A large model driven knowledge graph construction and teaching resource generation method

Abstract

This invention proposes a large-model-driven method for knowledge graph construction and teaching resource generation, comprising: multi-granularity semantic extraction: using a large language model to perform layout analysis and semantic parsing on imported unstructured corpora, extracting knowledge triples containing entities, attributes, and logical relationships between entities; dynamic graph construction and alignment: mapping the knowledge triples to a graph database, and fusing entities with the same name in heterogeneous corpora by calculating entity vector similarity, constructing a teaching knowledge graph with a topological structure; utilizing a large language model combined with a layout recognition model and sliding window technology to achieve multi-granularity semantic parsing of unstructured teaching corpora, accurately extracting knowledge triples containing entities, attributes, and long-range semantic relationships across paragraphs, improving knowledge extraction efficiency compared to traditional manual annotation or rule extraction methods.

Description

Technical Field

[0001] This invention relates to the field of natural language processing, and more particularly to a method for constructing knowledge graphs and generating teaching resources driven by large models. Background Technology

[0002] With the deepening of digital transformation in smart education, the education industry has an increasingly urgent need for structured modeling of knowledge and generation of intelligent teaching resources.

[0003] Existing technologies rely heavily on manual annotation or traditional rule-based methods for entity and relation extraction, which is difficult to adapt to the multi-granularity semantic parsing of unstructured teaching corpora. This results in uneven granularity of knowledge point extraction, omission of entity relations, long construction cycles, and high costs, making it difficult to meet the needs of large-scale teaching knowledge modeling. For multi-source heterogeneous teaching corpora, existing technologies cannot effectively solve the problems of disambiguation of entities with the same name and fusion of synonymous entities, which can easily lead to redundant nodes and chaotic relationships in the knowledge graph, making it impossible to form a unified and standardized teaching knowledge system. Summary of the Invention

[0004] The technical problem to be solved by this invention is to overcome the defects of the existing technology. This invention proposes a method for knowledge graph construction and teaching resource generation driven by large models.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: Firstly, this invention provides a method for large-model-driven knowledge graph construction and teaching resource generation, including: S1: Multi-granularity semantic extraction: Utilize a large language model to perform layout analysis and semantic parsing on the imported unstructured corpus, and extract knowledge triples containing entities, attributes, and logical relationships between entities; S2: Dynamic Graph Construction and Alignment: The knowledge triples are mapped to a graph database, and entities with the same name in heterogeneous corpora are fused by calculating entity vector similarity to construct a teaching knowledge graph with a topological structure. S3: Bidirectional binding of multimodal resources: Establishing an index mapping between knowledge graph nodes and semantic blocks of the original corpus and external multimodal material libraries to form enhanced knowledge nodes that combine knowledge graph logical mapping and corpus materials; S4: Teaching plan planning for knowledge graph topology: Based on the teaching objectives input by the user, perform topological sorting and path planning on the knowledge graph, call the preset lesson plan template, fill the path nodes and their bound materials into the template, and automatically generate structured lesson plans and visual courseware. S5: Graph-constrained retrieval enhanced question answering: Based on user questions about knowledge nodes, it combines the subgraph structure of the knowledge graph with the vector knowledge base for joint retrieval to generate auxiliary teaching questions and answers with traceable evidence.

[0006] Preferably, the semantic parsing in step S1 includes: using a layout recognition model to locate hierarchical headings, formulas, and chart indexes in unstructured corpora; and using a sliding window technique in conjunction with an overlap area mechanism to extract long-range semantic association triples across paragraphs using a large language model.

[0007] Preferably, the calculation of entity vector similarity in step S2 uses the following formula. When the weighted similarity exceeds a preset threshold, the nodes are merged and the graph adjacency matrix is ​​updated.

[0008] Preferably, the bidirectional binding of multimodal resources in step S3 includes: assigning a globally unique identifier to each knowledge node and using it as the ID of the original corpus fragment and the URL index of the multimodal material in the vector database.

[0009] Preferably, the path planning in step S4 includes: constructing a directed acyclic graph based on the prior dependencies in the knowledge graph; calculating the teaching weight of nodes using the PageRank algorithm; and generating a logical sequence that conforms to the principle of gradual teaching by combining a breadth-first search strategy.

[0010] Preferably, the automatic generation of structured lesson plans in step S4 includes: converting the generated logical sequence into an intermediate state description language; and mapping the intermediate state description language to preset presentation page placeholders through a style engine to realize the layout and multimodal content filling of the courseware.

[0011] Preferably, the graph-constrained retrieval enhanced question answering in step S5 includes: extracting the core entity from the query request, retrieving its K-order domain subgraph in the knowledge graph to obtain contextual logic; concatenating the subgraph logical relationship with the text slices retrieved from the vector library to form enhanced prompt words, and generating an answer containing original text citation tags by the large language model.

[0012] Secondly, an electronic device includes: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the large model-driven knowledge graph construction and teaching resource generation method.

[0013] Thirdly, a computer-readable storage medium stores computer-executable instructions that, when executed by a processor, implement the steps of the large model-driven knowledge graph construction and teaching resource generation method.

[0014] Compared with existing technologies, the beneficial effects of this invention include: utilizing a large language model combined with a layout recognition model and sliding window technology to achieve multi-granular semantic parsing of unstructured teaching corpora, accurately extracting knowledge triples of entities, attributes, and long-range semantic relationships across paragraphs. Compared with traditional manual annotation or rule extraction methods, the knowledge extraction efficiency is improved, and the accuracy of triples is significantly improved. A weighted matching mechanism combining entity vector similarity and string edit distance is introduced to achieve intelligent disambiguation and fusion of entities with the same name or synonyms in heterogeneous corpora, eliminate redundancy of knowledge graph nodes, and construct a teaching knowledge graph with a standardized structure and clear associations, laying a high-quality data foundation for subsequent teaching applications. Attached Figure Description

[0015] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts. Wherein: Figure 1 The flowchart shows the method for large-model-driven knowledge graph construction and teaching resource generation in this application. Figure 2 The diagram illustrates an electronic device for the method of knowledge graph construction and teaching resource generation driven by the large model according to the present invention. Detailed Implementation

[0016] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0017] Figure 1 The flowchart of the method for large model-driven knowledge graph construction and teaching resource generation provided in this embodiment of the invention is shown below. Figure 1 The method includes: S1, Multi-granularity semantic extraction The system calls a large language model to perform two core processing steps on the imported unstructured corpus, ultimately extracting knowledge triples. The specific process is as follows: Page layout analysis: Using a page layout recognition model, unstructured corpora are analyzed globally to accurately locate hierarchical headings (such as chapter titles and subsection titles), formula content, and chart indexes in the corpus, clarifying the structural hierarchy and core information location of the corpus, and providing layout support for subsequent semantic analysis.

[0018] Semantic parsing and triple extraction: The sliding window technique combined with the overlap region mechanism is used to solve the problem of difficulty in capturing semantic associations across paragraphs. Through a large language model, the semantic analysis of the corpus is performed paragraph by paragraph and across paragraphs to accurately identify entities (such as teaching knowledge points, concepts, and terms), entity attributes (such as entity definitions, features, and parameters), and logical relationships between entities (such as inclusion relationships, causal relationships, and subordinate relationships). The three are combined to form knowledge triples, entity 1-relation-entity 2 or entity 1-relation-entity attribute, to complete the transformation of unstructured corpus into structured knowledge units.

[0019] Specifically, text segmentation and vectorization The corpus is divided into n semantic blocks. The Embedding model is used to map it to a vector space. Thus, we obtain the knowledge triplet.

[0020] Among them, v i For the i-th entity / concept / text fragment c i The corresponding embedding vector is used to represent the semantic information of the object.

[0021] f emb The embedded model function is the core computational logic that maps unstructured inputs into vectors.

[0022] c i The i-th input object refers to an entity, concept, knowledge point, or fragment of original corpus in the teaching knowledge graph.

[0023] S2, Dynamic Graph Construction and Alignment Using all the knowledge triples extracted from S1 as the basic data, a knowledge graph is constructed and optimized to ensure its accuracy and completeness. The specific process is as follows: Graph mapping: The extracted knowledge triples are mapped one by one to a pre-set graph database. The entities in the triples are used as graph nodes, and the logical relationships between entities are used as edges between nodes, thus initially constructing a knowledge graph prototype with a basic association structure.

[0024] Entity fusion and graph alignment: Since the imported corpus may be heterogeneous (such as different versions of textbooks or lecture notes from different sources), there may be entities with the same name but different meanings or the same meaning but different names. The system calculates the similarity of entity vectors. When the weighted similarity exceeds a preset threshold, it fuses entities with the same name in the heterogeneous corpus, eliminates redundant nodes, and updates the adjacency matrix of the graph. Finally, it constructs a teaching knowledge graph with a clear structure, accurate associations, and a topological structure.

[0025] Specifically, the calculation of triple confidence. Extracting triples using LLM (Head entity, relation, tail entity). To ensure accuracy, a confidence score formula is introduced: (1) in, The overall confidence score of the knowledge triple T is the final weighted value used to measure the effectiveness of entity fusion or triples. The probability score generated for the large model; count(T) is the frequency of triple T in the heterogeneous corpus; This is the sum of the frequencies of all candidate triplets, and this term represents the normalized proportion of the frequencies. ω1 represents the cosine similarity between entities, used to measure the semantic rationality of the relationship; ω1, ω2, and ω3 are the weight coefficients of the three dimensions, satisfying ω1+ω2+ω3=1, used to adjust the proportion of each feature in the final score.

[0026] Ontology Fusion and Disambiguation in Knowledge Graphs To prevent discrepancies in the representation of the same concept across different corpora, node fusion is necessary.

[0027] Entity similarity measurement Calculate the newly extracted entity With existing entities in the map Overall similarity: (2) in, The cosine similarity of the embedding vectors of the new and old entities. The preset weighting coefficients are (0 ≤ α ≤ 1). When S match When the threshold is exceeded, the entities are identified as the same node, the entities with the same name are merged, and the knowledge graph adjacency matrix is ​​updated. This is the Levenstein distance, used to measure the difference between two strings.

[0028] Graph Topology Construction Construct an adjacency matrix A for the integrated teaching knowledge graph, where matrix elements A... ij Used to represent the connection relationship between knowledge point i and knowledge point j: If A ij =1, indicating that knowledge points i and j have a pre-teaching dependency or semantic association; if A ij =0 indicates that there is no direct relationship between the two. This adjacency matrix provides topological data support for subsequent teaching path planning and knowledge point weight calculation.

[0029] S3, Two-way binding of multimodal resources: To enhance the practicality and pedagogical adaptability of knowledge graphs, connections are established between knowledge graphs and various materials to form enhanced knowledge nodes. The specific process is as follows: The system assigns a globally unique identifier (ID) to each knowledge node in the teaching knowledge graph. This identifier serves as the core association index, linking to the corresponding original corpus fragment ID in the vector database to achieve a bidirectional mapping between the knowledge node and the semantic blocks of the original corpus (the node can be quickly traced back to the original corpus, and the corresponding knowledge node can be located through the original corpus). On the other hand, it links to the URL index of the corresponding material in an external multimodal material library. The multimodal materials may include teaching materials such as images, videos, audio, and animations, thus forming an enhanced knowledge node that integrates the logical mapping of the knowledge graph, original corpus fragments, and multimodal teaching materials, providing material support for the subsequent generation of teaching resources.

[0030] Specifically, the teaching scheme planning based on knowledge graph topology The system is based on the teaching objective nodes. It automatically recommends the teaching sequence.

[0031] Node importance assessment The coreness of each knowledge node in the teaching system was calculated using an improved PageRank algorithm as follows: (3) in, The importance of the node as a foundation, For topology dependency propagation terms, We define custom weight enhancements for teaching and obtain the teaching weights of each node through iterative calculation, providing a basis for generating knowledge sequences that are adapted to teaching objectives using breadth-first search.

[0032] Topological sorting generates sequences Solve a sequence while satisfying the prerequisite dependencies. This ensures that the sequence covers all core nodes and minimizes path cost.

[0033] S4. Teaching plan for knowledge graph topology Based on the completed teaching knowledge graph and combined with the teaching objectives input by the user, structured lesson plans and visual courseware are automatically generated. The specific process consists of two steps: Path planning: First, based on the preorder dependencies of each knowledge node in the teaching knowledge graph (such as the order of basic knowledge points and advanced knowledge points), a directed acyclic graph is constructed; then, the PageRank algorithm is used to calculate the teaching weight of each knowledge node (the weight corresponds to the importance of the knowledge point in teaching), and combined with the breadth-first search strategy, the logical sequence of knowledge nodes that conforms to the principle of gradual teaching is selected from the knowledge graph. This sequence is the teaching content order that is adapted to the teaching objectives.

[0034] Automatic generation of teaching resources: The system calls preset lesson plan templates (such as standard lesson plan templates for different subjects and different class periods), converts the generated knowledge node logical sequence into intermediate state description language; then, through the style engine, it maps the intermediate state description language to preset presentation page placeholders, and automatically extracts the multimodal materials bound to each knowledge node (called through URL index) and fills them into the corresponding positions in the courseware, finally completing the automatic generation of structured lesson plans (including teaching objectives, teaching content, teaching steps, etc.) and visual courseware (including knowledge point presentation, material display, etc.).

[0035] Specifically, template mapping logic The generated path sequence Input into the prompt container .

[0036] Content generation function (4) Among them, C prompt To pre-set teaching question and answer prompts, v i `v1` represents the embedding vector of the core knowledge node (representing the logical context of the K-order subgraph), `Content(vi)` represents the original corpus fragment (including source markers) bound to the node, and `⊕` represents the information fusion operation. Based on the fused enhanced prompt words, the large language model generates auxiliary teaching question-and-answer content that combines logical relevance and source evidence.

[0037] The system is based on nodes The bound multimodal index retrieves the corresponding images or charts, and the text and multimodal data are rendered into lesson plans and courseware using the lesson plan library and slide library.

[0038] S5, Graph-Constrained Search Enhanced Question Answering In response to user questions related to teaching knowledge, precise answers with traceable evidence are provided based on knowledge graphs to assist in teaching. The specific process is as follows: Core entity extraction: The system performs semantic parsing on the user's input questions and extracts the core entities (i.e., the teaching knowledge points or concepts that the question refers to).

[0039] Subgraph retrieval and context retrieval: Centered on the extracted core entity, retrieve its K-order domain subgraph (i.e., the surrounding nodes and relationships directly and indirectly associated with the core entity) in the teaching knowledge graph to obtain the contextual logical relationship of the core entity, ensuring the logic and relevance of the answer.

[0040] Enhanced prompts and answer generation: The retrieved subgraph logical relationships are combined with the corresponding corpus text slices retrieved from the vector database (associated through the globally unique identifiers of knowledge nodes) to form enhanced prompt words, which are then input into the large language model. Based on these enhanced prompt words, the large language model generates answers containing original text citation tags (marking the original corpus fragments or knowledge nodes from which the answer originates), ensuring the traceability of the answers and providing reliable support for teaching.

[0041] Specifically, a bidirectional question-answering mechanism based on GraphRAG. During the question-and-answer process, the system does not directly search the text, but instead locates the information in the graph and then performs retrieval enhancement to help users better trace the source of the questions and answers.

[0042] Search Enhancement Computation By combining the weighted scores of vector retrieval and graph retrieval, node origination is performed. The specific weighting method is as follows: (5) in, To query the semantic similarity with candidate text fragments, To query the sum of weights of neighboring nodes of the core node, λ is a preset weight coefficient. This formula is used to filter out text fragments that have both semantic relevance and logical rationality, providing support for the subsequent generation of traceable teaching questions and answers by the large language model.

[0043] Inference generation The retrieved graph structure information and text fragments are submitted together as Context to LLM to generate a response with traceability capabilities.

[0044] This embodiment also provides a schematic diagram of an electronic device structure for a method of large-model-driven knowledge graph construction and teaching resource generation. The electronic device is intended to represent various forms of digital computers, such as laptops, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframes, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and its… The functions described herein are merely illustrative and are not intended to limit the implementation of the invention as described and / or claimed herein.

[0045] Electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer programs stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 may also store various programs and data required for the operation of electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0046] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication... Communication unit 19, such as a network interface card (NIC), modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0047] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as a... A method for knowledge graph construction and teaching resource generation driven by large models.

[0048] In some embodiments, the large model-driven knowledge graph construction and teaching resource generation method can be implemented as a computer program, which is used by... It is contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, a portion or part of the computer program is included in the storage medium. All of these can be loaded and / or installed onto the electronic device 10 via ROM 12 and / or communication unit 19. When the computer program When the sequence is loaded into RAM 13 and executed by processor 11, one or more steps of the large model-driven knowledge graph construction and instructional resource generation method described above can be performed. Alternatively, in other embodiments, processor 11 can be configured to perform the large model-driven knowledge graph construction and instructional resource generation method by any other suitable means (e.g., by means of firmware).

[0049] The various implementations of the systems and technologies described above in this article can be applied to digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), and systems-on-chips. The implementation may be carried out in a system of components (SOC), a load-programmable logic device (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include: implementation in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transferring data and instructions to the storage system, the at least one input device, and the at least one output device.

[0050] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0051] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable memory (EPM), and other similar media. Read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above.

[0052] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or an LCD (liquid crystal display) for displaying information to the user. The device includes a monitor and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or haptic feedback); and input from the user can be received in any form (including voice input, speech input, or haptic input).

[0053] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0054] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0055] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0056] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for constructing knowledge graphs and generating teaching resources driven by a large model, characterized in that, include: S1: Multi-granularity semantic extraction: Utilize a large language model to perform layout analysis and semantic parsing on the imported unstructured corpus, and extract knowledge triples containing entities, attributes, and logical relationships between entities; S2: Dynamic Graph Construction and Alignment: The knowledge triples are mapped to a graph database, and entities with the same name in heterogeneous corpora are fused by calculating entity vector similarity to construct a teaching knowledge graph with a topological structure. S3: Bidirectional binding of multimodal resources: Establishing an index mapping between knowledge graph nodes and semantic blocks of the original corpus and external multimodal material libraries to form enhanced knowledge nodes that combine knowledge graph logical mapping and corpus materials; S4: Teaching plan planning for knowledge graph topology: Based on the teaching objectives input by the user, perform topological sorting and path planning on the knowledge graph, call the preset lesson plan template, fill the path nodes and their bound materials into the template, and automatically generate structured lesson plans and visual courseware. S5: Graph-constrained retrieval enhanced question answering: Based on user questions about knowledge nodes, it combines the subgraph structure of the knowledge graph with the vector knowledge base for joint retrieval to generate auxiliary teaching questions and answers with traceable evidence.

2. The method for large-model-driven knowledge graph construction and teaching resource generation according to claim 1, characterized in that, The semantic parsing in step S1 includes: using a layout recognition model to locate hierarchical headings, formulas, and chart indexes in unstructured corpora; and using a sliding window technique combined with an overlap area mechanism to extract long-range semantic association triples across paragraphs using a large language model.

3. The method for large-model-driven knowledge graph construction and teaching resource generation according to claim 1, characterized in that, The calculation of entity vector similarity in step S2 uses the following formula. When the weighted similarity exceeds a preset threshold, the nodes are merged and the graph adjacency matrix is ​​updated.

4. The method for large-model-driven knowledge graph construction and teaching resource generation according to claim 1, characterized in that, Step S3, which involves two-way binding of multimodal resources, includes: assigning a globally unique identifier to each knowledge node and using it as the ID of the original corpus fragment and the URL index of the multimodal material in the vector database.

5. The method for large-model-driven knowledge graph construction and teaching resource generation according to claim 1, characterized in that, The path planning described in step S4 includes: constructing a directed acyclic graph based on the prior dependencies in the knowledge graph; calculating the teaching weight of nodes using the PageRank algorithm; and generating a logical sequence that conforms to the principle of gradual teaching by combining a breadth-first search strategy.

6. The method for large-model-driven knowledge graph construction and teaching resource generation according to claim 1, characterized in that, Step S4, which involves automatically generating structured lesson plans, includes: converting the generated logical sequence into an intermediate state description language; and mapping the intermediate state description language to preset presentation page placeholders through a style engine to achieve the layout and multimodal content filling of the courseware.

7. The method for large-model-driven knowledge graph construction and teaching resource generation according to claim 1, characterized in that, Step S5, the graph-constrained retrieval enhanced question answering, includes: extracting the core entities from the query request, retrieving their K-order domain subgraphs in the knowledge graph to obtain contextual logic; concatenating the subgraph logical relationships with the text slices retrieved from the vector library to form enhanced prompt words, and generating an answer containing original text citation tags by the large language model.

8. An electronic device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the large model-driven knowledge graph construction and teaching resource generation method according to any one of claims 1 to 7.

9. A computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the large model-driven knowledge graph construction and teaching resource generation method according to any one of claims 1 to 7.

Images