Modular content generation method and system for embedded teaching

Abstract

The invention discloses a modular content generation method and system for embedded teaching, and relates to the technical field of teaching semantic generation. The invention discloses a modularized content generation system for embedded teaching. The modularized content generation system comprises an embedded teaching module division module and an embedded teaching content generation module, according to the method, the problems of loose organization and poor reusability of the traditional embedded teaching content are solved through semantic analysis and structured processing of the original embedded teaching content; by introducing a skill dependence analysis mechanism, a clear preposed skill set and a clear output skill set are distributed for each initial teaching module, a teaching path network oriented to capability growth is constructed, and the definition of teaching logic and the controllability of a learning path are improved; learning fatigue evaluation is carried out by fusing multi-dimensional factors such as skill span, path complexity and cognitive load, and cognitive overload or content redundancy in the learning process of students is effectively avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of teaching semantic generation, and in particular to a modular content generation method and system for embedded teaching. Background Art

[0002] Existing embedded learning platforms often use a linear, fixed curriculum structure, making it difficult to dynamically adjust teaching content based on individual student differences. This leads to problems such as mismatched learning pace and uneven cognitive load. Furthermore, teaching resources often consist of static courseware, lab manuals, or scattered code examples, lacking structured organization and skill-related analysis, making them incapable of supporting modular reconstruction and personalized generation. Furthermore, traditional teaching models lack real-time feedback on students' learning status, resulting in a disconnect between teaching content and learners' abilities, impacting teaching effectiveness.

[0003] Therefore, there is an urgent need for a modular embedded teaching method and system that can dynamically optimize the teaching structure based on semantic analysis and skill modeling according to cognitive laws, and generate adaptive learning content based on student portraits. Summary of the Invention

[0004] The present invention aims to provide a modular content generation method and system for embedded teaching, which can dynamically optimize the teaching structure according to cognitive laws based on semantic analysis and skill modeling, and generate adaptive learning content in combination with student portraits.

[0005] A modular content generation method for embedded teaching, comprising the following steps: Obtain original teaching content related to embedded teaching; perform semantic analysis and structural processing on the original teaching content, and divide the original teaching content into several initial teaching modules; assign corresponding prerequisite skill sets and output skill sets to each initial teaching module; and sort all initial teaching modules according to the teaching sequence relationship; For any initial teaching module, a learning fatigue assessment is performed on the initial teaching module based on the pre-skill set and the output skill set to obtain a module fatigue assessment result; the module fatigue assessment result is compared with the preset fatigue threshold: if the module fatigue assessment result exceeds the preset fatigue threshold, the corresponding initial teaching module is split again to obtain at least two split embedded teaching modules; if the module fatigue assessment result is far below the preset fatigue threshold, multiple related initial teaching modules are selected and combined to obtain a combined embedded teaching module; Based on the step of comparing the module fatigue assessment result with the preset fatigue threshold, all initial teaching modules are traversed and compared, and all the obtained split embedded teaching modules and combined embedded teaching modules are renamed to obtain a number of embedded teaching modules; Generate corresponding embedded module learning content based on the student learning profile and the embedded teaching modules that need to be learned; learn based on the corresponding embedded module learning content and collect learning feedback data to update the student learning profile.

[0006] As a preferred technical solution of the present invention, the specific steps of allocating a corresponding prerequisite skill set and output skill set to each initial teaching module include: identifying the first n initial teaching modules and the last n initial teaching modules of the initial teaching module; Performing semantic analysis on the teaching content contained in the initial teaching module to obtain the knowledge set corresponding to the initial teaching module; Perform semantic analysis on the first n initial teaching modules to obtain the output basic ability set; perform semantic analysis on the last n initial teaching modules to obtain the prerequisite basic ability set; based on the dependency relationship between knowledge points, comprehensively analyze the knowledge set, prerequisite basic ability set and output basic ability set to obtain the prerequisite skill set and output skill set of the initial teaching module.

[0007] As a preferred technical solution of the present invention, the specific steps of performing learning fatigue assessment on the initial teaching module based on the pre-skill set and the output skill set include: Construct a skill learning span model based on prior knowledge; conduct a quantitative analysis of the skill differences between the pre-skill set and the output skill set to obtain the skill cognitive transition intensity value; analyze the average cognitive load carrying capacity of students based on big data, and determine the preset fatigue threshold based on the average cognitive load carrying capacity; Calculate the number and length of skill conversion paths included in the initial teaching module, and quantify the number and length of skill conversion paths as the learning fatigue index; The skill cognitive transition intensity value and learning fatigue index are fused to obtain the module fatigue assessment result.

[0008] As a preferred technical solution of the present invention, if the module fatigue assessment result exceeds the preset fatigue threshold, the specific steps include: Based on the skill span distribution between the pre-skill set and the output skill set, identify the skill transition mutation points in the initial teaching module; based on the skill transition mutation points, determine the cognitive overload sub-areas in the initial teaching module; A graph neural network is used to generate a learning content chain diagram for the cognitive overload sub-area. Cluster analysis is performed based on the learning content chain diagram to extract candidate sets of teaching sub-modules with longer operation chains. Identify the number of the shortest paths, the longest paths, and the cross paths of the learning content link graph, and perform load balancing splitting on the teaching submodule candidate set based on the identified number to obtain at least two split embedded teaching modules; The teaching sequence relationship of the split embedded teaching module is re-established, and the prerequisite skill set and output skill set corresponding to the split embedded teaching module are updated.

[0009] As a preferred technical solution of the present invention, if the module fatigue assessment result is far below the preset fatigue threshold, the specific steps include: Extract other initial teaching modules that have similar output skill sets to the current initial teaching module and whose module fatigue assessment results are far below the preset fatigue threshold as associated initial teaching modules; Identify the learning content link graphs of multiple related initial teaching modules and construct a module merging candidate pair corresponding to the current initial teaching module; Introducing a learning coherence scoring model; using the learning coherence scoring model to screen module merging candidate pairs and select module merging candidate pairs that meet the preset learning coherence requirements; Module merging candidate pairs that meet the preset learning coherence requirements are combined to obtain a combined embedded teaching module; the teaching sequence relationship of the combined embedded teaching module is re-established, and the corresponding prerequisite skill set and output skill set of the combined embedded teaching module are updated.

[0010] As a preferred technical solution of the present invention, the specific steps of generating corresponding embedded module learning content based on the student learning profile and the embedded teaching module to be learned include: Obtain student learning profiles; student learning profiles include current learning cognitive load, learning style preferences, and prior knowledge level; match semantic generation granularity based on prior knowledge level; generate semantic generation preferences based on learning style preferences; and match semantic generation density based on current learning cognitive load. The corresponding embedded module learning content of the embedded teaching module is generated based on the semantic generation granularity, semantic generation preference and semantic generation density.

[0011] A modular content generation system for embedded teaching, comprising: The embedded teaching module division module includes an initial module generation unit and an embedded teaching module generation unit; the initial module generation unit is used to obtain original teaching content related to embedded teaching; the original teaching content is semantically parsed and structured, and the original teaching content is divided into several initial teaching modules; each initial teaching module is assigned a corresponding prerequisite skill set and output skill set; all initial teaching modules are sorted according to the teaching sequence relationship; the embedded teaching module generation unit is used to perform learning fatigue evaluation on the initial teaching module based on the prerequisite skill set and the output skill set for any initial teaching module, and obtain a module fatigue evaluation result; the module fatigue evaluation result is compared with a preset fatigue threshold to judge: if the module fatigue evaluation result exceeds the preset fatigue threshold, the corresponding initial teaching module is split again to obtain at least two split embedded teaching modules; if the module fatigue evaluation result is far below the preset fatigue threshold, multiple related initial teaching modules are selected and combined to obtain a combined embedded teaching module; Based on the step of comparing the module fatigue assessment result with the preset fatigue threshold, all initial teaching modules are traversed and compared, and all the obtained split embedded teaching modules and combined embedded teaching modules are renamed to obtain a number of embedded teaching modules; The embedded teaching content generation module includes a module content generation unit; the module content generation unit is used to generate corresponding embedded module learning content based on the student learning profile and the embedded teaching module that needs to be learned; learning is carried out based on the corresponding embedded module learning content and learning feedback data is collected to update the student learning profile.

[0012] The present invention has the following advantages: 1. The present invention converts unstructured, multimodal teaching resources into analyzable and operable teaching modules through semantic analysis and structural processing of the original embedded teaching content, thereby solving the problems of loose organization and poor reusability of traditional embedded teaching content; by introducing a skill dependency analysis mechanism, a clear set of prerequisite skills and output skills is assigned to each initial teaching module, and a teaching path network oriented to ability growth is constructed, which improves the clarity of teaching logic and the controllability of learning paths; further, by integrating multidimensional factors such as skill span, path complexity and cognitive load to evaluate learning fatigue, it can dynamically determine whether a module is too heavy or too light, support intelligent splitting and reconstruction of modules, and effectively avoid cognitive overload or content redundancy in students' learning process.

[0013] 2. By introducing student learning portraits, including prior knowledge level, cognitive load and learning style preferences, the present invention can realize personalized dynamic generation of the granularity, density and form of embedded teaching content, so that the learning content is more in line with individual differences, and improve students' sense of participation and learning efficiency; at the same time, the feedback data collected during the learning process will continuously update the learning portrait, realize two-way adaptation and closed-loop optimization of teaching content and learning status, and promote the formation of a student-centered, task-driven, skill-oriented precise teaching mechanism. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is a schematic diagram of the structure of a modular content generation system for embedded teaching adopted in an embodiment of the present invention. DETAILED DESCRIPTION

[0015] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0016] Example 1, a modular content generation method for embedded teaching, comprising the following steps: Obtain original teaching content related to embedded teaching; original teaching content includes but is not limited to text, code, schematics, embedded development board simulation models, development board teaching videos, audio and 3D hardware models; obtaining original teaching content for embedded teaching usually requires gradual collection, screening and integration through multiple channels; for example, you can obtain systematic embedded course materials through the official teaching resource library of colleges and universities. These resources include detailed teaching outlines, course handouts, experimental instructions, schematic files and supporting source code; secondly, you can visit large technical communities and professional forums, such as electronic engineering, embedded development or single-chip microcomputer application websites, where you can find original teaching content, development board training documents, real project code and circuit diagrams shared by engineers or teachers. These contents are usually published in the form of posts, compressed packages or file attachments, with demonstration videos or development board simulation model files, in formats including Proteus simulation diagrams, Multisim project files, etc., which can be downloaded and used for local learning or simulation experiments; in addition, you can also obtain original code, circuit diagrams and 3D hardware models of development boards through open source hardware platforms.

[0017] Perform semantic analysis and structural processing on the original teaching content, and divide the original teaching content into several initial teaching modules; Specific steps: First, different types of content need to be uniformly converted into a processable semantic data format. For text content, such as teaching handouts, experimental instructions, source code comments, etc., the word segmentation, syntactic analysis and named entity recognition methods in natural language processing technology can be used to extract key terms, concept definitions and knowledge system architecture; for video and audio materials, voice recognition and image recognition must be performed first to transcribe the voice into text, and identify the operating steps, device labels or code demonstration screens in the screen, and extract teaching information synchronously with the voice content; circuit schematics and three-dimensional hardware models use image recognition or model analysis tools to extract components, connection relationships and hardware composition logic, and then convert them into processable data. Data describing hardware structure and functionality; identifying knowledge points and skill points within the parsed semantic content. Based on knowledge graph construction methods, information such as theoretical concepts of embedded systems (e.g., interrupts, timers, serial communication), programming techniques (e.g., register manipulation, peripheral drivers, bare-metal development), and hardware operations (e.g., circuit soldering, development board debugging, and simulation platform usage) are annotated within the semantic structure. Combined with actual project operational processes, practical skill points are further identified, such as timer-based LED flashing control and data loopback testing using serial debugging tools. This process incorporates both rule-based reasoning and the integration of deep learning models to identify implicit skill expressions within unstructured content. After forming a complete map of knowledge and skill points, the content needs to be initially structured and divided according to teaching logic. This division is based on the hierarchical relationship between knowledge and skills, the learning sequence, and the dependencies between content, and the corresponding unit boundaries are delineated. Ultimately, each initial teaching module should include a main knowledge point, several related skill points, and a complete practical goal, forming a basic unit structure with clear content, clear goals, and easy teaching organization.

[0018] Assign a corresponding prerequisite skill set and output skill set to each initial teaching module; the prerequisite skill set refers to the basic skills or prerequisite abilities that learners should have before learning a certain initial teaching module. These skills are the prerequisites for the smooth learning, understanding and completion of practical tasks of this unit; prerequisite skills may include basic programming syntax, basic circuit knowledge, development environment configuration, common tool usage methods, etc. Without these skills, learners may not be able to effectively follow up the teaching content, resulting in knowledge gaps or practical failures; the output skill set refers to the new skills or improved abilities that learners should master after completing a certain initial teaching module. These skills are the embodiment of the teaching objectives of the module, usually including the ability to understand specific knowledge, hands-on operation ability, problem-solving ability or actual control of a certain type of hardware module. Output skills can be basic (such as lighting an LED light) or complex (such as completing the master-slave communication mechanism based on the serial port protocol), depending on the teaching depth of the unit and the complexity of the task.

[0019] All initial teaching modules are sorted according to the teaching sequence relationship; in order to sort all initial teaching modules according to the teaching sequence relationship, it is necessary to construct a complete skill dependency graph; the graph uses each initial teaching module as a node, and the dependency relationship between the prerequisite skill sets and the output skill sets between modules as directed edges, thus forming a directed graph structure; by comparing the skill sets of each module, if the prerequisite skills of a module happen to be a subset of the output skills of another module or are completely included in it, it can be determined that the former should precede the latter in the teaching sequence; the constructed skill dependency graph not only describes the flow direction of the teaching logic, but also reveals the sequence between modules and the skill evolution path.

[0020] The specific steps for assigning corresponding prerequisite skill sets and output skill sets to each initial teaching module include: Identify the first n initial teaching modules and the last n initial teaching modules of the initial teaching module; the specific value of n is manually set by professional technicians; these modules represent the possible leading teaching content and subsequent connection content of the module respectively; constructing a local teaching window in this way can help the system understand the logical connection and dependency path between the module and the surrounding knowledge.

[0021] Conduct semantic analysis on the teaching content contained in the initial teaching module to obtain the knowledge set corresponding to the initial teaching module; use natural language processing technology to parse the text content, code comments, experimental steps and other information in the module, extract core knowledge units with semantic meaning, and construct the knowledge set of the module; the knowledge set can include theoretical concepts, operating procedures, interface mechanisms, hardware characteristics, etc., which are the basic corpus resources for subsequent skill reasoning.

[0022] Perform semantic analysis on the first n initial teaching modules to obtain the output basic ability set; perform semantic analysis on the last n initial teaching modules to obtain the prerequisite basic ability set; based on the dependency relationship between knowledge points, conduct a comprehensive analysis on the knowledge set, prerequisite basic ability set and output basic ability set to obtain the prerequisite skill set and output skill set of the initial teaching module; perform similar semantic analysis on the first n initial teaching modules of the current module one by one, and extract the set of ability units that these modules should have after completion, which is called the output basic ability set. This set represents the prerequisite knowledge and operational experience that learners of the current module may have; at the same time, perform semantic analysis on the last n modules to obtain the prerequisite ability set they rely on, that is, the basic skills required for the successful learning of these modules. This set can reflect the skill transfer role of the current module in the future teaching path and provide a basis for the setting of skill output.

[0023] The knowledge set of the current initial teaching module, the output basic competence set of the first n modules, and the pre-condition basic competence set of the last n modules are integrated and analyzed. Based on the logical dependencies between knowledge points and the skill development path map, the analysis identifies the minimum required competence units for the current module as the pre-condition skill set, and extracts the new skills that should be output after learning the current module as the output skill set. This comprehensive analysis ensures that the skill chain forms a continuous and effective transmission mechanism between teaching modules, thereby constructing a skill structure system that conforms to both cognitive laws and teaching logic.

[0024] For any initial teaching module, a learning fatigue assessment is performed on the initial teaching module based on the pre-skill set and the output skill set to obtain a module fatigue assessment result; the module fatigue assessment result is compared with the preset fatigue threshold: if the module fatigue assessment result exceeds the preset fatigue threshold, the corresponding initial teaching module is split again to obtain at least two split embedded teaching modules; if the module fatigue assessment result is far below the preset fatigue threshold, multiple related initial teaching modules are selected and combined to obtain a combined embedded teaching module; Based on the step of comparing the module fatigue assessment result with the preset fatigue threshold, all initial teaching modules are traversed and compared, and all the obtained split embedded teaching modules and combined embedded teaching modules are renamed to obtain a number of embedded teaching modules; The specific steps for assessing learning fatigue in the initial teaching module based on the prerequisite skill set and the output skill set include: Construct a skill learning span model based on prior knowledge; conduct a quantitative analysis of the skill differences between the pre-skill set and the output skill set to obtain the skill cognitive transition intensity value; analyze the average cognitive load carrying capacity of students based on big data, and determine the preset fatigue threshold based on the average cognitive load carrying capacity; The Skill Learning Span Model is used to measure the learning span of a teaching module at the knowledge level. This is the depth of knowledge and complexity of skill types required to reach the target skill from a student's existing skills. This model, typically driven by educational experts and domain knowledge graphs, encompasses multiple dimensions, including skill difficulty, knowledge dependency, operational abstraction, and cross-domain integration, to characterize the transition costs between different skills. Once constructed, this model will serve as a foundational tool for evaluating the cognitive leaps within a teaching module. Based on the skill gap between the pre-skill set and the output skill set in the initial teaching module, the skill learning span model is used to quantify the gap and obtain the skill cognitive transition strength value. The skill cognitive transition strength value represents the cognitive transformation intensity required to transition from pre-skill to output skill, that is, the number of new cognitive structures and operational models required by the learner. The larger the value, the greater the skill improvement span within the module, requiring learners to engage in deeper understanding, conceptual reconstruction, or skill integration, and thus being more prone to learning fatigue. Based on big data analysis, the average cognitive load carrying capacity of the student population is analyzed to determine the corresponding preset fatigue threshold. Cognitive load carrying capacity refers to the upper limit of new information or skills that students can effectively absorb, process and master within a unit of learning time or learning stage. This capacity is obtained by analyzing the learning behavior data of a large number of students under different skill difficulties and module structures, including completion time, failure rate, answer accuracy, number of operations, etc.; through statistical modeling and cluster analysis, the fatigue limit that students can generally withstand in a specific teaching environment can be derived, thereby setting a preset fatigue threshold for subsequent fatigue level judgment.

[0025] Calculate the number and length of skill conversion paths included in the initial teaching module, and quantify the number and length of skill conversion paths as the learning fatigue index; The skill conversion paths contained in the initial teaching module are calculated, including the number of paths and the length of the paths. The number of paths refers to the total number of all feasible learning paths from any skill in the prerequisite skill set to the output skill set, and the path length refers to the number of intermediate skill conversion steps that need to be completed on each path. These two indicators reflect the complexity of the skill structure and the depth of the cognitive process. The more paths and the longer the length, the more complex the skill transition chain within the module, and the longer the operational load and cognitive concentration time. The number of paths and the path length are combined and weighted to obtain the learning fatigue index, which is used to measure the potential learning fatigue caused by an overly long skill chain or overly dense paths.

[0026] The skill cognitive transition intensity value and learning fatigue index are integrated to obtain the module fatigue assessment result; The module fatigue assessment results are generated by integrating the skill cognitive transition intensity value with the learning fatigue index. This fusion method can employ weighted averaging, fuzzy reasoning, or machine learning models, with specific strategies flexibly adjusted based on the nature of the course. The module fatigue assessment result is a total value indicator used to determine whether the current module is too dense or the target span is too large. If the value exceeds the preset fatigue threshold, the module is at risk of overload and requires splitting and optimization. If the result is low, the module skills are too fragmented or repetitive, and merging them may be considered to improve learning efficiency and content coherence. This assessment mechanism makes teaching module design more scientific and adaptable to individual needs, effectively improving the learner experience and course implementation quality.

[0027] If the module fatigue assessment result exceeds the preset fatigue threshold, the specific steps include: Based on the skill span distribution between the pre-skill set and the output skill set, identify the skill transition mutation points in the initial teaching module; based on the skill transition mutation points, determine the cognitive overload sub-areas in the initial teaching module; Skill span distribution refers to the changes in complexity, connections, and continuity in skill transformation requirements across skill points within the initial teaching module. When the skill span at a certain location shows a sudden change in a local area (i.e., a significant increase in difficulty or dependencies), it is a skill transition mutation point. These mutation points often indicate areas of sharp increase in cognitive load and sudden learning difficulty, and require special attention. After identifying the mutation points of skill transition, it is necessary to further determine the cognitive overload sub-areas in the initial teaching module based on these mutation points; cognitive overload sub-areas refer to those teaching content segments that contain high-difficulty skill combinations, dense intersection of knowledge points, and complex operational logic; in these areas, students need to complete more skill conversion, information processing, and abstract reasoning, which can easily lead to learning fatigue; therefore, these sub-areas need to be separated from the overall module and used as separate research and split objects for further optimization.

[0028] A graph neural network is used to generate a learning content chain diagram for the cognitive overload sub-area. Cluster analysis is performed based on the learning content chain diagram to extract candidate sets of teaching sub-modules with longer operation chains. For the cognitive overload sub-areas mentioned above, graph neural network technology is used to structurally model the teaching content within them, generating a learning content link graph. This link graph uses skill points and knowledge points in the teaching content as nodes, and skill dependencies, concept references, and operation sequences as edges, expressing the logical relationships and learning paths between the various contents. The graph neural network not only considers the attribute characteristics of the nodes themselves but also learns the semantic associations and structural embeddings between nodes through a multi-layer message passing mechanism, forming a higher-dimensional representation of the learning path and providing basic data for subsequent content reorganization. After obtaining the learning content link graph, a cluster analysis is performed on the graph structure. A graph segmentation algorithm is used to identify sub-graph regions with similar structures and highly relevant content. These regions are then considered as candidate sets of teaching sub-modules. Candidate sub-modules are typically composed of one or more operation chains, each representing the operational process of gradually evolving from a basic skill to a specific target skill. By calculating the length and node density of these chains, the complexity of each sub-module can be assessed to determine whether it contains excessively long operation paths or too many concurrent skill points.

[0029] Identify the number of the shortest paths, the longest paths, and the cross paths of the learning content link graph, and perform load balancing splitting on the teaching submodule candidate set based on the identified number to obtain at least two split embedded teaching modules; Identify the shortest path, longest path and number of cross paths in the learning content link diagram; the shortest path reflects the minimum skill steps required to achieve basic learning goals in the module, the longest path reveals the deepest skill dependency chain that may exist in the module, and the number of cross paths reflects the degree of reuse and multi-path coupling between skills; based on these three indicators, the system performs load balancing splitting on the candidate sub-modules. This process splits the high-complexity area into at least two split embedded teaching modules with smaller cognitive span, shallower path depth and less skill intersection, so as to achieve a balanced distribution of skills within the module and reduce the cognitive burden of learners.

[0030] Re-establish the teaching sequence relationship of the split-embedded teaching modules after splitting, and update the prerequisite skill set and output skill set corresponding to the split-embedded teaching modules; reconstruct the teaching sequence of at least two split-embedded teaching modules; reconstruct the teaching path map according to the sequence of their original positions, skill dependency direction and knowledge development logic to ensure smooth connection between modules and natural skill progression; reallocate the prerequisite skill set and output skill set for each newly generated split-embedded teaching module so that it can independently adapt to personalized learning path planning, forming a controllable teaching unit with clear structure and moderate granularity.

[0031] If the module fatigue assessment result is far below the preset fatigue threshold, the specific steps include: "Far below the preset fatigue threshold" is a criterion used to determine whether the initial teaching module has too light a learning load or too small a teaching granularity. "Far below" is not an absolute value, but rather a quantitative deviation range relative to the threshold. This judgment is usually set by combining statistical methods with teaching cognitive models. The specific judgment method is as follows: The preset fatigue threshold is a critical value determined based on the average cognitive load capacity of the student population. It represents the upper limit of the reasonable cognitive stress that students can withstand to complete a module under normal learning conditions. Generally, this threshold is a range, not a single point value. Far below is usually quantified using the difference-to-proportion method. The difference-to-proportion method states that if the learning fatigue index for a module falls below a certain percentage of the fatigue threshold (e.g., 30% or 40%), it is considered far below. For example, if the preset fatigue threshold is 80, and a module assessment score of 45 is less than 30% of 80, or at least 24 points, it is considered far below.

[0032] Extract other initial teaching modules that have similar output skill sets to the current initial teaching module and whose module fatigue assessment results are far below the preset fatigue threshold as associated initial teaching modules; Extract other initial teaching modules that are similar to the output skill set of the current initial teaching module, and the fatigue assessment results of these modules are also in a low-load state, indicating that these modules are scattered in content, have small granularity, or have a light cognitive burden, which may cause interruption of learning rhythm or content redundancy; at this time, skill set similarity calculation, such as semantic similarity matching based on knowledge graph or vectorized embedding model, can be used to identify modules with similar skill output directions, and construct a set of associated initial teaching modules for the current module to prepare for subsequent content fusion.

[0033] Identify the learning content link graphs of multiple related initial teaching modules and construct module merging candidate pairs corresponding to the current initial teaching module; after identifying multiple related initial teaching modules, it is necessary to extract the learning content link graphs of these modules respectively; the learning content link graph is a directed graph structure constructed with the knowledge points, skill points, operation steps, etc. within the module as nodes and the logic or dependency relationship between them, which is used to represent the structured process of the content within the module; by comparing the link graph of the current module with the link graph of the related modules, several module merging candidate pairs can be combined, each pair representing two teaching modules that may have merging potential in terms of content logic, skill direction and learning path.

[0034] A learning coherence scoring model is introduced; the learning coherence scoring model is used to screen module merging candidate pairs and select module merging candidate pairs that meet the preset learning coherence requirements; a learning coherence scoring model is introduced to evaluate the consistency of teaching rhythm and cognitive process between module merging candidate pairs; the learning coherence scoring model is an evaluation tool constructed based on multi-dimensional indicators such as structural similarity, skill transfer fluency, and teaching goal consistency. Its function is to avoid abrupt learning rhythm or logical jumps caused by simple splicing of content; the learning coherence scoring model takes into account the conceptual continuity, operational connection and semantic transition smoothness between modules, and sets a coherence threshold to ensure that candidate pairs can only be further merged under the conditions of logical coherence and reasonable skill docking.

[0035] The construction of the learning coherence scoring model is to measure whether two or more teaching modules have a natural connection in terms of knowledge logic, skill transfer, task rhythm, etc., to avoid cognitive jumps or learning fragmentation after the modules are merged; the construction method of the model is first based on the structured feature extraction of the teaching module, and generates its semantic representation by vectorizing the knowledge point sequence, skill path diagram, operation flow chart, etc. of each module; on this basis, a deep matching network or a dual-tower semantic similarity model is used, and the two modules are used as input pairs to learn the learning coherence score between them through training; the training process is mainly supervised learning, and a group of module combinations annotated by teaching experts are selected. The data is used as the training set, where positive examples are module combinations that are considered to have natural teaching transitions and reasonable rhythms, while negative examples are combinations that are logically abrupt and have disconnected skills. The model gradually learns effective parameters for measuring teaching coherence by optimizing the gap between scores and manual scores. The training set sources mainly include historical course design plans, classic module pairs in the teaching case library, module combinations that perform well and fail in actual student learning path data, and a coherence evaluation control set constructed through manual scoring by the expert review system. The final scoring model can input any two teaching modules to be merged and output a coherence score to determine whether they are suitable for integration into a unified embedded teaching module.

[0036] Modules that meet the preset learning continuity requirements are combined into candidate pairs to obtain combined embedded teaching modules; the teaching sequence relationship of the combined embedded teaching modules is re-established, and the corresponding prerequisite skill sets and output skill sets of the combined embedded teaching modules are updated; the preset learning continuity requirements are manually set by professional technicians; All candidate pairs are screened using the learning coherence scoring model, and module combinations with substandard coherence scores are eliminated, while merging candidate pairs that meet the preset learning coherence requirements are retained. This process ensures that the new modules generated by subsequent mergers will not affect the learning effect due to content mixing, skill conflicts or cognitive rhythm disorder, and also ensures the natural connection of the teaching path and the consistency of the content. Module merging operations are performed based on the screened module merging candidate pairs to obtain combined embedded teaching modules. Each combined embedded teaching module should have stronger content integrity, task coherence and skill effectiveness in teaching functions, which not only improves teaching efficiency but also reduces the risk of fragmentation. After the module merger is completed, the teaching sequence relationship needs to be reconstructed to ensure that the merged combined embedded teaching module is still in the appropriate sequence position in the entire teaching path, and its prerequisite skill set should cover the prerequisite skill sets of all original sub-modules, and its output skill set should cover the skill outputs of all merged modules, so as to ensure that the combined embedded teaching module has reasonable input and output interfaces in the skill structure to support the smooth connection and progressive learning of subsequent modules.

[0037] Rename all the split embedded teaching modules and combined embedded teaching modules. This step means: within the system, the split embedded teaching modules obtained by splitting the initial teaching modules, or the combined embedded teaching modules obtained by recombining the initial teaching modules, and the initial teaching modules that have not been split or combined are integrated and unified as embedded teaching modules for subsequent teaching content generation operations; Specifically, the teaching sequence relationship corresponding to the split embedded teaching module, the teaching sequence relationship corresponding to the combined embedded teaching module, and the teaching sequence relationship corresponding to the initial teaching module are used as the teaching sequence relationship of the corresponding embedded teaching modules and are re-arranged.

[0038] Generate corresponding embedded module learning content based on the student's learning profile and the embedded teaching module that needs to be learned; learn based on the corresponding embedded module learning content and collect learning feedback data to update the student's learning profile; Based on the student learning profile and the embedded teaching modules that need to be learned, the specific steps for generating the corresponding embedded module learning content include: Obtain student learning profiles; student learning profiles include current learning cognitive load, learning style preferences, and prior knowledge level; match semantic generation granularity based on prior knowledge level; generate semantic generation preferences based on learning style preferences; and match semantic generation density based on current learning cognitive load. The student learning portrait is a dynamic feature model constructed with the individual student as the center, mainly including three core dimensions: current learning cognitive load, learning style preference and prior knowledge level. Among them, learning cognitive load reflects the student's psychological carrying capacity for knowledge absorption, skill mastery and operation execution in a certain learning stage; learning style preference reflects its preference in information reception, processing and expression, such as tendency towards graphical display, text explanation or interactive practice; prior knowledge level describes the knowledge reserve and skill mastery depth of the student before entering a certain teaching module; the acquisition of learning portrait usually depends on behavioral data, answer records, task completion status, learning feedback and periodic assessment results during the learning process.

[0039] Generate corresponding embedded module learning content of the embedded teaching module based on semantic generation granularity, semantic generation preference and semantic generation density; After understanding the students' prior knowledge level, it is necessary to match the semantic generation granularity based on this level; semantic generation granularity refers to the degree of refinement or the depth of information decomposition when generating learning content, and its role is to adjust the complexity and information density of teaching materials; if students' prior knowledge is weak, it is necessary to adopt a fine-grained strategy to break down a complex knowledge point into multiple basic concepts and attach a large amount of guiding explanations; if students already have a high knowledge base, coarse-grained expressions can be used to simplify descriptions, focus on key knowledge and practical goals, improve learning efficiency and challenge, and thus achieve personalized matching.

[0040] Determine semantic generation preferences based on students' learning style preferences; semantic generation preferences refer to the generation tendency in the form of content presentation, which ensures that the generated learning content can conform to the way students are most easily accepted and processed; for example, students who prefer visual learning can generate graphical content such as flow charts, structure diagrams, circuit diagrams, etc.; students who prefer language-based learning can generate detailed text descriptions, concept definitions and process explanations; and students who prefer hands-on operations can highlight operation steps, real cases and interactive simulation experiments. This preference matching can significantly improve students' learning motivation and acceptance.

[0041] It is also necessary to match the semantic generation density according to the current level of learning cognitive load; semantic generation density refers to the amount of information or the level of complexity in a unit of content, which determines the compactness of the learning content in logical advancement; if students are currently in a high cognitive load state, the semantic density should be appropriately reduced, and cognitive pressure should be reduced by inserting blank transition points, supplementing examples, simplifying language, etc.; on the contrary, when cognitive resources are sufficient, the density can be increased, knowledge points can be integrated, and expanded content can be added or the difficulty of questions can be increased to maintain challenge and efficiency.

[0042] By integrating the three personalized adjustment parameters of semantic generation granularity, semantic generation preference and semantic generation density, the generation engine is called to dynamically generate content for the knowledge points, skill points and operation processes in the embedded teaching module, and output embedded module learning content suitable for the student; these contents will be organized in a form suitable for their learning style, according to a reasonable complexity and structural hierarchy, to ensure that they cover both teaching objectives and are in line with the students' current cognitive status and growth path, thereby effectively supporting the implementation of personalized embedded learning.

[0043] After students begin learning the corresponding embedded module content, the learning process is monitored and recorded in real time. This process includes data collection from multiple dimensions, including learning task start time, completion time, number of interruptions, video viewing duration, interactive operation frequency, code writing and debugging behavior, and simulation experiment results. This aims to comprehensively capture students' actual learning behavior trajectory within the module. Secondly, students are guided to complete key test questions, experimental operation questions, and reflective questions embedded in the module. Their knowledge mastery and skill application ability are judged based on their answer accuracy, error types, and submission delay. The system also extracts features of error frequency and debugging time during the learning process to reflect cognitive load. The system then inputs this feedback data into the student profile update model. Combined with the existing learning profile, it dynamically adjusts and updates the student's current cognitive level, skill mastery, behavioral habits, and emotional tendencies, thereby forming a more accurate learning profile. This updated profile will serve as an important basis for the next round of learning content generation and teaching strategy recommendation, ensuring that the teaching process continuously adapts to the student's development status and achieving closed-loop optimization of personalized embedded teaching.

[0044] For example, in a university's intelligent electronics course, a basic peripheral control teaching module was built around the STM32 development board. The main content included GPIO configuration, LED lighting experiments, key input reading, and simple interrupt control. A student's learning profile showed that their current cognitive load was above average. Their prior knowledge mainly consisted of basic C language syntax and digital electronics principles, but they had no systematic exposure to embedded development. Their learning style preferred a combination of text and graphics, with a preference for hands-on learning. According to the portrait, first of all, the semantic generation granularity of the teaching module is set to medium-fine granularity based on the student's prior knowledge level, that is, the GPIO control process is broken down into steps, such as explaining the sub-steps such as clock enable, port configuration, and pin direction setting respectively; the semantic generation preference adopts a graphic process priority strategy, with supporting development board pin diagrams, circuit connection diagrams and register configuration flow charts; the semantic generation density is controlled at a medium-low level, and appropriate comments and common error prompts are inserted between key steps to avoid cognitive overload.

[0045] The final personalized embedded learning content includes: a video explanation starting with the real wiring diagram of the development board, which gradually guides the student to configure the project in the Keil environment, complete the GPIO initialization settings in CubeMX, and manually supplement the LED control statements after exporting the code; the accompanying text materials show the meaning of each register bit in the form of a chart, and are accompanied by visual debugging screenshots of each stage; after the learning content is put into use in actual teaching, by collecting the student's code writing accuracy, experiment completion time and learning behavior trajectory data generated by the system, the feedback is used to optimize the generation strategy of subsequent interrupt and timer modules, and finally a personalized teaching path from fixed modules to dynamic content generation is achieved, which demonstrates the feasibility and adaptability of this method in actual embedded teaching.

[0046] Example 2, a modular content generation system for embedded teaching, see Figure 1 Shown, including: The embedded teaching module division module includes an initial module generation unit and an embedded teaching module generation unit; the initial module generation unit is used to obtain original teaching content related to embedded teaching; the original teaching content is semantically parsed and structured, and the original teaching content is divided into several initial teaching modules; each initial teaching module is assigned a corresponding prerequisite skill set and output skill set; all initial teaching modules are sorted according to the teaching sequence relationship; the embedded teaching module generation unit is used to perform learning fatigue evaluation on the initial teaching module based on the prerequisite skill set and the output skill set for any initial teaching module, and obtain a module fatigue evaluation result; the module fatigue evaluation result is compared with a preset fatigue threshold to judge: if the module fatigue evaluation result exceeds the preset fatigue threshold, the corresponding initial teaching module is split again to obtain at least two split embedded teaching modules; if the module fatigue evaluation result is far below the preset fatigue threshold, multiple related initial teaching modules are selected and combined to obtain a combined embedded teaching module; Based on the step of comparing the module fatigue assessment result with the preset fatigue threshold, all initial teaching modules are traversed and compared, and all the obtained split embedded teaching modules and combined embedded teaching modules are renamed to obtain a number of embedded teaching modules; The embedded teaching content generation module includes a module content generation unit; the module content generation unit is used to generate corresponding embedded module learning content based on the student learning profile and the embedded teaching module that needs to be learned; learning is carried out based on the corresponding embedded module learning content and learning feedback data is collected to update the student learning profile.

[0047] It should be understood that those skilled in the art may make improvements or modifications based on the above description, and all such improvements and modifications shall fall within the scope of protection of the appended claims. Any portion of this specification not described in detail is prior art known to those skilled in the art.

Claims

1. A modular content generation method for embedded teaching, characterized in that: The following steps are involved: Obtain original teaching content related to embedded teaching; perform semantic analysis and structural processing on the original teaching content, and divide the original teaching content into several initial teaching modules; assign corresponding prerequisite skill sets and output skill sets to each initial teaching module; and sort all initial teaching modules according to the teaching sequence relationship; For any initial teaching module, the learning fatigue assessment of the initial teaching module is performed based on the pre-skill set and the output skill set to obtain the module fatigue assessment result; Compare the module fatigue assessment result with the preset fatigue threshold and make a judgment: if the module fatigue assessment result exceeds the preset fatigue threshold, the corresponding initial teaching module is split again to obtain at least two split embedded teaching modules; If the module fatigue assessment result is far below the preset fatigue threshold, multiple associated initial teaching modules are selected and combined to obtain a combined embedded teaching module; Based on the step of comparing the module fatigue assessment result with the preset fatigue threshold, all initial teaching modules are traversed and compared, and all the obtained split embedded teaching modules and combined embedded teaching modules are renamed to obtain a number of embedded teaching modules; Generate corresponding embedded module learning content based on students' learning profiles and the embedded teaching modules they need to learn; Study based on the corresponding embedded module learning content and collect learning feedback data to update the student learning profile.

2. A modular content generation method for embedded teaching according to claim 1, characterized in that: The specific steps for assigning corresponding prerequisite skill sets and output skill sets to each initial teaching module include: identifying the first n initial teaching modules and the last n initial teaching modules of the initial teaching module; Performing semantic analysis on the teaching content contained in the initial teaching module to obtain the knowledge set corresponding to the initial teaching module; Perform semantic analysis on the first n initial teaching modules to obtain the output basic ability set; perform semantic analysis on the last n initial teaching modules to obtain the prerequisite basic ability set; based on the dependency relationship between knowledge points, comprehensively analyze the knowledge set, prerequisite basic ability set and output basic ability set to obtain the prerequisite skill set and output skill set of the initial teaching module.

3. A modular content generation method for embedded teaching according to claim 2, characterized in that: The specific steps for assessing learning fatigue in the initial teaching module based on the prerequisite skill set and the output skill set include: Construct a skill learning span model based on prior knowledge; conduct a quantitative analysis of the skill differences between the pre-skill set and the output skill set to obtain the skill cognitive transition intensity value; analyze the average cognitive load carrying capacity of students based on big data, and determine the preset fatigue threshold based on the average cognitive load carrying capacity; Calculate the number and length of skill conversion paths included in the initial teaching module, and quantify the number and length of skill conversion paths as the learning fatigue index; The skill cognitive transition intensity value and learning fatigue index are fused to obtain the module fatigue assessment result.

4. A modular content generation method for embedded teaching according to claim 3, characterized in that: If the module fatigue assessment result exceeds the preset fatigue threshold, the specific steps include: Based on the skill span distribution between the pre-skill set and the output skill set, identify the skill transition mutation points in the initial teaching module; based on the skill transition mutation points, determine the cognitive overload sub-areas in the initial teaching module; A graph neural network is used to generate a learning content chain diagram for the cognitive overload sub-area. Cluster analysis is performed based on the learning content chain diagram to extract candidate sets of teaching sub-modules with longer operation chains. Identify the number of the shortest paths, the longest paths, and the cross paths of the learning content link graph, and perform load balancing splitting on the teaching submodule candidate set based on the identified number to obtain at least two split embedded teaching modules; The teaching sequence relationship of the split embedded teaching module is re-established, and the prerequisite skill set and output skill set corresponding to the split embedded teaching module are updated.

5. A modular content generation method for embedded teaching according to claim 4, characterized in that: If the module fatigue assessment result is far below the preset fatigue threshold, the specific steps include: Extract other initial teaching modules that have similar output skill sets to the current initial teaching module and whose module fatigue assessment results are far below the preset fatigue threshold as associated initial teaching modules; Identify the learning content link graphs of multiple related initial teaching modules and construct a module merging candidate pair corresponding to the current initial teaching module; Introducing a learning coherence scoring model; using the learning coherence scoring model to screen module merging candidate pairs and select module merging candidate pairs that meet the preset learning coherence requirements; Module merging candidate pairs that meet the preset learning coherence requirements are combined to obtain a combined embedded teaching module; the teaching sequence relationship of the combined embedded teaching module is re-established, and the corresponding prerequisite skill set and output skill set of the combined embedded teaching module are updated.

6. A modular content generation method for embedded teaching according to claim 5, characterized in that: Based on the student learning profile and the embedded teaching modules that need to be learned, the specific steps for generating the corresponding embedded module learning content include: Obtain student learning profiles; student learning profiles include current learning cognitive load, learning style preferences, and prior knowledge level; match semantic generation granularity based on prior knowledge level; generate semantic generation preferences based on learning style preferences; and match semantic generation density based on current learning cognitive load. The corresponding embedded module learning content of the embedded teaching module is generated based on the semantic generation granularity, semantic generation preference and semantic generation density.

7. A modular content generation system for embedded teaching, characterized in that: The system applies a modular content generation method for embedded teaching according to any one of claims 1 to 6, comprising: The embedded teaching module division module includes an initial module generation unit and an embedded teaching module generation unit; the initial module generation unit is used to obtain original teaching content related to embedded teaching; the original teaching content is semantically parsed and structured, and the original teaching content is divided into several initial teaching modules; each initial teaching module is assigned a corresponding prerequisite skill set and output skill set; all initial teaching modules are sorted according to the teaching sequence relationship; the embedded teaching module generation unit is used to perform learning fatigue evaluation on the initial teaching module based on the prerequisite skill set and the output skill set for any initial teaching module, and obtain a module fatigue evaluation result; the module fatigue evaluation result is compared with a preset fatigue threshold to judge: if the module fatigue evaluation result exceeds the preset fatigue threshold, the corresponding initial teaching module is split again to obtain at least two split embedded teaching modules; if the module fatigue evaluation result is far below the preset fatigue threshold, multiple related initial teaching modules are selected and combined to obtain a combined embedded teaching module; Based on the step of comparing the module fatigue assessment result with the preset fatigue threshold, all initial teaching modules are traversed and compared, and all the obtained split embedded teaching modules and combined embedded teaching modules are renamed to obtain a number of embedded teaching modules; The embedded teaching content generation module includes a module content generation unit; the module content generation unit is used to generate corresponding embedded module learning content based on the student learning profile and the embedded teaching module that needs to be learned; learning is carried out based on the corresponding embedded module learning content and learning feedback data is collected to update the student learning profile.

Images