Digital course generation method and system based on large language model

Abstract

The invention discloses a digital curriculum generation method and system based on a large language model, and the method comprises the steps: carrying out the tight combination of the acquisition of curriculum data, the chapter splitting and semantic analysis based on the large language model, and the construction of a multi-modal index, automatically and uniformly mapping the multi-source heterogeneous information in the teaching data into a clear chapter hierarchical structure, and achieving the uniform mapping of the multi-source heterogeneous information in the teaching data. On the basis, a structured model is quickly generated, so that the organization efficiency and the processing consistency of the course content are improved; then semantic feature extraction is carried out on the structured model, chapter level information is fused to construct a knowledge graph, and efficient conversion from original teaching materials to a knowledge network capable of being intelligently retrieved and visually presented is achieved; and finally, seamlessly integrating the generated digital human image as well as the expression and the teaching action thereof with the 3D teaching scene, the structured model and the knowledge graph, and completely outputting to form a digital course, thereby greatly shortening the overall process duration from material preparation to course delivery.

Description

Technical Field

[0001] The present invention relates to the field of digital course technology, and in particular to a digital course generation method and system based on a large language model. Background Art

[0002] Currently, with the application of big data and artificial intelligence technologies, digital textbooks and teaching platforms can achieve personalized learning and intelligent recommendations, improving learning efficiency and effectiveness. Furthermore, the integration of virtual reality (VR) and augmented reality (AR) technologies makes online teaching content more vivid and immersive, enhancing the interactive experience and helping students understand complex knowledge. Furthermore, the use of learning behavior data analysis can build adaptive learning paths, adjust teaching strategies in a timely manner, and promote individualized teaching. Generative AI significantly reduces the burden on teachers in content creation and automatic script generation, enabling intelligent content production. The remote and scalable nature of digital courses effectively meets the learning needs of students in remote areas or during special periods, promoting the balanced distribution of educational resources.

[0003] Although existing digital course generation methods can improve content production efficiency to a certain extent, they also have many shortcomings: First, the generation process often relies on fixed templates and preset rules, lacks personalized customization for different learning objects, and is difficult to meet diverse learning needs; second, automated content production often ignores teaching design principles, and the quality of the generated courseware and question banks varies, making it difficult to ensure depth and accuracy; third, these methods generally lack real-time teacher-student interaction and emotional feedback mechanisms, resulting in a relatively one-way learning experience, which affects teaching participation; in addition, the underlying technical architecture and algorithm implementation often have high barriers to entry, requiring a large amount of hardware and maintenance costs, making them difficult to popularize; at the same time, the knowledge update cycle of the generated content is long, and the knowledge base is prone to information lag and obsolescence; third, data privacy and copyright protection have not yet formed unified standards, which may cause security and compliance risks in large-scale applications; finally, due to the lack of unified standards and interfaces between different platforms, there are obstacles to cross-system sharing and compatibility of course content, which increases the complexity of subsequent maintenance and upgrades. Summary of the Invention

[0004] The present invention provides a digital course generation method and system based on a large language model, so as to improve the flexibility of digital course generation and the quality of digital course generation.

[0005] In order to solve the above technical problems, an embodiment of the present invention provides a digital course generation method based on a large language model, comprising: Acquire teaching materials, perform chapter segmentation and semantic analysis on the teaching materials based on a large language model, obtain chapter-level information of the teaching materials, and construct a multimodal index based on the chapter-level information to generate a course structured model; Performing feature extraction on the course structured model based on the large language model to obtain a semantic feature vector, and constructing a course knowledge graph based on the semantic feature vector and the chapter level information; Generate a digital human image based on three-dimensional facial modeling technology, and generate a digital human expression sequence and a teaching action sequence based on the digital human image, preset voice data, the course structured model and a preset teaching action library; Based on the digital human image, the digital human expression sequence and the teaching action sequence, the digital human is embedded in the 3D teaching scene model, and the digital course is synthesized based on the 3D teaching scene model, the course structured model and the course knowledge graph.

[0006] The present invention closely combines the acquisition of course materials, chapter segmentation and semantic analysis based on a large language model, and the construction of a multimodal index, automatically mapping the multi-source heterogeneous information in the teaching materials into a clear chapter hierarchy, and quickly generating a structured model on this basis, thereby improving the organizational efficiency and processing consistency of the course content; subsequently, by extracting semantic features from the structured model and integrating the chapter hierarchy information into the knowledge graph, an efficient conversion from the original teaching materials to a knowledge network that can be intelligently retrieved and visualized is achieved; finally, the generated digital human image, its expression and teaching movements are seamlessly integrated with the 3D teaching scene, the structured model and the knowledge graph, and the complete output constitutes a digital course, which greatly shortens the overall process time from material preparation to course delivery, and improves the intelligent and modular teaching generation capabilities.

[0007] Furthermore, the teaching materials include text materials, voice materials, and 3D teaching materials. The obtaining of the teaching materials, performing chapter segmentation and semantic analysis on the teaching materials based on a large language model, obtaining chapter-level information of the teaching materials, and constructing a multimodal index based on the chapter-level information to generate a course structured model includes: Acquire teaching materials, split the text materials into chapters based on the large language model, and obtain a list of chapter text segments of the teaching materials; Performing semantic analysis on the chapter text segment list segment by segment based on the large language model to obtain keywords and semantic summaries of each chapter text segment of the teaching material, and generating chapter-level information based on the keywords and semantic summaries of the chapter text segments and the chapter text segment list; Numbering the chapter-level information, and constructing multimodal index entries between the chapter-level information and text materials, voice materials, and 3D teaching materials based on the numbers; Based on the chapter level information and multimodal index entries, a course structured model is constructed.

[0008] The present invention uses a large language model to perform fine chapter segmentation and semantic analysis on text materials, obtains an accurate list of chapter text segments and their keywords and summaries, and ensures the logical integrity of the course framework; then numbers the chapter level information, and automatically establishes multimodal index entries for text, voice and 3D materials, realizing fine-grained association between multiple resources and chapters; finally, a unified course structured model is constructed through these index entries, which can provide a consistent and easily expandable data foundation for subsequent action synthesis, knowledge retrieval and interactive applications, thereby effectively improving the system's organization and management efficiency of multi-source teaching materials.

[0009] Furthermore, the feature extraction of the course structured model based on the large language model to obtain a semantic feature vector, and the construction of a course knowledge graph based on the semantic feature vector and the chapter level information includes: Performing feature extraction on the course structured model based on the large language model to obtain a semantic feature vector; Creating knowledge nodes based on the semantic feature vector and the chapter level information, and calculating the semantic similarity between the knowledge nodes; The chapter parent-child relationship of the knowledge node is obtained based on the chapter level information, and the associated edges between the knowledge nodes are generated based on the semantic similarity and the chapter parent-child relationship to complete the construction of the knowledge graph.

[0010] The present invention uses a large language model to extract semantic feature vectors and perform deep semantic encoding on the text of each chapter; it creates knowledge nodes based on chapter-level information and calculates the semantic similarity between nodes, which can accurately reflect the intrinsic relationship between concepts; and then generates associated edges based on the parent-child relationship and similarity of chapters. The knowledge graph finally constructed takes into account both the logical structure of the chapters and the semantic closeness between knowledge points, which is helpful for applications such as intelligent question answering, personalized recommendations and learning path planning, and effectively enhances the integrity and interpretability of knowledge representation.

[0011] Furthermore, the generation of a digital human image based on three-dimensional facial modeling technology, and the generation of a digital human expression sequence and a teaching action sequence based on the digital human image, preset voice data, the course structured model and a preset teaching action library, include: generating a digital human image based on the course structured model and three-dimensional facial modeling technology; Performing speech frame analysis and prosody analysis on the speech data to obtain semantic features; Identifying teaching intentions based on the large language model and the course structured model, and generating semantic vectors for each teaching intention; The semantic features, the semantic vectors and the preset teaching action library are weightedly fused based on the half-cosine curve function to generate a digital human expression sequence and a teaching action sequence synchronized with the current semantic intention and semantic features.

[0012] The present invention generates a high-precision digital human image based on a course structured model and three-dimensional facial modeling technology, ensuring the consistency of the virtual teacher's appearance with the teaching materials. It obtains voice features by performing voice frame segmentation and rhythm analysis on the voice data, and combines the teaching intention semantic vector identified by the large language model. It uses a half-cosine weighted fusion algorithm to fuse voice, text and action templates, so that the generated expression sequences and action sequences are highly consistent with the explanation rhythm and can accurately present the teaching intention, breaking through the limitations of a fixed action library and realizing a more natural, coherent and semantically driven interactive digital human demonstration.

[0013] Furthermore, after the knowledge graph is constructed, the following steps are further included: Collecting question and answer records from the user's learning behavior data, and classifying the question and answer records according to knowledge nodes to obtain question and answer data; Based on the question and answer data, the access frequency, answer accuracy rate and average response time of each knowledge node are counted to obtain statistical results; Based on the statistical results, the weight of each knowledge node and associated edge in the knowledge graph is dynamically updated using a preset weighted fusion model to dynamically optimize the knowledge graph.

[0014] After the knowledge graph is constructed, the present invention continues to collect and classify users' question and answer records, and counts the access frequency, answer accuracy and average response time of each knowledge node to form learning behavior data; based on these statistical results, a preset weighted fusion model is used to dynamically adjust the weights of knowledge graph nodes and associated edges, so that the graph can reflect the user's mastery and weak links in real time, thereby realizing adaptive optimization of the knowledge graph, and then supporting more accurate knowledge push and personalized tutoring.

[0015] In a second aspect, the present invention provides a digital course generation system based on a large language model, comprising: a data aggregation layer, an agent training layer, a digital human interaction layer, and an application service layer; The data aggregation layer is used to obtain teaching materials, perform chapter segmentation and semantic analysis on the teaching materials based on a large language model, obtain chapter-level information of the teaching materials, and construct a multimodal index based on the chapter-level information to generate a course structured model; The agent training layer is used to extract features from the course structured model based on the large language model, obtain semantic feature vectors, and construct a course knowledge graph based on the semantic feature vectors and the chapter level information; The digital human interaction layer is used to generate a digital human image based on three-dimensional facial modeling technology, and to generate a digital human expression sequence and a teaching action sequence based on the digital human image, preset voice data, the course structured model and a preset teaching action library; The application service layer is used to embed the digital human into the 3D teaching scene model based on the digital human image, the digital human expression sequence and the teaching action sequence, and synthesize the digital course based on the 3D teaching scene model, the course structured model and the course knowledge graph.

[0016] Furthermore, the data aggregation layer is used for teaching materials including text materials, voice materials, and 3D teaching materials. The teaching materials are obtained, the teaching materials are divided into chapters and semantically analyzed based on the large language model, the chapter-level information of the teaching materials is obtained, and a multimodal index is constructed based on the chapter-level information to generate a course structured model, including: Acquire teaching materials, split the text materials into chapters based on the large language model, and obtain a list of chapter text segments of the teaching materials; Performing semantic analysis on the chapter text segment list segment by segment based on the large language model to obtain keywords and semantic summaries of each chapter text segment of the teaching material, and generating chapter-level information based on the keywords and semantic summaries of the chapter text segments and the chapter text segment list; Numbering the chapter-level information, and constructing multimodal index entries between the chapter-level information and text materials, voice materials, and 3D teaching materials based on the numbers; Based on the chapter level information and multimodal index entries, a course structured model is constructed.

[0017] Furthermore, the agent training layer is used to extract features from the course structured model based on the large language model, obtain semantic feature vectors, and construct a course knowledge graph based on the semantic feature vectors and the chapter-level information, including: Performing feature extraction on the course structured model based on the large language model to obtain a semantic feature vector; Creating knowledge nodes based on the semantic feature vector and the chapter level information, and calculating the semantic similarity between the knowledge nodes; The chapter parent-child relationship of the knowledge node is obtained based on the chapter level information, and the associated edges between the knowledge nodes are generated based on the semantic similarity and the chapter parent-child relationship to complete the construction of the knowledge graph.

[0018] Furthermore, the digital human interaction layer is used to generate a digital human image based on three-dimensional facial modeling technology, and to generate a digital human expression sequence and a teaching action sequence based on the digital human image, preset voice data, the course structured model, and a preset teaching action library, including: generating a digital human image based on the course structured model and three-dimensional facial modeling technology; Performing speech frame analysis and prosody analysis on the speech data to obtain semantic features; Identifying teaching intentions based on the large language model and the course structured model, and generating semantic vectors for each teaching intention; The semantic features, the semantic vectors and the preset teaching action library are weightedly fused based on the half-cosine curve function to generate a digital human expression sequence and a teaching action sequence synchronized with the current semantic intention and semantic features.

[0019] Furthermore, the agent training layer is also used to: Collecting question and answer records from the user's learning behavior data, and classifying the question and answer records according to knowledge nodes to obtain question and answer data; Based on the question and answer data, the access frequency, answer accuracy rate and average response time of each knowledge node are counted to obtain statistical results; Based on the statistical results, the weight of each knowledge node and associated edge in the knowledge graph is dynamically updated using a preset weighted fusion model to dynamically optimize the knowledge graph. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 A flowchart of a method for generating digital courses based on a large language model is provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0021] The following embodiments of the present invention are described in further detail with reference to the accompanying drawings and examples. The following examples are used to illustrate the present invention but are not intended to limit the scope of the present invention.

[0022] The terms "first," "second," and the like in the specification, claims, and drawings of this application are used to distinguish between different objects, not to describe a particular order. Furthermore, the terms "including," "having," and any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or elements is not limited to the listed steps or elements, but may optionally include steps or elements not listed, or may optionally include other steps or elements inherent to the process, method, product, or apparatus.

[0023] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments. Example

[0024] See also Figure 1 , Figure 1 A flowchart of a method for generating a digital course based on a large language model provided by an embodiment of the present invention. The method includes steps 101 to 104, which are as follows: Step 101: Acquire teaching materials, perform chapter segmentation and semantic analysis on the teaching materials based on a large language model, obtain chapter-level information of the teaching materials, and construct a multimodal index based on the chapter-level information to generate a course structured model; In this embodiment, the teaching materials include text materials, voice materials, and 3D teaching materials. The steps of obtaining the teaching materials, performing chapter segmentation and semantic analysis on the teaching materials based on a large language model, obtaining chapter-level information of the teaching materials, and constructing a multimodal index based on the chapter-level information to generate a course structured model include: Acquire teaching materials, split the text materials into chapters based on the large language model, and obtain a list of chapter text segments of the teaching materials; Performing semantic analysis on the chapter text segment list segment by segment based on the large language model to obtain keywords and semantic summaries of each chapter text segment of the teaching material, and generating chapter-level information based on the keywords and semantic summaries of the chapter text segments and the chapter text segment list; Numbering the chapter-level information, and constructing multimodal index entries between the chapter-level information and text materials, voice materials, and 3D teaching materials based on the numbers; Based on the chapter level information and multimodal index entries, a course structured model is constructed.

[0025] In this embodiment, the system first obtains the course materials uploaded by the teacher, which include text materials (such as PPT slides and lecture documents), voice materials (audio files of explanations), and 3D teaching materials (such as models or animation assets). When "obtaining teaching materials", the above multimodal resources are stored in the object storage, and the metadata of each record is registered in the resource management module. Then, the pre-tuned chapter splitting engine based on the large language model is called to perform boundary detection and semantic segmentation on the text materials segment by segment, and output a "chapter text segment list" (each chapter segment contains a set of continuous text paragraphs and their source identifiers); then, the semantic analyzer of the same large language model is used to perform keyword extraction and summary generation on each chapter text segment in turn, obtain the core term set and brief topic description of each chapter segment, and jointly construct chapter hierarchical information (including chapter number, parent-child relationship and paragraph range) based on the "chapter text segment list", "keyword set" and "semantic summary". On this basis, the system numbers the chapter-level information and associates it with the corresponding text material paragraphs, audio material timestamps, and 3D teaching material resource identifiers to generate a multimodal index entry for each chapter. The entry records the mapping relationship between "chapter number → text paragraph ID list", "audio timestamp interval" and "3D material ID". Finally, the platform summarizes all multimodal index entries in chapter order to generate a "course structure model" that can be used by downstream modules.

[0026] In this embodiment, the course structuring model not only retains the chapter tree structure and semantic overview, but also realizes the fine association of text, voice and 3D materials, serving as a unified data foundation for subsequent functional modules such as knowledge graph construction, digital human generation and intelligent question and answer.

[0027] The present invention uses a large language model to perform fine chapter segmentation and semantic analysis on text materials, obtains an accurate list of chapter text segments and their keywords and summaries, and ensures the logical integrity of the course framework; then numbers the chapter level information, and automatically establishes multimodal index entries for text, voice and 3D materials, realizing fine-grained association between multiple resources and chapters; finally, a unified course structured model is constructed through these index entries, which can provide a consistent and easily expandable data foundation for subsequent action synthesis, knowledge retrieval and interactive applications, thereby effectively improving the system's organization and management efficiency of multi-source teaching materials.

[0028] Step 102: extracting features from the course structured model based on the large language model to obtain a semantic feature vector, and constructing a course knowledge graph based on the semantic feature vector and the chapter-level information; In this embodiment, the feature extraction of the course structured model based on the large language model to obtain a semantic feature vector, and the construction of a course knowledge graph based on the semantic feature vector and the chapter level information include: Performing feature extraction on the course structured model based on the large language model to obtain a semantic feature vector; Creating knowledge nodes based on the semantic feature vector and the chapter level information, and calculating the semantic similarity between the knowledge nodes; The chapter parent-child relationship of the knowledge node is obtained based on the chapter level information, and the associated edges between the knowledge nodes are generated based on the semantic similarity and the chapter parent-child relationship to complete the construction of the knowledge graph.

[0029] In this embodiment, a pre-trained and fine-tuned large language model is loaded to perform deep semantic encoding on each chapter and its multimodal index entry in the generated course structured model, thereby extracting its semantic feature vector. Specifically, the large language model uses a self-attention mechanism to contextually fuse the chapter text, keywords, abstracts, and related 3D material descriptions, outputting a fixed-dimensional vector representation that fully captures the semantic connotation of each chapter. The system then maps each chapter's semantic feature vector and its corresponding chapter-level information into a knowledge node, which contains both conceptual entities at the "chapter" level and can be refined to the "knowledge point" level.

[0030] In this embodiment, after generating knowledge points, the semantic association strength is measured for any two nodes by calculating their vector cosine similarity, and node pairs that exceed a preset threshold are marked as "semantically related." Finally, the system establishes a "parent-child" structural edge for each knowledge node based on the parent-child relationship defined in the chapter hierarchy information. Based on the aforementioned semantic similarity judgment results, it adds "cross-chapter" or "same-level" association edges between nodes. This completes the construction of a course knowledge graph in the graph database that retains the course organizational structure while incorporating deep semantic connections, providing a solid data foundation for subsequent intelligent question-answering, learning path recommendations, and knowledge visualization.

[0031] In this embodiment, by utilizing a large language model to extract semantic feature vectors, deep semantic encoding is performed on the text of each chapter; knowledge nodes are created in combination with chapter-level information and the semantic similarity between nodes is calculated, which can accurately reflect the intrinsic relationship between concepts; then, association edges are jointly generated based on the parent-child relationship and similarity of the chapters. The knowledge graph finally constructed takes into account both the logical structure of the chapters and the semantic closeness between knowledge points, which is helpful for applications such as intelligent question answering, personalized recommendations, and learning path planning, and effectively enhances the integrity and interpretability of knowledge representation.

[0032] In this embodiment, after the knowledge graph is constructed, the following steps are further included: Collecting question and answer records from the user's learning behavior data, and classifying the question and answer records according to knowledge nodes to obtain question and answer data; Based on the question and answer data, the access frequency, answer accuracy rate and average response time of each knowledge node are counted to obtain statistical results; Based on the statistical results, the weight of each knowledge node and associated edge in the knowledge graph is dynamically updated using a preset weighted fusion model to dynamically optimize the knowledge graph.

[0033] In this embodiment, after the knowledge graph is constructed, the question and answer records generated by students during the question and answer interaction process are collected in real time through the collection module, and based on the predefined knowledge node index in the knowledge graph, each question and answer detail (including the question content, system answer, whether the answer is correct, and response time, etc.) is mapped to the corresponding node; then, the statistical analysis engine summarizes the mapped question and answer data by node dimension in a batch or streaming manner to calculate indicators such as access frequency, accuracy, and average response time; finally, the system calls the pre-designed weighted fusion model, and merges the above statistical results with the original graph weight according to the preset weight strategy, and dynamically adjusts the weight value of each knowledge node and its associated edges in the graph database, so as to realize real-time feedback and adaptive optimization of the knowledge graph on students' learning behavior, thereby continuously improving the accuracy of intelligent question and answer and the effect of personalized recommendation.

[0034] After the knowledge graph is constructed, the present invention continues to collect and classify users' question and answer records, and counts the access frequency, answer accuracy and average response time of each knowledge node to form learning behavior data; based on these statistical results, a preset weighted fusion model is used to dynamically adjust the weights of knowledge graph nodes and associated edges, so that the graph can reflect the user's mastery and weak links in real time, thereby realizing adaptive optimization of the knowledge graph, and then supporting more accurate knowledge push and personalized tutoring.

[0035] Step 103: Generate a digital human image based on 3D facial modeling technology, and generate a digital human expression sequence and a teaching action sequence based on the digital human image, preset voice data, the course structured model, and a preset teaching action library; In this embodiment, the method of generating a digital human image based on three-dimensional facial modeling technology, and generating a digital human expression sequence and a teaching action sequence based on the digital human image, preset voice data, the course structured model, and a preset teaching action library includes: generating a digital human image based on the course structured model and three-dimensional facial modeling technology; Performing speech frame analysis and prosody analysis on the speech data to obtain semantic features; Identifying teaching intentions based on the large language model and the course structured model, and generating semantic vectors for each teaching intention; The semantic features, the semantic vectors and the preset teaching action library are weightedly fused based on the half-cosine curve function to generate a digital human expression sequence and a teaching action sequence synchronized with the current semantic intention and semantic features.

[0036] In this embodiment, based on the chapter text, keywords, and associated 3D teaching material information in the aforementioned course structured model, 3D facial modeling technology (such as a 3DMM-based face reconstruction algorithm) is first used to reconstruct a digital human facial geometry mesh with 0.1mm accuracy. This is then combined with expression-driven parameters to generate a complete digital human image. Next, the voice data provided by the teacher is subjected to voice frame segmentation and prosody analysis to extract voice features such as phoneme boundaries, stress positions, and speech rate variations to reflect the rhythm and emotional intonation of the lecture. The system then combines the chapter text in the course structured model with the large language model, identifying the teaching intent contained in each chapter or key sentence through a self-attention mechanism and outputting corresponding semantic vectors as content guidance for action generation. Finally, the system uses a half-cosine curve function to perform a weighted fusion of the aforementioned voice features, teaching intent semantic vectors, and facial expression and gesture action segments from a pre-collected and labeled teaching action library. This generates digital human expression and teaching action sequences that are both strictly aligned with the speech prosody and accurately convey the teaching intent. This enables natural, coherent, and semantically driven dynamic demonstrations of the digital human in different teaching scenarios.

[0037] In this embodiment, a high-precision digital human image is generated based on a course structured model and three-dimensional facial modeling technology, ensuring the consistency of the virtual teacher's appearance with the teaching materials. Voice features are obtained by performing voice frame segmentation and rhythm analysis on the voice data, and combined with the teaching intention semantic vector identified by the large language model, a half-cosine weighted fusion algorithm is used to fuse voice, text and action templates. The generated expression sequences and action sequences are highly consistent with the explanation rhythm and can accurately present the teaching intention, breaking through the limitations of a fixed action library and achieving a more natural, coherent and semantically driven interactive digital human demonstration.

[0038] Step 104: Based on the digital human image, the digital human expression sequence and the teaching action sequence, the digital human is embedded in the 3D teaching scene model, and a digital course is synthesized based on the 3D teaching scene model, the course structured model and the course knowledge graph.

[0039] In this embodiment, a pre-built 3D teaching scene model, including various visual elements such as classroom settings, experimental equipment, or sample objects, is first loaded into a 3D rendering engine (such as Unity or Unreal Engine). Subsequently, the digital human image, along with the corresponding expression sequence and teaching action sequence, generated by the 3D facial modeling and action synthesis module, is jittered into the scene in a timed manner. Skeletal skinning and facial drive parameters drive its real-time demonstration in the virtual environment. Next, the camera path and shot switching strategy within the scene are automatically configured based on the chapter sequence in the course structured model and the associated logic of each knowledge node in the knowledge graph to ensure the coherence and emphasis of the digital human's explanation. At each knowledge node or chapter node, the corresponding 3D teaching model (such as a neural network structure or mechanical principle diagram) is automatically triggered for local interactive display, and the presentation duration and camera focus are dynamically adjusted based on the node weight information in the graph. Finally, the rendering engine synthesizes all chapter segments into a high-definition video with subtitles and interactive annotations, outputting a complete digital course file. This achieves a deep integration of the digital human, 3D scene, and knowledge model, while ensuring the structured presentation and visual interaction of the teaching content.

[0040] The embodiment of the present invention also provides a digital course generation device based on a large language model, comprising: a data aggregation layer, an intelligent agent training layer, a digital human interaction layer, and an application service layer; The data aggregation layer is used to obtain teaching materials, perform chapter segmentation and semantic analysis on the teaching materials based on a large language model, obtain chapter-level information of the teaching materials, and construct a multimodal index based on the chapter-level information to generate a course structured model; The agent training layer is used to extract features from the course structured model based on the large language model, obtain semantic feature vectors, and construct a course knowledge graph based on the semantic feature vectors and the chapter level information; The digital human interaction layer is used to generate a digital human image based on three-dimensional facial modeling technology, and to generate a digital human expression sequence and a teaching action sequence based on the digital human image, preset voice data, the course structured model and a preset teaching action library; The application service layer is used to embed the digital human into the 3D teaching scene model based on the digital human image, the digital human expression sequence and the teaching action sequence, and synthesize the digital course based on the 3D teaching scene model, the course structured model and the course knowledge graph.

[0041] In this embodiment, the data aggregation layer is used for teaching materials including text materials, voice materials, and 3D teaching materials. The acquisition of teaching materials, the chapter segmentation and semantic analysis of the teaching materials based on the large language model, the acquisition of chapter-level information of the teaching materials, and the construction of a multimodal index based on the chapter-level information to generate a course structured model include: Acquire teaching materials, split the text materials into chapters based on the large language model, and obtain a list of chapter text segments of the teaching materials; Performing semantic analysis on the chapter text segment list segment by segment based on the large language model to obtain keywords and semantic summaries of each chapter text segment of the teaching material, and generating chapter-level information based on the keywords and semantic summaries of the chapter text segments and the chapter text segment list; Numbering the chapter-level information, and constructing multimodal index entries between the chapter-level information and text materials, voice materials, and 3D teaching materials based on the numbers; Based on the chapter level information and multimodal index entries, a course structured model is constructed.

[0042] In this embodiment, the agent training layer is used to extract features from the course structured model based on the large language model, obtain semantic feature vectors, and construct a course knowledge graph based on the semantic feature vectors and the chapter-level information, including: Performing feature extraction on the course structured model based on the large language model to obtain a semantic feature vector; Creating knowledge nodes based on the semantic feature vector and the chapter level information, and calculating the semantic similarity between the knowledge nodes; The chapter parent-child relationship of the knowledge node is obtained based on the chapter level information, and the associated edges between the knowledge nodes are generated based on the semantic similarity and the chapter parent-child relationship to complete the construction of the knowledge graph.

[0043] In this embodiment, the digital human interaction layer is used to generate a digital human image based on three-dimensional facial modeling technology, and to generate a digital human expression sequence and a teaching action sequence based on the digital human image, preset voice data, the course structured model, and a preset teaching action library, including: generating a digital human image based on the course structured model and three-dimensional facial modeling technology; Performing speech frame analysis and prosody analysis on the speech data to obtain semantic features; Identifying teaching intentions based on the large language model and the course structured model, and generating semantic vectors for each teaching intention; The semantic features, the semantic vectors and the preset teaching action library are weightedly fused based on the half-cosine curve function to generate a digital human expression sequence and a teaching action sequence synchronized with the current semantic intention and semantic features.

[0044] In this embodiment, the agent training layer is further used to: Collecting question and answer records from the user's learning behavior data, and classifying the question and answer records according to knowledge nodes to obtain question and answer data; Based on the question and answer data, the access frequency, answer accuracy rate and average response time of each knowledge node are counted to obtain statistical results; Based on the statistical results, the weight of each knowledge node and associated edge in the knowledge graph is dynamically updated using a preset weighted fusion model to dynamically optimize the knowledge graph.

[0045] In this embodiment, the digital course generation system adopts a four-layer B / S architecture, in which the data aggregation layer is mainly responsible for the acquisition and preprocessing of course materials - teachers upload text materials (PPT, lecture notes), voice materials (audio files) and 3D teaching materials to the cloud or local server through a browser or local client. The data cleaning module in the aggregation layer calls the fine-tuned large language model to perform chapter splitting and semantic analysis on the text materials, outputs a "chapter text segment list", keywords and summaries, and combines the voice frame timestamp and 3D material identifier to build a multimodal index, and finally generates a course structured model; the intelligent agent training layer extracts features from the structured model based on the same large language model, obtains the semantic feature vector of each chapter, and creates knowledge nodes in the graph database based on chapter level information, calculates the semantic similarity and parent-child relationship between nodes, and dynamically constructs and optimizes the course knowledge graph; the digital human interaction layer uses the course structured model and three-dimensional facial modeling technology to generate a high-precision digital human image, combining voice rhythm features and LLM The identified teaching intention vector is used to generate expression sequences and teaching action sequences synchronized with semantics and voice through the half-cosine weighted fusion algorithm. The digital human is then embedded in the 3D teaching scene model and rendered in collaboration with the knowledge graph and structured model to output digital course videos with subtitles and interactive annotations. The application service layer ultimately seamlessly connects service modules such as the intelligent body Q&A square, teacher preparation toolbox, and student learning path recommendation engine with the above generation process, allowing teachers to easily create, manage, and publish digital courses. Students can also initiate questions and answers in real time during the learning process, view learning progress, and obtain personalized learning path recommendations.

[0046] In this embodiment, the system collaboratively completes the digital course generation according to a four-layer architecture: First, in the data aggregation layer, the teacher uploads the "Machine Learning" PPT courseware through the browser, and the platform automatically calls the large language model to perform chapter splitting and semantic analysis on the PPT text material, extracts the title, keywords and abstract of each chapter, and combines the corresponding 3D teaching materials and voice scripts to build a multimodal index and course structured model; then, entering the intelligent agent training layer, the system calls LLM based on the structured model to perform in-depth semantic feature extraction, generates semantic feature vectors for each chapter, constructs a mind-map knowledge graph, and relies on the hybrid question-answering model to generate two query subsets of "teaching cases" and "ideological and political cases" in the knowledge base; then, in the digital human interaction At the interaction layer, the platform reconstructs the image of the digital human based on 3D facial modeling technology, and combines the course structured model with 3D models such as the neural network structure represented by the knowledge graph. It fuses the semantic vector of teaching intention and the rhythmic features of speech through the semi-cosine weighted fusion algorithm, triggers dynamic actions such as "indicative gestures" to generate expressions and body sequences, and renders and synthesizes explanation videos with subtitles in real time; finally, at the application service layer, when a student asks a question, the intelligent agent simultaneously retrieves the exclusive course knowledge base and the large model pre-training data and outputs the answer; if the teacher marks it as an error, the feedback is injected into the intelligent agent training layer through the recall-correction-retraining process, so that the knowledge base can be dynamically iterated and updated, thereby improving the answer accuracy in a closed loop and ensuring the continuous optimization of the course intelligent agent and the teaching adaptability.

[0047] In this embodiment, the digital course generation system is also a course intelligent teaching assistant system. Student users can initiate various types of intelligent question-and-answer services such as "AI guidance", "AI learning assistance", "AI training assistance" or "AI learning extension" at any time through the front-end learning interface. The interactive prompt words or intelligent bodies are independently created and maintained by teachers in the background; the system associates courseware, teaching videos, generated teaching cases with ideological and political cases, prompt words and intelligent body resources according to the multi-layer chapter structure of the course, and students can quickly retrieve and call corresponding content in any chapter dimension; the question-and-answer module supports three modes: exclusive data question and answer (course knowledge base only), hybrid data question and answer (course knowledge base and large model pre-training data) and exclusive question and answer based on chapter intelligent bodies to meet learning needs of different depth and breadth; in addition, the system provides learning plan formulation and resource recommendation functions, and automatically generates personalized learning paths based on students' knowledge mastery; students can also conduct online evaluation of their own learning effects at any time to grasp their learning progress and weak links.

[0048] In this embodiment, the platform's "Intelligent Agent Plaza" allows administrators or teachers to publish intelligent agents related to learning, courses, and even life. It supports access to third-party intelligent agents through web links or SDK parameters, and AI can automatically generate introductions and configure access permissions. Teachers can also create and manage prompt word templates for each chapter, which also supports AI-generated introductions and status settings, and centrally initiate prompt word or agent-guided Q&A in the AI workbench. The system has a built-in courseware document training function: after students or teachers upload untrained files and choose to train, they can build a dedicated course knowledge base in the large model. If the training results are not ideal, they can be recalled and modified and retrained at any time. At the same time, the platform can automatically generate a mind map-style course map based on the chapter level and related data, and provide learning record analysis and map display, learning path recommendation, and problem tracking services. Teachers and administrators can also query all question and answer conversation records of students in the class for teaching effectiveness evaluation and personalized tutoring. The online learning module supports breakpoint resumption and material download. The learning interface intuitively presents the course structure in the form of a mind map, helping students to sort out the knowledge context. In addition, the system integrates intelligent assisted programming functions to provide students with one-stop programming support such as code generation, parsing, optimization, error detection and formatting, greatly improving learning efficiency and experience.

[0049] In this embodiment, teachers can also quickly create and manage courses through the system: First, the teacher enters the course name and core keywords in the "Course Management" interface, and the system calls the large language model to automatically generate a course introduction, and supports setting multi-level access rights such as students, teachers or administrators; then, the teacher can quickly build a course chapter system by manually adding by level, importing XMind files, AI one-click generation of chapter structure, or automatically extracting chapters directly from uploaded PPT / documents, and use drag-and-drop in the chapter management panel to adjust the superior-subordinate relationship, upgrade or downgrade nodes, show / hide chapters, edit titles, or batch delete useless nodes; under each chapter, teachers can upload or associate teaching materials in various formats such as documents, videos, compressed packages, etc. to build a resource library, and use natural language input to generate teaching cases and ideological and political cases. The generated content can be freely output or apply preset templates; the resource management function supports uploading, editing, moving, downloading and pinning of multi-level materials to achieve courseware reuse and unified management; at the same time Teachers can centrally configure multiple intelligent bodies in the AI workbench, including allocating access rights by chapter granularity or importing third-party intelligent bodies, and set their display and calling strategies in the "Intelligent Body Plaza"; the platform also allows multiple teachers to be invited to jointly write courses, and to invite students to join in batches through invitation codes, student numbers or list imports, and to set personalized access and disclosure strategies for different classes; in addition, this system also includes a one-click teaching assistant toolbox, and teachers can directly call teaching cases, ideological and political cases, intelligent question setting, intelligent assessment, video generation and prompt word templates in the "one-click teaching assistant toolbox" to greatly improve preparation efficiency and teaching quality.

[0050] In this embodiment, a student-oriented intelligent course learning and tutoring system has been constructed, integrating multi-level course content management, intelligent agent Q&A, personalized learning support, and AI teaching assistant tools. The system supports teachers creating course structures through an AI workbench. Teachers can manually add chapters, import mind maps, automatically extract chapters based on PowerPoint presentations, or use AI to generate course chapters with one click, flexibly building a multi-level course system. Each chapter can be associated with teaching resources such as courseware, videos, teaching cases, and ideological and political case studies. It also supports natural language generation of case content and output according to templates. During the learning process, students can conduct AI Q&A based on the courseware content. Three modes are supported: dedicated data Q&A, hybrid data Q&A, and chapter-based agent Q&A, comprehensively enhancing the depth of course understanding and interactive experience. The system also supports students to develop learning plans, conduct self-assessment, generate learning maps, analyze learning records, recommend learning paths, and query personal and class learning Q&A records. Teachers can create and manage prompts and agents, configure third-party access methods, control their display and access rights, and use the courseware training function to build a dedicated course knowledge base. The platform also provides an "Intelligent Platform" where students can access intelligent agents related to their daily lives and studies, providing around-the-clock tutoring services. Furthermore, the system integrates one-stop intelligent assisted programming capabilities, covering tasks such as code generation, parsing, optimization, error detection, and formatting, empowering students to enhance their interdisciplinary skills. This solution effectively achieves a deep integration of course content and intelligent interaction, providing systematic support for personalized learning and intelligent teaching.

[0051] In an embodiment of the present invention, a terminal device is also provided, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the above-mentioned digital course generation method based on the large language model is implemented.

[0052] In an embodiment of the present invention, a computer-readable storage medium is also provided. The computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the above-mentioned digital course generation method based on the large language model.

[0053] For example, a computer program may be divided into one or more modules, one or more of which are stored in a memory and executed by a processor to implement the present invention. One or more modules may be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program in a terminal device.

[0054] The terminal device may be a computing device such as a desktop computer, laptop, PDA, or cloud server. The terminal device may include, but is not limited to, a processor, memory, and display. Those skilled in the art will appreciate that the aforementioned components are merely examples of terminal devices and do not constitute a limitation of the terminal device. The terminal device may include more or fewer components, or a combination of certain components, or different components. For example, the terminal device may also include input / output devices, network access devices, buses, and the like.

[0055] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting various parts of the entire terminal device using various interfaces and lines.

[0056] The memory can be used to store computer programs and / or modules. The processor implements various functions of the terminal device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area. The program storage area can store an operating system and at least one application required for a function (such as a sound playback function, a text conversion function, etc.); the data storage area can store data generated based on the use of the mobile phone (such as audio data, text message data, etc.). In addition, the memory can include high-speed random access memory and non-volatile memory, such as a hard disk, internal memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0057] Among them, if the module based on the digital course generation based on the large language model is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the process in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, it can implement the steps of the above-mentioned various method embodiments. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. Computer-readable media may include: any entity or device capable of carrying computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. Ordinary technicians in this field can understand and implement it without paying creative work.

[0058] The specific embodiments described above further illustrate the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A digital course generation method based on a large language model, characterized in that: include: Acquire teaching materials, perform chapter segmentation and semantic analysis on the teaching materials based on a large language model, obtain chapter-level information of the teaching materials, and construct a multimodal index based on the chapter-level information to generate a course structured model; Performing feature extraction on the course structured model based on the large language model to obtain a semantic feature vector, and constructing a course knowledge graph based on the semantic feature vector and the chapter level information; Generate a digital human image based on three-dimensional facial modeling technology, and generate a digital human expression sequence and a teaching action sequence based on the digital human image, preset voice data, the course structured model and a preset teaching action library; Based on the digital human image, the digital human expression sequence and the teaching action sequence, the digital human is embedded in the 3D teaching scene model, and the digital course is synthesized based on the 3D teaching scene model, the course structured model and the course knowledge graph.

2. A digital course generation method based on a large language model as claimed in claim 1, characterized in that: The teaching materials include text materials, voice materials, and 3D teaching materials. The obtaining of the teaching materials, performing chapter segmentation and semantic analysis on the teaching materials based on a large language model, obtaining chapter-level information of the teaching materials, and constructing a multimodal index based on the chapter-level information to generate a course structured model include: Acquire teaching materials, split the text materials into chapters based on the large language model, and obtain a list of chapter text segments of the teaching materials; Performing semantic analysis on the chapter text segment list segment by segment based on the large language model to obtain keywords and semantic summaries of each chapter text segment of the teaching material, and generating chapter-level information based on the keywords and semantic summaries of the chapter text segments and the chapter text segment list; Numbering the chapter-level information, and constructing multimodal index entries between the chapter-level information and text materials, voice materials, and 3D teaching materials based on the numbers; Based on the chapter level information and multimodal index entries, a course structured model is constructed.

3. A digital course generation method based on a large language model as described in claim 2, characterized in that: The step of extracting features from the course structured model based on the large language model to obtain a semantic feature vector, and constructing a course knowledge graph based on the semantic feature vector and the chapter level information includes: Performing feature extraction on the course structured model based on the large language model to obtain a semantic feature vector; Creating knowledge nodes based on the semantic feature vector and the chapter level information, and calculating the semantic similarity between the knowledge nodes; The chapter parent-child relationship of the knowledge node is obtained based on the chapter level information, and the associated edges between the knowledge nodes are generated based on the semantic similarity and the chapter parent-child relationship to complete the construction of the knowledge graph.

4. A digital course generation method based on a large language model as claimed in claim 3, characterized in that: The method of generating a digital human image based on the three-dimensional facial modeling technology, and generating a digital human expression sequence and a teaching action sequence based on the digital human image, preset voice data, the course structured model and a preset teaching action library, includes: generating a digital human image based on the course structured model and three-dimensional facial modeling technology; Performing speech frame analysis and prosody analysis on the speech data to obtain semantic features; Identifying teaching intentions based on the large language model and the course structured model, and generating semantic vectors for each teaching intention; The semantic features, the semantic vectors and the preset teaching action library are weightedly fused based on the half-cosine curve function to generate a digital human expression sequence and a teaching action sequence synchronized with the current semantic intention and semantic features.

5. A digital course generation method based on a large language model as claimed in claim 3, characterized in that: After the construction of the knowledge graph is completed, the following steps are also included: Collecting question and answer records from the user's learning behavior data, and classifying the question and answer records according to knowledge nodes to obtain question and answer data; Based on the question and answer data, the access frequency, answer accuracy rate and average response time of each knowledge node are counted to obtain statistical results; Based on the statistical results, the weight of each knowledge node and associated edge in the knowledge graph is dynamically updated using a preset weighted fusion model to dynamically optimize the knowledge graph.

6. A digital course generation system based on a large language model, characterized by: include: Data aggregation layer, intelligent agent training layer, digital human interaction layer and application service layer; The data aggregation layer is used to obtain teaching materials, perform chapter segmentation and semantic analysis on the teaching materials based on a large language model, obtain chapter-level information of the teaching materials, and construct a multimodal index based on the chapter-level information to generate a course structured model; The agent training layer is used to extract features from the course structured model based on the large language model, obtain semantic feature vectors, and construct a course knowledge graph based on the semantic feature vectors and the chapter level information; The digital human interaction layer is used to generate a digital human image based on three-dimensional facial modeling technology, and to generate a digital human expression sequence and a teaching action sequence based on the digital human image, preset voice data, the course structured model and a preset teaching action library; The application service layer is used to embed the digital human into the 3D teaching scene model based on the digital human image, the digital human expression sequence and the teaching action sequence, and synthesize the digital course based on the 3D teaching scene model, the course structured model and the course knowledge graph.

7. A digital course generation system based on a large language model as claimed in claim 6, characterized in that: The data aggregation layer is used for teaching materials including text materials, voice materials, and 3D teaching materials. The teaching materials are obtained, the teaching materials are divided into chapters and semantically analyzed based on the large language model, the chapter-level information of the teaching materials is obtained, and a multimodal index is constructed based on the chapter-level information to generate a course structured model, including: Acquire teaching materials, split the text materials into chapters based on the large language model, and obtain a list of chapter text segments of the teaching materials; Performing semantic analysis on the chapter text segment list segment by segment based on the large language model to obtain keywords and semantic summaries of each chapter text segment of the teaching material, and generating chapter-level information based on the keywords and semantic summaries of the chapter text segments and the chapter text segment list; Numbering the chapter-level information, and constructing multimodal index entries between the chapter-level information and text materials, voice materials, and 3D teaching materials based on the numbers; Based on the chapter level information and multimodal index entries, a course structured model is constructed.

8. A digital course generation system based on a large language model as claimed in claim 7, characterized in that: The agent training layer is used to extract features from the course structured model based on the large language model, obtain semantic feature vectors, and construct a course knowledge graph based on the semantic feature vectors and the chapter-level information, including: Performing feature extraction on the course structured model based on the large language model to obtain a semantic feature vector; Creating knowledge nodes based on the semantic feature vector and the chapter level information, and calculating the semantic similarity between the knowledge nodes; The chapter parent-child relationship of the knowledge node is obtained based on the chapter level information, and the associated edges between the knowledge nodes are generated based on the semantic similarity and the chapter parent-child relationship to complete the construction of the knowledge graph.

9. A digital course generation system based on a large language model as claimed in claim 8, characterized in that: The digital human interaction layer is used to generate a digital human image based on three-dimensional facial modeling technology, and to generate a digital human expression sequence and a teaching action sequence based on the digital human image, preset voice data, the course structured model, and a preset teaching action library, including: generating a digital human image based on the course structured model and three-dimensional facial modeling technology; Performing speech frame analysis and prosody analysis on the speech data to obtain semantic features; Identifying teaching intentions based on the large language model and the course structured model, and generating semantic vectors for each teaching intention; The semantic features, the semantic vectors and the preset teaching action library are weightedly fused based on the half-cosine curve function to generate a digital human expression sequence and a teaching action sequence synchronized with the current semantic intention and semantic features.

10. A digital course generation system based on a large language model as claimed in claim 8, characterized in that: The agent training layer is also used to: Collecting question and answer records from the user's learning behavior data, and classifying the question and answer records according to knowledge nodes to obtain question and answer data; Based on the question and answer data, the access frequency, answer accuracy rate and average response time of each knowledge node are counted to obtain statistical results; Based on the statistical results, the weight of each knowledge node and associated edge in the knowledge graph is dynamically updated using a preset weighted fusion model to dynamically optimize the knowledge graph.

Images