Intelligent ideological and political education resource matching method and system based on knowledge graph

Abstract

The invention relates to an intelligent ideological and political education resource matching method and system based on a knowledge graph. The method comprises the following steps: acquiring an ideological and political education and teaching material data set, and extracting knowledge concepts to obtain an ideological and political education and teaching material-knowledge concept association relationship set; in response to the obtained mark of the learner on the knowledge concept corresponding to the ideological and political education and teaching material, establishing a learner personalized knowledge graph according to the ideological and political education and teaching material-knowledge concept association relationship set; constructing a learner knowledge gap model according to the personalized knowledge graph; based on an ideological and political education resource library, performing ideological and political education resource matching according to the learner knowledge gap model to obtain an intelligent recommendation result; the ideological and political education resource library comprises a plurality of resource vectors. By adopting the method, the knowledge graph, the learner knowledge gap modeling and the intelligent recommendation technology can be fused, educational resources and auxiliary teaching support are realized for ideological and political education, and a traditional ideological and political teaching mode is promoted to be upgraded to a personalized and intelligent direction.

Description

Technical Field

[0001] This invention belongs to the field of teaching signal technology, and in particular relates to an intelligent matching method and system for ideological and political education resources based on knowledge graphs. Background Technology

[0002] With the development of educational information technology, intelligent education technology based on knowledge graphs has emerged. Through the association structure between complex concepts, it can perform intelligent recommendation, intelligent question answering, learning path planning, etc., and has the advantages of clear structure, flexible expression and support for reasoning.

[0003] Existing knowledge graph-based teaching resources typically establish a mapping with pre-defined knowledge points through keyword extraction, entity linking, and other methods. Then, based on the course content or the goals set by the teacher, resources are retrieved and matched for recommendation.

[0004] However, the knowledge graphs described above are typically static structures built uniformly for courses or textbooks, failing to dynamically adjust according to individual learners' cognitive levels or knowledge gaps. Recommended resources are based solely on semantic matching of content and knowledge points, rather than on supplementary directions truly needed by learners, resulting in insufficient personalization. There is a lack of mechanisms to collect data from learners' actual learning process to update the graph structure or recommendation strategies, failing to form a closed-loop knowledge modeling and resource recommendation optimization mechanism. Operating at the knowledge point-to-resource level ignores the more detailed structural layers within the teaching content, leading to a disconnect between recommended resources and current learning content, impacting learning efficiency and coherence. Summary of the Invention

[0005] Therefore, it is necessary to provide a knowledge graph-based intelligent matching method and system for ideological and political education resources that can provide personalized recommendations for learners, addressing the aforementioned technical issues.

[0006] Firstly, this application provides an intelligent matching method for ideological and political education resources based on knowledge graphs, including:

[0007] A dataset of ideological and political education teaching materials was obtained, and knowledge concepts were extracted to obtain a set of relationships between ideological and political education teaching materials and knowledge concepts. The set of relationships between ideological and political education teaching materials and knowledge concepts includes multiple ideological and political education teaching materials and their corresponding knowledge concepts.

[0008] In response to the learners' markings of knowledge concepts corresponding to ideological and political education teaching materials, a personalized knowledge graph for learners is established based on the set of relationships between ideological and political education teaching materials and knowledge concepts. The markings include understanding markings and non-understanding markings. The personalized knowledge graph includes nodes and corresponding relationship edges. The nodes include ideological and political education teaching material nodes, knowledge concept nodes, and learner nodes. The relationship edges correspond to the relationships of read, included, understood, and not understood.

[0009] Construct learner knowledge gap models based on personalized knowledge graphs;

[0010] Based on the ideological and political education resource database, the system matches ideological and political education resources according to the learner's knowledge gap model to obtain intelligent recommendation results; the ideological and political education resource database includes multiple resource vectors.

[0011] In one embodiment, a dataset of ideological and political education teaching materials is obtained, and knowledge concepts are extracted to obtain a set of relationships between ideological and political education teaching materials and knowledge concepts, including:

[0012] Keyword extraction algorithms were used to extract key phrases from each ideological and political education teaching material in the dataset, resulting in multiple candidate keyword phrases.

[0013] A Chinese entity linking algorithm is used to disambiguate multiple candidate keyword phrases to obtain knowledge concepts.

[0014] By associating knowledge concepts with corresponding ideological and political education teaching materials, a set of association relationships between ideological and political education teaching materials and knowledge concepts is obtained.

[0015] In one embodiment, the learner-personalized knowledge graph is obtained through the following method:

[0016] Based on the corresponding ideological and political education teaching materials, construct the read relationship edges between the ideological and political education teaching material nodes and the learner nodes;

[0017] Based on the set of relationships between ideological and political education teaching materials and knowledge concepts, establish inclusion relationship edges between nodes of ideological and political education teaching materials and nodes of knowledge concepts;

[0018] Based on the understanding tags, establish understanding relationship edges between knowledge concept nodes and learner nodes;

[0019] Based on the misunderstanding markers, establish misunderstanding relationship edges between knowledge concept nodes and learner nodes;

[0020] By integrating nodes of ideological and political education teaching materials, learner nodes, and knowledge concept nodes, as well as their corresponding relational edges, a personalized knowledge graph for learners is obtained.

[0021] In one embodiment, constructing a learner knowledge gap model based on a personalized knowledge graph includes:

[0022] Based on the pre-trained Chinese SEBRT model, the concept name of each knowledge concept node is encoded to obtain the initial concept semantic embedding vector;

[0023] Based on the pre-trained Chinese SEBRT model, the text content corresponding to each ideological and political education teaching material node is extracted to obtain the initial ideological and political education teaching material embedding vector.

[0024] Based on graph convolutional networks, multiple rounds of information aggregation are performed on each node of the personalized knowledge graph according to the cosine similarity of SEBRT vectors to obtain the final embedding vector of each node; the SEBRT vector consists of the initial concept semantic embedding vector and the initial ideological and political education teaching material embedding vector.

[0025] Extracting relational edges from the personalized knowledge graph to identify nodes representing concepts that learners do not understand, thus obtaining nodes representing concepts that learners do not understand.

[0026] The cosine similarity between the initial semantic embedding vector of the concept node that the learner does not understand and the initial embedding vector of the ideological and political education teaching materials is used to obtain the knowledge gap weight.

[0027] Based on weighted averaging, a learner knowledge gap model is synthesized by combining the initial conceptual semantic embedding vector corresponding to the concept nodes that learners do not understand with the knowledge gap weights.

[0028] In one embodiment, based on the ideological and political education resource database, ideological and political education resources are matched according to the learner's knowledge gap model to obtain intelligent recommendation results, including:

[0029] Based on learners' lack of understanding of the knowledge concepts corresponding to concept nodes, keyword pre-screening is performed on resource vectors in the ideological and political education resource database to obtain multiple candidate resource vectors.

[0030] Based on the final embedding vector of each node, the cosine similarity between the learner's knowledge gap model and multiple candidate resource vectors is calculated.

[0031] The resource vectors are sorted in descending order of cosine similarity to obtain the recommendation list;

[0032] Based on a preset recommendation quantity threshold, the educational resources corresponding to the recommendation list are extracted to obtain intelligent recommendation results.

[0033] In one embodiment, the ideological and political education resource database is constructed in the following manner:

[0034] Access to external educational resources;

[0035] Metadata is entered into multiple educational resources contained in external educational resources to obtain a metadataset; the metadataset includes multiple educational resources and their corresponding metadata; the metadata includes title, summary, keywords and category labels;

[0036] The metadata of the metadata dataset is encoded based on a pre-trained Chinese SEBRT model to obtain multiple resource vectors;

[0037] A resource database for ideological and political education is constructed based on multiple resource vectors.

[0038] In one embodiment, the method further includes:

[0039] Obtain learners' feedback data on the educational resources corresponding to the intelligent recommendation results; feedback data includes favorites, read marks, and irrelevant feedback;

[0040] The knowledge gap weights are adjusted based on the feedback data, and the recommendation list is reordered to obtain new intelligent recommendation results.

[0041] Secondly, this application also provides an intelligent matching system for ideological and political education resources based on knowledge graphs, including:

[0042] The learning resources module is used to acquire a dataset of ideological and political education teaching materials and extract knowledge concepts to obtain a set of relationships between ideological and political education teaching materials and knowledge concepts. The set of relationships between ideological and political education teaching materials and knowledge concepts includes multiple ideological and political education teaching materials and their corresponding knowledge concepts.

[0043] The learner-personalized knowledge graph module is used to respond to the learner's marking of knowledge concepts corresponding to ideological and political education teaching materials, and to build a learner-personalized knowledge graph based on the set of associations between ideological and political education teaching materials and knowledge concepts.

[0044] The knowledge gap module is used to build learner knowledge gap models based on personalized knowledge graphs.

[0045] The resource matching module is used to match ideological and political education resources based on the ideological and political education resource database and learners' knowledge gap model to obtain intelligent recommendation results.

[0046] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of any of the above-mentioned knowledge graph-based intelligent matching methods for ideological and political education resources.

[0047] Fourthly, this application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of any of the above-mentioned knowledge graph-based intelligent matching methods for ideological and political education resources.

[0048] The aforementioned knowledge graph-based intelligent matching method and system for ideological and political education resources utilizes knowledge graphs to construct personalized knowledge structure models for learners. Combined with learning status markers, this enables differentiated instruction and on-demand recommendations, significantly improving the matching degree between ideological and political education resources and learners' actual needs. It organically connects teaching materials, knowledge concepts, and learners' understanding states through structured graphs, forming a knowledge network with clear semantic levels and logical relationships, facilitating in-depth management and intelligent mining of teaching resources. By combining the vectorized representation of teaching resources with a knowledge gap model matching mechanism, it achieves efficient resource retrieval and precise delivery, effectively improving resource utilization and reducing the problem of duplicate content recommendations. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is a flowchart illustrating the intelligent matching method for ideological and political education resources based on knowledge graphs according to the present invention.

[0051] Figure 2 This is a flowchart illustrating the steps of step S103.

[0052] Figure 3 This is a flowchart illustrating the steps of step S104.

[0053] Figure 4 This is a structural diagram of the knowledge graph-based intelligent matching system for ideological and political education resources of the present invention. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0055] In one embodiment, such as Figure 1 As shown, a knowledge graph-based intelligent matching method for ideological and political education resources is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, and further to a system including both a terminal and a server, and is implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0056] S101. Obtain the dataset of ideological and political education teaching materials and extract knowledge concepts to obtain the set of relationships between ideological and political education teaching materials and knowledge concepts; the set of relationships between ideological and political education teaching materials and knowledge concepts includes multiple ideological and political education teaching materials and their corresponding knowledge concepts.

[0057] In the illustrative sense, the ideological and political education teaching materials dataset can include various forms of teaching materials, such as PowerPoint presentations (PPT), course handouts, or teaching manuscripts. Knowledge concepts refer to keywords or terminological phrases that can express core ideas, political theories, historical context, or policy terms. The knowledge concept extraction process can employ keyword extraction and entity recognition methods from natural language processing technologies. For example, this can involve unsupervised keyword extraction based on TextRank (a graph-based ranking algorithm), candidate phrase recognition combining part-of-speech and word frequency, and calling a Chinese encyclopedia-like entity linking system to achieve concept standardization. Furthermore, each piece of ideological and political education teaching material is associated with multiple knowledge concepts, thus forming a set of associations between ideological and political education teaching materials and their corresponding knowledge concepts.

[0058] S102. In response to the learner's marking of the knowledge concepts corresponding to the ideological and political education teaching materials, a personalized knowledge graph for the learner is established based on the set of associations between ideological and political education teaching materials and knowledge concepts. The markings include understanding markings and non-understanding markings. The personalized knowledge graph includes nodes and corresponding relation edges. The nodes include ideological and political education teaching material nodes, knowledge concept nodes, and learner nodes. The relation edges correspond to the relationships of read, included, understood, and not understood.

[0059] This method illustratively records learners' behaviors in studying ideological and political education materials and collects their subjective cognitive states regarding each knowledge concept based on interactive design. Specifically, interactive marking functions can be set below each slide or section of handouts. For example, "Understand" and "Don't Understand" buttons can be provided, allowing learners to mark the knowledge concepts automatically extracted from the page. Responding to learners' marking actions, a learner-centered personalized knowledge graph is constructed. This personalized knowledge graph consists of three types of nodes: teaching material nodes, knowledge concept nodes, and learner nodes; and several relational edges, including inclusion relationships indicating that the ideological and political education materials contain certain knowledge concepts, read relationships indicating learners' reading status of the materials, understanding relationships indicating learners' understanding of certain knowledge concepts, and non-understanding relationships indicating learners' non-understanding of certain knowledge concepts. Through the combination of these nodes and edges, it is possible to represent which knowledge points learners encounter in specific materials, and which knowledge concepts they understand or do not understand, thus forming a dynamically updated structured cognitive graph.

[0060] S103. Construct a learner knowledge gap model based on personalized knowledge graphs.

[0061] This paper illustrates the construction of a learner knowledge gap model based on a personalized knowledge graph to characterize learners' current knowledge comprehension blind spots. The model considers both the semantic features of the knowledge concepts themselves and their structural context within the graph. Specifically, each learner-uncomprehended knowledge concept node can be vectorized, obtained by encoding its definition or related description using a pre-trained language model. Further, the adjacency structure of these nodes within the personalized knowledge graph is analyzed, including which ideological and political education materials they connect to and whether they are associated with multiple other uncomprehended knowledge concept nodes. By calculating the aggregated representation of the set of currently uncomprehended knowledge concepts, the blank areas or interest clusters in the learner's current knowledge state can be identified, thus yielding the learner knowledge gap model.

[0062] S104. Based on the ideological and political education resource database, ideological and political education resources are matched according to the learner's knowledge gap model to obtain intelligent recommendation results; the ideological and political education resource database includes multiple resource vectors.

[0063] This example illustrates how a resource matching and recommendation task is performed based on a pre-established ideological and political education resource database, addressing learners' knowledge gaps. This database can contain various types of supplementary learning materials, including policy news interpretations, excerpts of authoritative speeches, videos of ideological and political case studies, and illustrated popular science articles. For instance, each supplementary learning resource can have its semantic vector constructed using meta-information such as title, abstract, and tags upon entry into the database, and this vector is uniformly encoded and stored within the database. Furthermore, by comparing the semantic similarity between the learner's knowledge gap model vector and each resource vector, cosine similarity or other distance metrics can be used for scoring and ranking, thereby selecting the resources most relevant to the learner's current cognitive blind spots. Optionally, the intelligent recommendation results can not only provide resource titles and abstracts but also display the core concepts associated with the resource or suggest their correspondence with unfamiliar knowledge concepts, enhancing the interpretability of the recommendations.

[0064] The aforementioned knowledge graph-based intelligent matching method for ideological and political education resources, by constructing a personalized knowledge graph for each learner, can accurately identify each learner's understanding of specific knowledge concepts, enabling personalized recommendations of teaching resources and improving the accuracy of resource matching. Utilizing a learner knowledge gap model, it can precisely locate weaknesses in learners' knowledge structures, improving the effectiveness of ideological and political education resource matching. By modeling the relationship between teaching materials and knowledge concepts, a graph-based mapping between teaching content and knowledge points is established, facilitating the systematic management and semantic-level linkage of subsequent educational resources and improving the intelligence of resource scheduling. Based on the matching mechanism between resource vectors and the knowledge gap model, it can quickly filter out the most valuable content for the current learner from massive ideological and political resources, improving resource distribution efficiency and reducing resource redundancy.

[0065] In one embodiment, a dataset of ideological and political education teaching materials is obtained, and knowledge concepts are extracted to obtain a set of relationships between ideological and political education teaching materials and knowledge concepts, including:

[0066] S11. Use a keyword extraction algorithm to extract key phrases from each ideological and political education teaching material in the ideological and political education teaching material dataset to obtain multiple candidate keyword phrases.

[0067] This example illustrates how a keyphrase extraction algorithm is used to extract key phrases from each document in the ideological and political education teaching materials dataset. The keyphrase extraction algorithm can employ graph-based algorithms such as TextRank, SingleRank (a graph-based keyphrase extraction algorithm), statistical weight-based TF-IDF (Term Frequency-Inverse Document Frequency), or attention-based algorithms using pre-trained language models. Specifically, a set of phrases potentially possessing pedagogical semantic value is initially identified, forming a candidate keyphrase pool for subsequent semantic standardization processing.

[0068] S12. Use the Chinese entity linking algorithm to perform entity disambiguation on multiple candidate keyword phrases to obtain knowledge concepts.

[0069] This example illustrates how entity linking is performed on a set of candidate keyword phrases to achieve semantic uniformity and ambiguity resolution. Entity linking refers to mapping phrases appearing in natural language to standardized conceptual entities in a knowledge base or semantic system. Entity linking algorithms can employ open-source entity linking tools for Chinese, including linking systems based on CN-DBpedia Spotlight (an entity linking tool), or a combination algorithm using HanLP (a natural language processing toolkit) and a self-built terminology dictionary for the ideological and political education field. Specifically, the candidate keyword phrases undergo semantic discrimination, their true meaning is determined within the context, and the most matching conceptual entries are retrieved from the existing educational knowledge base, thereby ensuring semantic normalization and distinguishing potentially ambiguous phrases.

[0070] S13. Associate knowledge concepts with corresponding ideological and political education teaching materials to obtain a set of association relationships between ideological and political education teaching materials and knowledge concepts.

[0071] This example illustrates how a structured association is established between standard knowledge concepts, after entity disambiguation, and their source ideological and political education teaching materials, forming a set of association relationships between ideological and political education teaching materials and knowledge concepts. For instance, this set of relationships can be expressed in the form of triples, i.e., the format (teaching material identifier, containment, knowledge concept), clearly representing the semantic membership relationship between teaching materials and knowledge concepts. Optionally, the triples can be stored as edge information in a graph database, or standardized and stored using structured formats such as JSON-LD or RDF.

[0072] The above method realizes the entire process of automatically extracting and structurally modeling knowledge concepts from original teaching materials, ensuring the traceability, computability, and recommendability of ideological and political education content at the semantic level.

[0073] In one embodiment, the learner-personalized knowledge graph is obtained through the following method:

[0074] S21. Based on the ideological and political education teaching materials corresponding to the tags, construct the read relationship edges between the ideological and political education teaching material nodes and the learner nodes.

[0075] In a schematic way, read-ahead edges are established based on learners' reading behavior of teaching materials. Specifically, whenever a learner browses a slide, opens a handout, or clicks on a teaching material item, the operation is automatically recorded, and a read-ahead edge is established in the graph, pointing from the learner node to the corresponding teaching material node. The existence of this edge indicates the learner's actual contact with a certain material.

[0076] S22. Based on the set of relationships between ideological and political education teaching materials and knowledge concepts, establish inclusion relationship edges between nodes of ideological and political education teaching materials and nodes of knowledge concepts.

[0077] By utilizing the set of relationships between ideological and political education teaching materials and knowledge concepts, an inclusion relationship edge is established between each node of ideological and political education teaching materials and the knowledge concept nodes involved. The inclusion edge indicates which core ideological and political concepts the material involves.

[0078] S23. Based on the understanding tags, establish the understanding relationship edges between knowledge concept nodes and learner nodes.

[0079] Based on learners' understanding labels, understanding relationship edges are constructed from learner nodes to knowledge concept nodes. During the learning process, learners can subjectively label each knowledge concept. If a learner believes they have mastered a knowledge concept, they record the probability of that knowledge as an understanding label, and generate corresponding relationship edges in the graph. This reflects the knowledge area that the learner believes they have mastered, helping to eliminate unnecessary resource recommendations related to the probability of not understanding the knowledge.

[0080] S24. Based on the misunderstanding markers, establish misunderstanding relationship edges between knowledge concept nodes and learner nodes.

[0081] The learner generates "ununderstood" relationship edges based on their "ununderstood" markers. If a knowledge concept is marked as "ununderstood," it indicates that the concept belongs to the learner's current knowledge blind spot and is a target concept that needs to be focused on during resource matching.

[0082] S25. Integrate the nodes of ideological and political education teaching materials, learner nodes, and knowledge concept nodes, as well as their corresponding relational edges, to obtain a personalized knowledge graph for learners.

[0083] All constructed nodes and edges are uniformly integrated to form a learner's personalized knowledge graph. This graph structure not only presents the behavioral connections between learners and teaching content, but also embeds knowledge semantics and its level of understanding, laying a complete structural foundation for subsequent graph convolutional network modeling, semantic aggregation, and resource recommendation.

[0084] In one embodiment, such as Figure 2 As shown, a learner knowledge gap model is constructed based on a personalized knowledge graph, including:

[0085] S201. Based on the pre-trained Chinese SEBRT model, the concept name of each knowledge concept node is encoded to obtain the initial concept semantic embedding vector.

[0086] This example illustrates how a pre-trained Chinese semantic encoding model (SBERT, Sentence-BERT) is used to vectorize each knowledge concept node in the graph. SBERT is a pre-trained model that optimizes sentence vectors on top of BERT (Bidirectional Encoder Representations from Transformers), enabling it to map short texts or terms into high-dimensional semantic embedding vectors. For instance, publicly available Chinese SBERT models such as shibing624-text2vec-base-chinese can be used, taking the text of each knowledge concept node as input and outputting its corresponding semantic representation vector.

[0087] S202. Based on the pre-trained Chinese SEBRT model, the text content corresponding to each ideological and political education teaching material node is extracted to obtain the initial ideological and political education teaching material embedding vector.

[0088] This example illustrates how the same SBERT vectorization operation is performed on each node of ideological and political education teaching materials in the personalized knowledge graph. The original text of these materials can be slide content, textbook paragraphs, or teacher handouts. After appropriate text preprocessing, such as removing whitespace and concatenating multiple text segments, it is input into the same SBERT model to generate embedding vectors representing the semantic content of the material. Since teaching materials are typically long, the SBERT encoding results can effectively capture their overall ideological content, which is helpful for subsequently constructing high-level semantic relationships between ideological and political education teaching materials and knowledge concepts.

[0089] S203. Based on graph convolutional networks, multiple rounds of information aggregation are performed on each node of the personalized knowledge graph according to the cosine similarity of the SEBRT vectors to obtain the final embedding vector of each node; the SEBRT vector consists of the initial concept semantic embedding vector and the initial ideological and political education teaching material embedding vector.

[0090] Graph Convolutional Networks (GCNs) are used to aggregate structural information across the entire personalized knowledge graph in multiple rounds. For example, to achieve lightweight and efficient modeling, graph convolutional network frameworks such as LightGCN (Lightweight Graph Convolutional Network) are preferred. Specifically, during the aggregation process, the embedding vector of each node is weighted and integrated with the information of its neighboring nodes, i.e. in, For neighboring nodes; The final embedding vector is represented by l; the number of iterations for aggregation is l; and the weights are w. vuThe semantic cosine similarity between the initial SBERT vectors of nodes is dynamically calculated, thus preserving semantic features while injecting the graph structural context. For example, through two rounds of iteration, each node finally obtains a final embedding vector that integrates the structural information of its neighbors and its own semantic features.

[0091] S204. Extract relational edges from the personalized knowledge graph to identify nodes representing concepts that the learner does not understand, thus obtaining the nodes representing concepts that the learner does not understand.

[0092] This is illustrative. All knowledge concept nodes that have unresolved relationships with learner nodes are selected from the personalized knowledge graph. This set represents the key knowledge units that the learner explicitly indicates they have not mastered in actual learning.

[0093] S205. Calculate the cosine similarity between the initial concept semantic embedding vector and the initial ideological and political education teaching material embedding vector corresponding to the concept node that the learner does not understand, and obtain the knowledge gap weight.

[0094] Furthermore, the cosine similarity between the initial conceptual semantic embedding vector of each node whose concept is not understood and the initial ideological and political education teaching material embedding vector of its corresponding node is calculated. This cosine similarity serves as the knowledge gap weight for that concept in the current learning context. This cosine similarity reflects the severity of the learner's comprehension difficulties when encountering the concept in specific learning materials. A higher similarity indicates that the concept is more crucial in the current learning content, and the consequences of the learner's failure to grasp it may be more severe. Therefore, this weight can be used to represent recommendation priority.

[0095] S206. Based on weighted average, synthesize the learner knowledge gap model according to the initial concept semantic embedding vector corresponding to the concept node that the learner does not understand and the knowledge gap weight.

[0096] Indicatively, a weighted average method is used to synthesize a unified vector representing the learner's current knowledge gap model. Specifically, the embedding vectors of all ununderstood concept nodes are weighted, summed, and normalized according to their knowledge gap weights to obtain a high-dimensional vector expression, which serves as the overall semantic representation of the learner's current "unmastered knowledge domains." in, α is the initial conceptual semantic embedding vector; i Weights for knowledge gaps.

[0097] In one embodiment, such as Figure 3 As shown, based on the ideological and political education resource database, and according to the learner's knowledge gap model, ideological and political education resources are matched to obtain intelligent recommendation results, including:

[0098] S301. Based on the learner's lack of understanding of the knowledge concepts corresponding to the concept nodes, keyword pre-screening is performed on the resource vectors in the ideological and political education resource database to obtain multiple candidate resource vectors.

[0099] This is illustrative of a preliminary screening of the ideological and political education resource database based on learners' lack of understanding of the knowledge concepts corresponding to the concept nodes, in order to significantly narrow down the range of candidate resources and avoid high-cost vector comparisons across the entire resource database.

[0100] S302. Based on the final embedding vector of each node, calculate the cosine similarity between the learner's knowledge gap model and multiple candidate resource vectors.

[0101] The learner knowledge gap model vector is a high-dimensional representation formed by weighted aggregation of embedding vectors of multiple unfamiliar knowledge concepts, while each resource vector can be derived from the SBERT-encoded representation of the resource title and summary text. A cosine similarity score is calculated for each pair of vectors as a matching score; a matching score closer to 1 indicates greater semantic similarity.

[0102] S303. Sort the corresponding resource vectors in descending order of cosine similarity to obtain the recommendation list.

[0103] Furthermore, all candidate resource vectors are sorted from high to low based on their similarity scores to generate a recommended candidate list.

[0104] S304. Based on the preset recommendation quantity threshold, extract the educational resources corresponding to the recommendation list to obtain the intelligent recommendation results.

[0105] For example, if the preset recommendation quantity is 5, then the top 5 resources with the highest similarity will be selected to recommend to the user.

[0106] In one embodiment, the ideological and political education resource database is constructed in the following manner:

[0107] S31. Obtain external educational resources.

[0108] External educational resources can include, but are not limited to, publicly available ideological and political education video courses released by national education authorities, policy interpretations published by authoritative media, theoretical articles written by experts and scholars, campus ideological and political education case studies, and Party history learning materials. Optionally, these resources can be obtained through batch import from the platform, API (Application Programming Interface) data integration, or manually uploaded local files by teachers or administrators. Resource types can include videos, audio, web pages with mixed text and images, and PDF (Portable Document Format) documents.

[0109] S32. Input metadata into multiple educational resources contained in external educational resources to obtain a metadata dataset; the metadata dataset includes multiple educational resources and their corresponding metadata; the metadata includes titles, abstracts, keywords and category labels.

[0110] This example illustrates the process of inputting structured metadata into the acquired educational resources to form a resource metadata dataset. Specifically, each educational resource requires the extraction and storage of its basic attribute information during input, typically including but not limited to title, abstract, keywords, and category tags. The title is the explicit name of the resource, the abstract is a summary of its content, keywords are core terms extracted manually or by algorithms, and category tags reflect its subject matter. Optionally, the metadata can be filled in by the resource provider during upload or automatically generated with the assistance of natural language processing tools.

[0111] S33. The metadata of the metadata dataset is encoded based on the pre-trained Chinese SEBRT model to obtain multiple resource vectors.

[0112] As an example, semantic encoding is performed on each resource in the metadata set to generate an embedding representation that can be used for vectorized matching. For instance, a pre-trained SBERT model is used as the embedding tool. The title and summary of each resource are concatenated into a short text and input into the SBERT model to generate a high-dimensional semantic vector corresponding to the resource, so as to comprehensively capture the core content, domain semantics and keyword information of the resource.

[0113] S34. Construct a resource database for ideological and political education based on multiple resource vectors.

[0114] Furthermore, all generated resource vectors are uniformly stored in a unified database to construct an ideological and political education resource database. Each resource vector corresponds one-to-one with its original metadata and complete resource content, and is stored in a vector index database. Optionally, the ideological and political education resource database can adopt a structure that separates the traditional relational database from the vector storage module, or it can be a hybrid database system that integrates vector retrieval functionality.

[0115] In one embodiment, the method further includes:

[0116] S41. Obtain learners' feedback data on the educational resources corresponding to the intelligent recommendation results; feedback data includes favorites, read marks, and irrelevant feedback.

[0117] S42. Adjust the knowledge gap weights based on the feedback data and reorder the recommendation list to obtain new intelligent recommendation results.

[0118] As an example, a tabbed interface can be set up to present video and text recommendation results separately, and users can mark resources as read, indicate as irrelevant, or favorite them. Furthermore, personalized knowledge graphs and recommendation strategies can be dynamically updated based on user feedback, thereby achieving a continuously evolving intelligent resource matching function.

[0119] The knowledge graph-based intelligent matching method for ideological and political education resources supports operation on teaching systems, online education platforms, or interactive learning terminals. Taking slideshows as an example, after teachers upload courseware, a knowledge concept mapping is automatically generated. Learners log in, study page by page, and mark their notes. The backend updates the knowledge graph and gap model in real time and pushes targeted resources after each course.

[0120] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0121] Based on the same inventive concept, this application also provides a knowledge graph-based intelligent matching system for ideological and political education resources, used to implement the aforementioned knowledge graph-based intelligent matching method for ideological and political education resources. The solution provided by this device is similar to the implementation scheme described in the above method. Therefore, the specific limitations of one or more knowledge graph-based intelligent matching system embodiments provided below can be found in the limitations of the knowledge graph-based intelligent matching method for ideological and political education resources described above, and will not be repeated here.

[0122] In one exemplary embodiment, such as Figure 4 As shown, a knowledge graph-based intelligent matching system for ideological and political education resources is provided, including:

[0123] The learning resource module 401 is used to acquire the dataset of ideological and political education teaching materials and extract knowledge concepts to obtain the set of relationships between ideological and political education teaching materials and knowledge concepts. The set of relationships between ideological and political education teaching materials and knowledge concepts includes multiple ideological and political education teaching materials and their corresponding knowledge concepts.

[0124] The learner-personalized knowledge graph module 402 is used to respond to the learner's marking of knowledge concepts corresponding to ideological and political education teaching materials, and to build a learner-personalized knowledge graph based on the set of associations between ideological and political education teaching materials and knowledge concepts.

[0125] Knowledge Gap Module 403 is used to build learner knowledge gap models based on personalized knowledge graphs.

[0126] The resource matching module 404 is used to match ideological and political education resources based on the ideological and political education resource database and the learner's knowledge gap model to obtain intelligent recommendation results.

[0127] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps in the above method embodiments.

[0128] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0129] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0130] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A knowledge graph-based intelligent matching method for ideological and political education resources, characterized in that, The method includes: A dataset of ideological and political education teaching materials is obtained, and knowledge concepts are extracted to obtain a set of relationships between ideological and political education teaching materials and knowledge concepts; the set of relationships between ideological and political education teaching materials and knowledge concepts includes multiple ideological and political education teaching materials and their corresponding knowledge concepts; In response to the learner's markings of the knowledge concepts corresponding to the ideological and political education teaching materials, a personalized knowledge graph for the learner is established based on the set of associations between the ideological and political education teaching materials and the knowledge concepts. The markings include understanding markings and non-understanding markings. The personalized knowledge graph includes nodes and corresponding relation edges. The nodes include ideological and political education teaching material nodes, knowledge concept nodes, and learner nodes. The relation edges correspond to relations including read, included, understood, and not understood. Construct a learner knowledge gap model based on the personalized knowledge graph; Based on the ideological and political education resource database, ideological and political education resources are matched according to the learner knowledge gap model to obtain intelligent recommendation results; the ideological and political education resource database includes multiple resource vectors.

2. The method according to claim 1, characterized in that, The process of acquiring a dataset of ideological and political education teaching materials and extracting knowledge concepts to obtain a set of relationships between ideological and political education teaching materials and knowledge concepts includes: The keyword extraction algorithm is used to extract key phrases from each ideological and political education teaching material in the dataset to obtain multiple candidate keyword phrases. The knowledge concept is obtained by performing entity disambiguation on multiple candidate keyword phrases using a Chinese entity linking algorithm. By associating the knowledge concepts with the corresponding ideological and political education teaching materials, a set of association relationships between ideological and political education teaching materials and knowledge concepts is obtained.

3. The method according to claim 1, characterized in that, The learner's personalized knowledge graph was obtained through the following methods: Based on the ideological and political education teaching materials corresponding to the marked information, construct the read relationship edge between the ideological and political education teaching material node and the learner node; Based on the set of relationships between ideological and political education teaching materials and knowledge concepts, establish the inclusion relationship edges between the nodes of the ideological and political education teaching materials and the nodes of the knowledge concepts; Based on the understanding tags, establish the understanding relationship edges between the knowledge concept nodes and the learner nodes; Based on the misunderstanding markers, establish the misunderstanding relationship edges between the knowledge concept nodes and the learner nodes; By integrating the nodes of ideological and political education teaching materials, the nodes of learners, and the nodes of knowledge concepts, as well as their corresponding relational edges, the learner's personalized knowledge graph is obtained.

4. The method according to claim 1, characterized in that, The step of constructing a learner knowledge gap model based on the personalized knowledge graph includes: Based on the pre-trained Chinese SEBRT model, the concept name of each knowledge concept node is encoded to obtain an initial concept semantic embedding vector; Based on the pre-trained Chinese SEBRT model, the text content corresponding to each node of the ideological and political education teaching materials is extracted to obtain the initial embedding vector of the ideological and political education teaching materials. Based on graph convolutional networks, multiple rounds of information aggregation are performed on each node of the personalized knowledge graph according to the cosine similarity of the SEBRT vectors to obtain the final embedding vector of each node; the SEBRT vector is composed of the initial concept semantic embedding vector and the initial ideological and political education teaching material embedding vector; The relational edges are extracted from the personalized knowledge graph to identify the concept nodes that the learner does not understand, thus obtaining the concept nodes that the learner does not understand. The cosine similarity between the initial concept semantic embedding vector corresponding to the concept node that the learner does not understand and the initial ideological and political education teaching material embedding vector is calculated to obtain the knowledge gap weight. Based on a weighted average, the learner knowledge gap model is synthesized according to the initial concept semantic embedding vector corresponding to the concept node that the learner does not understand and the knowledge gap weight.

5. The method according to claim 4, characterized in that, The process of matching ideological and political education resources based on the ideological and political education resource database and the learner's knowledge gap model to obtain intelligent recommendation results includes: Based on the learner's lack of understanding of the knowledge concept corresponding to the concept node, keyword pre-screening is performed on the resource vectors in the ideological and political education resource database to obtain multiple candidate resource vectors; Based on the final embedding vector of each node, the cosine similarity between the learner knowledge gap model and multiple candidate resource vectors is calculated. The resource vectors are sorted in descending order of cosine similarity to obtain a recommendation list; Based on a preset recommendation quantity threshold, the educational resources corresponding to the recommendation list are extracted to obtain the intelligent recommendation result.

6. The method according to claim 1, characterized in that, The ideological and political education resource database is constructed in the following ways: Access to external educational resources; Metadata is entered into the external educational resources to obtain a metadata dataset; the metadata dataset includes multiple educational resources and their corresponding metadata; the metadata includes title, summary, keywords and category labels; The metadata of the metadata dataset is encoded based on a pre-trained Chinese SEBRT model to obtain multiple resource vectors; The ideological and political education resource library is constructed based on multiple resource vectors.

7. The method according to claim 5, characterized in that, The method further includes: Obtain learners' feedback data on the educational resources corresponding to the intelligent recommendation results; the feedback data includes favorites, read marks, and irrelevant feedback; The knowledge gap weights are adjusted based on the feedback data, and the recommendation list is reordered to obtain new intelligent recommendation results.

8. A knowledge graph-based intelligent matching system for ideological and political education resources, characterized in that, The system includes: The learning resources module is used to acquire a dataset of ideological and political education teaching materials and extract knowledge concepts to obtain a set of relationships between ideological and political education teaching materials and knowledge concepts; the set of relationships between ideological and political education teaching materials and knowledge concepts includes multiple ideological and political education teaching materials and their corresponding knowledge concepts; The learner personalized knowledge graph module is used to respond to the learner's marking of the knowledge concepts corresponding to the ideological and political education teaching materials, and to establish a learner personalized knowledge graph based on the set of associations between the ideological and political education teaching materials and knowledge concepts. The knowledge gap module is used to construct a learner's knowledge gap model based on the personalized knowledge graph. The resource matching module is used to match ideological and political education resources based on the ideological and political education resource library and the learner's knowledge gap model to obtain intelligent recommendation results.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

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

Images