for loss calculation and the model parameters are updated through back propagation until the model converges; Model building mainly includes the following modules: Graph update result extraction: the current moment The picture and The picture Compare and delete The graph content at the moment, get the updated graph , and then Processing on Feature Embedding Module: At each layer, for each edge type Assign one Dimensional Embedding , and use both edge type embeddings and node embeddings to compute the attention score:

[0043] in Representation Node and nodes The type of edge between is a learnable matrix used to transform type embeddings; Residual link module: adds pre-activated residual connections for cross-layer node representation, assuming It is the edge The attention weight, is the activation function. When the feature dimension of the layer changes, the scheme uses a learnable linear transformation Transform the features and perform residual aggregation on the transformed features:

[0044] Similarly, the residual module is also introduced for the learning of edge features, as follows:

[0045] Multi-head attention module: In order to further enhance the learning ability of the model, a multi-head attention mechanism is introduced. The independent attention modules learn the features separately and concatenate the learned features to obtain the final feature representation. The corresponding update rules are as follows:

[0046]

[0047]

[0048] Final output layer The output of the layer is:

[0049] Time series feature fusion: Before the moment The features learned at each moment are , and then the recurrent neural network is used to process the features and fuse the features containing historical information as follows:

[0050] Preference information prediction: The output features are then input into the prediction module based on the multi-layer perceptron. For each moment, the feature user features learned and resource characteristics , after concatenating the features, input them into the multi-layer perceptron to get the predicted score as follows

[0051] Model training mainly includes the following steps: Construct cross entropy: During the training process, the dynamic changes of the simulated graph are simulated to enhance the fit between the learned model and the actual evolution. Specifically, for each moment involved in the training process , respectively construct the corresponding training sets ,for The true labels and corresponding predicted labels contained in , construct the cross entropy loss as follows:

[0052] Model update: Then use the back propagation of the neural network to calculate the gradient and optimize the parameters, evaluate whether the stopping condition is reached, and if the model has not converged, repeat until the model converges.

[0053] The present invention uses an open knowledge graph as a barrier to connect the scientific and educational resources and user preference information stored by different educational institutions, which can effectively realize the learning of distributed recommendation models and achieve privacy-guaranteed multi-source scientific and educational resource recommendations; The present invention uses the knowledge graph as a feature propagation medium and combines it with a graph convolutional neural network to achieve effective propagation of resource features and user preference features between different organizations, achieve feature aggregation at the semantic level, and better resist the performance degradation problem of the recommendation model caused by user preference heterogeneity and resource heterogeneity, thereby effectively improving the model accuracy. The present invention comprehensively considers the dynamic nature of user preferences and the demand for efficiency in the scenario of recommending massive scientific and educational resources, and realizes dynamic adjustment of the model with the help of dynamic graph neural network, so that the dynamically changing preference information of users can be fitted without retraining the model.

[0054] In yet another embodiment of the present invention, a software-defined science and education resource dynamic collaborative recommendation system based on a knowledge graph is provided, which can be used to implement the above-mentioned software-defined science and education resource dynamic collaborative recommendation method based on a knowledge graph. Specifically, the system includes: The science and education resource knowledge graph acquisition module is used to construct an artificial intelligence knowledge graph based on artificial intelligence knowledge entities and the relationships between entities; obtain the science and education resource data of scientific and educational institutions, extract the knowledge entities in the science and education resource data, connect them to the artificial intelligence knowledge graph, and obtain the science and education resource knowledge graph; The feature extraction module is used to build a user preference network model that includes graph updates and temporal feature fusion, extract local resource features / user features, and obtain local resource features and user features of each client; The model training module is used to build a local knowledge graph based on the knowledge graph of scientific and educational resources. The local resource features of each client are used as initial features for model training until the model converges. The prediction output module is used to input users and resources into the graph neural network for prediction and recommend high-scoring resources to each user.

[0055] The division of modules in the embodiments of the present invention is schematic and is only a logical function division. There may be other division methods in actual implementation. In addition, each functional module in each embodiment of the present invention may be integrated into one processor, or may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated modules may be implemented in the form of hardware or in the form of software functional modules.

[0056] In another embodiment of the present invention, a computer device is provided, the computer device including a processor and a memory, the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, which are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions in the computer storage medium to implement the corresponding method flow or corresponding function; the processor described in the embodiment of the present invention can be used for the operation of the software-defined science and education resource dynamic collaborative recommendation method based on the knowledge graph.

[0057] In another embodiment of the present invention, the present invention also provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device for storing programs and data. It is understandable that the computer-readable storage medium here can include both built-in storage media in a computer device and, of course, extended storage media supported by the computer device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by a processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the software-defined science and education resource dynamic collaborative recommendation method based on the knowledge graph in the above embodiment.

[0058] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0059] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0060] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0061] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0062] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A software-defined science and education resource dynamic collaborative recommendation method based on knowledge graph, characterized in that: include: Build an AI knowledge graph based on AI knowledge entities and the relationships between them; Obtain the scientific and educational resource data of scientific and educational institutions, extract the knowledge entities in the scientific and educational resource data, connect them to the artificial intelligence knowledge graph, and obtain the scientific and educational resource knowledge graph; Build a user preference network model that includes graph updates and temporal feature fusion, extract local resource features / user features, and obtain local resource features and user features of each client; Build a local knowledge graph based on the knowledge graph of science and education resources, and use the local resource features of each client as the initial features for model training until the model converges; Users and resources are input into the graph neural network for prediction, and resources with high scores are recommended to each user.

2. The software-defined science and education resource dynamic collaborative recommendation method based on knowledge graph according to claim 1 is characterized in that: Based on AI knowledge entities and the relationships between entities, AI knowledge graphs are constructed, including: Build a public knowledge corpus in the field of artificial intelligence, use entity recognition algorithms and relationship extraction algorithms to extract knowledge entities and relationships between entities in the field of artificial intelligence, and form a knowledge graph in the field of artificial intelligence; specifically: Corpus construction: Collect and annotate AI-related texts to build a corpus; Named Entity Recognition: On a given corpus, the BERT language processing model is used to extract text features, and then the model is trained with the help of , and the performance is evaluated on the validation set. The trained model is applied to the complete corpus for named entity recognition; Relationship extraction: extract relevant features, train using a recurrent neural network, evaluate model performance on a validation set, and construct a knowledge graph with the extracted relationships and identified entities.

3. The software-defined science and education resource dynamic collaborative recommendation method based on knowledge graph according to claim 1 is characterized in that: Obtain the scientific and educational resource data of scientific and educational institutions, extract the knowledge entities in the scientific and educational resource data, connect them to the artificial intelligence knowledge graph, and obtain the scientific and educational resource knowledge graph, including: For each scientific and educational institution, different types of data, such as metadata, text, video, image, and file, contained in the scientific and educational resources it holds are processed, the knowledge entities and their associations contained therein are extracted, and they are aligned and connected with the constructed artificial intelligence knowledge graph to construct a software-defined scientific and educational resource knowledge graph that integrates artificial intelligence knowledge points and scientific and educational resources; specifically: Feature extraction of multimodal science and education resources: The science and education resources of different modalities including text, video, image, and audio are extracted separately. BERT is used to process text data, and convolutional networks are used to process image and video data. Transformer is used to perform weighted fusion on the processed results to generate a comprehensive semantic feature representation of science and education resources, and the feature representation is used to output the category of each entity. Entity matching from scientific and educational resources to knowledge graph: perform cosine similarity measurement on the learned features and knowledge point features in the knowledge graph, select the closest result for linking, and complete the knowledge link from scientific and educational resources to entities in KG.

4. The software-defined science and education resource dynamic collaborative recommendation method based on knowledge graph according to claim 1 is characterized in that: Build a user preference network model that includes graph updates and time series feature fusion, extract local resource features / user features, and obtain local resource features and user features of each client, including: On each client, a user preference network is constructed by graph modeling the "user-resource" preference relationship. The cross entropy loss function is constructed using the preference relationship to train the model. When the model converges, the local resource characteristics and user characteristics of each client are obtained.

5. The software-defined science and education resource dynamic collaborative recommendation method based on knowledge graph according to claim 4 is characterized in that: Based on the knowledge graph of science and education resources, a local knowledge graph is constructed. The local resource features of each client are used as initial features for model training until the model converges, including: Model building: Aiming at the scenario of recommending scientific and educational resources, a recommendation network is built based on graph convolutional neural network; Model training: Based on the constructed model, a loss function is constructed, and data is input into the model for loss calculation and the model parameters are updated through back propagation until the model converges.

6. The software-defined science and education resource dynamic collaborative recommendation method based on knowledge graph according to claim 5 is characterized in that: Model building, including: Graph update result extraction: the current moment The picture and The picture Compare and delete The graph content at the moment, get the updated graph , and then Processing on Feature Embedding Module: At each layer, for each edge type Assign one Dimensional Embedding , and use both edge type embeddings and node embeddings to compute the attention score: in Representation Node and nodes The type of edge between is a learnable matrix used to transform type embeddings; Residual link module: adds pre-activated residual connections for cross-layer node representation, assuming It is the edge The attention weight, is the activation function. When the feature dimension of a layer changes, use a learnable linear transformation Transform the features and perform residual aggregation on the transformed features: Similarly, the residual module is also introduced for the learning of edge features, as follows: Multi-head attention module: introduces a multi-head attention mechanism, using The independent attention modules learn the features separately and concatenate the learned features to obtain the final feature representation. The corresponding update rules are as follows: Final output layer The output of the layer is: Time series feature fusion: Before the moment The features learned at each moment are , and then the recurrent neural network is used to process the features and fuse the features containing historical information as follows: Preference information prediction: The output features are then input into the prediction module based on the multi-layer perceptron. For each moment, the feature user features learned and resource characteristics , after concatenating the features, input them into the multi-layer perceptron to get the predicted score as follows 。 7. The software-defined science and education resource dynamic collaborative recommendation method based on knowledge graph according to claim 5 is characterized in that: Model training, including: Construct cross entropy: During the training process, the dynamic changes of the simulated graph are simulated to enhance the fit between the learned model and the actual evolution. Specifically, for each moment involved in the training process , respectively construct the corresponding training sets ,for The true labels and corresponding predicted labels contained in , construct the cross entropy loss as follows: Model update: Then use the back propagation of the neural network to calculate the gradient and optimize the parameters to evaluate whether the stopping condition is reached. If the model has not converged, continue training until the model converges.

8. A software-defined science and education resource dynamic collaborative recommendation system based on knowledge graph, characterized by: include: The science and education resource knowledge graph acquisition module is used to build an artificial intelligence knowledge graph based on artificial intelligence knowledge entities and the relationships between entities; Obtain the scientific and educational resource data of scientific and educational institutions, extract the knowledge entities in the scientific and educational resource data, connect them to the artificial intelligence knowledge graph, and obtain the scientific and educational resource knowledge graph; The feature extraction module is used to build a user preference network model that includes graph updates and temporal feature fusion, extract local resource features / user features, and obtain local resource features and user features of each client; The model training module is used to build a local knowledge graph based on the knowledge graph of scientific and educational resources. The local resource features of each client are used as initial features for model training until the model converges. The prediction output module is used to input users and resources into the graph neural network for prediction and recommend high-scoring resources to each user.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the software-defined dynamic collaborative recommendation method for scientific and educational resources based on the knowledge graph are implemented as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the software-defined knowledge graph-based dynamic collaborative recommendation method for science and education resources are implemented as described in any one of claims 1 to 7.

Images