Interactive teaching content generation method and system supporting real-time data processing

Abstract

The invention relates to the technical field of interactive teaching, and discloses an interactive teaching content generation method and system supporting real-time data processing. The method comprises the following steps: acquiring real-time teaching data including student behavior data and teaching content data; performing learning state evaluation analysis on the student behavior data to obtain a learning state evaluation result; comparing the result with a preset state threshold value, and generating a teaching content adjusting signal when the result does not accord with the preset state threshold value; acquiring associated data of the teaching content based on the signal, and performing content reconstruction analysis to obtain reconstructed teaching content; and generating interactive teaching content based on the reconstructed teaching content. According to the method, learning behaviors and teaching content changes of students are captured in real time, the teaching content is adjusted in time in combination with learning state evaluation, the generated interactive teaching content can promote participation of the students, personalized learning requirements are met, and the pertinence and effectiveness of teaching are improved.

Description

Technical Field

[0001] This invention relates to the field of interactive teaching technology, specifically to a method and system for generating interactive teaching content that supports real-time data processing. Background Technology

[0002] In traditional teaching models, the generation of teaching content often relies on pre-set syllabi and teaching plans, making it difficult to dynamically adjust according to students' real-time learning progress. Teachers typically teach according to a uniform pace and standards, failing to accurately capture the behavioral changes and individual needs of each student during the learning process. For example, when students show signs of inattention or increased error rates in exercises, teachers struggle to detect these issues promptly and adjust the teaching content. This results in some students gradually losing interest in learning because they cannot keep up with the pace, while others do not benefit sufficiently from the overly simplistic content. With the development of online education, although some teaching content delivery methods based on preset rules have emerged, these methods mostly rely on the analysis of historical data and lack the ability to process real-time data. When a student's learning status changes rapidly, such as suddenly experiencing confusion or accelerated understanding of a certain knowledge point, existing systems cannot quickly respond and generate appropriate teaching content. This static or semi-static teaching content generation model struggles to meet the needs of personalized teaching and limits further improvement in teaching effectiveness. Existing methods for generating instructional content often lack effective interactive mechanisms. Students can only passively receive the generated content and cannot influence the generation process through their own feedback. This one-way information transmission model not only reduces student engagement but also makes it difficult to optimize and adjust the content based on real-time student feedback, thus affecting the relevance and effectiveness of instruction. Summary of the Invention

[0003] The purpose of this invention is to provide an interactive teaching content generation method and system that supports real-time data processing, so as to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides an interactive teaching content generation method supporting real-time data processing, the method comprising: Acquire real-time teaching data, which includes student behavior data and teaching content data; The student behavior data is analyzed to assess and obtain the learning status assessment results. The learning status assessment result is compared with a preset status threshold. When the learning status assessment result does not meet the preset status threshold, a teaching content adjustment signal is generated. Based on the teaching content adjustment signal, obtain the associated data of the teaching content, perform content reconstruction analysis on the associated data, and obtain the reconstructed teaching content; Based on the reconstructed teaching content, interactive teaching content is generated.

[0005] Preferably, the step of performing learning status assessment and analysis on the student behavior data includes: The student behavior data is divided into multiple time periods according to a preset time interval; Acquire student behavior data for each time period and calculate student behavior indicators for each time period; By comparing the student behavior indicators for each time period with the preset behavior indicator standard values, the learning status assessment value for each time period is obtained. The learning status assessment values ​​for each time period are weighted and summed to obtain the learning status assessment result.

[0006] Preferably, the calculation of student behavior indicators for each time period includes: Obtain the number of student interactions, the accuracy rate of student answers, and the student concentration within each time period; The number of student interactions, the accuracy rate of student answers, and the student concentration are standardized to obtain standardized student interaction counts, standardized student answer accuracy rates, and standardized student concentration. The standardized student interaction frequency, standardized student answer accuracy, and standardized student attention concentration are weighted and summed to obtain the student behavior indicators for each time period.

[0007] Preferably, the content reconstruction analysis of the associated data includes: Obtain the knowledge point association data and student historical learning data from the associated data; Based on the knowledge point association data, a knowledge point association network is constructed; Based on the students' historical learning data, obtain data on the students' mastery of knowledge points; Based on the data on students' mastery of knowledge points, weak knowledge points are identified in the knowledge point association network. Based on the identified weak knowledge points, restructured teaching content is generated.

[0008] Preferably, the step of constructing a knowledge point association network based on the knowledge point association data includes: Treat each knowledge point as a network node; The association strength between knowledge points is obtained, and a connection edge is established between two knowledge points whose association strength is greater than a preset association strength threshold to form a knowledge point association network.

[0009] Preferably, the step of identifying weak knowledge points in the knowledge point association network based on the students' knowledge point mastery data includes: The student's knowledge point mastery data is compared with the knowledge point mastery threshold. If the student's knowledge point mastery data is lower than the knowledge point mastery threshold, it is marked as a weak knowledge point. The weak knowledge points and their directly related knowledge points are extracted from the knowledge point association network and used as candidate knowledge points for reconstructing the teaching content.

[0010] Preferably, the generation of interactive teaching content includes: The reconstructed teaching content is assembled according to a preset interactive template to obtain interactive teaching content; The interactive template includes a question-and-answer section, a simulated operation section, and a real-time feedback section. The generation of the real-time feedback process includes: Set up feedback points in interactive teaching content; When a student arrives at a feedback point, their learning behavior data at that feedback point is collected. The learning behavior data is compared with preset feedback behavior standards to generate real-time feedback content.

[0011] Preferably, the acquisition of student attention concentration includes: Acquire students' facial image data via camera; Eye state data was obtained by analyzing the student's facial image data. The student's attention span is calculated based on eye condition data.

[0012] Preferably, after generating interactive teaching content, it also includes: The interactive teaching content is sent to the student's terminal; Receive real-time learning data from students during the interactive teaching process, and return the real-time learning data as new student behavior data to the step of obtaining real-time teaching data.

[0013] Preferably, the present invention further includes an interactive teaching content generation system supporting real-time data processing, used to implement the interactive teaching content generation method supporting real-time data processing as described above, the system comprising: The real-time data acquisition module is used to acquire real-time teaching data, which includes student behavior data and teaching content data. The learning status assessment module is connected to the real-time data acquisition module and is used to perform learning status assessment and analysis on the student behavior data to obtain learning status assessment results. The state comparison and signal generation module is connected to the learning state evaluation module and is used to compare the learning state evaluation result with a preset state threshold. When the learning state evaluation result does not meet the preset state threshold, a teaching content adjustment signal is generated. The associated data acquisition and content reconstruction module is connected to the state comparison and signal generation module. It is used to adjust the signal based on the teaching content, acquire the associated data of the teaching content, and perform content reconstruction analysis on the associated data to obtain the reconstructed teaching content. An interactive teaching content generation module, connected to the associated data acquisition and content reconstruction module, is used to generate interactive teaching content based on the reconstructed teaching content.

[0014] Compared with the prior art, the beneficial effects of the present invention are: By acquiring real-time teaching data, we can promptly capture the dynamic changes in students' learning behaviors and teaching content. This ensures that the generation of teaching content is based on real-time information, avoiding the lag that may result from relying on historical data or pre-set rules. Analyzing student behavior data to assess their learning status provides a deeper understanding of how students' learning progresses, such as their mastery of knowledge points and their learning pace, offering a basis for adjusting subsequent teaching content. When the learning status assessment results do not meet the preset threshold, a teaching content adjustment signal is generated. This makes the adjustment of teaching content more targeted, enabling timely responses based on students' actual learning status and preventing a disconnect between teaching content and student needs. By acquiring related data based on the teaching content adjustment signal and performing content reconstruction analysis, various information related to the current teaching content can be integrated to generate reconstructed teaching content that better suits students' learning status, thus enhancing the adaptability of the teaching content. Reconstructing teaching content to generate interactive teaching content can promote interaction between students and the content. Students are no longer passive recipients of knowledge but can influence the further generation of teaching content through their own feedback, thus increasing student engagement. Simultaneously, this interactive teaching content generation model allows the content to be continuously optimized based on real-time student feedback, creating a virtuous cycle that better adapts to students' personalized learning needs and enhances the flexibility and effectiveness of the teaching process. Attached Figure Description

[0015] Figure 1 This is a sequence diagram of the interactive teaching content generation method supporting real-time data processing described in this invention; Figure 2 A flowchart for calculating student behavior indicators; Figure 3 A flowchart for content restructuring analysis; Figure 4A flowchart for identifying weak knowledge points; Figure 5 A flowchart generated for interactive teaching content. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Please see Figure 1 The present invention provides a method for generating interactive teaching content that supports real-time data processing, the method comprising: The system acquires real-time teaching data, including student behavior data and teaching content data. It then performs learning status assessment and analysis on the student behavior data to obtain assessment results. These results are compared with preset thresholds; if the assessment results do not meet the thresholds, a teaching content adjustment signal is generated. Based on this signal, related data of the teaching content is acquired, and content reconstruction analysis is performed on this related data to obtain reconstructed teaching content. Finally, interactive teaching content is generated based on the reconstructed teaching content. In the overall implementation scheme, student behavior data includes student interaction information, operational behaviors, and other activity records reflecting learning status during the teaching process; teaching content data covers course materials, knowledge unit structures, and teaching resources. The system automatically completes this process through pre-built software modules or algorithms, achieving a closed-loop operation from data acquisition to content generation. All data flows are processed through a network communication interface to ensure real-time performance and responsiveness throughout the process. The initial stage involves installing sensor devices or software programs to capture student behavior in the classroom environment; subsequent analysis steps use mathematical models to calculate assessment values; the threshold comparison stage uses a rule engine to determine whether adjustments are needed; the content reconstruction stage calls upon a database or knowledge base; and the generation stage uses preset template technology to assemble interactive elements. The output results can be directly integrated into the teaching platform or terminal application to achieve a teaching experience that dynamically adapts to students' needs.

[0018] Example 1: See Figure 2This embodiment is based on the continuous collection and analysis of student behavior data. Student behavior data is dynamically acquired during system operation, including student operation records, response behaviors, and interaction trajectories within the teaching environment. Initially, this data is input as an event stream and processed in segments according to preset and configurable time intervals. The length of these time intervals can be adjusted according to the actual teaching scenario, for example, dividing the entire lesson data into fixed ten-minute units. Each segmented time period is considered an independent processing unit, and the system performs calculations on the student behavior data contained within that unit.

[0019] For a single time period, the system identifies and statistically analyzes three core behavioral dimensions: the total number of interactive behaviors initiated by students during that time period; the accuracy rate of students' answering tasks during that time period; and the raw measurement of students' attention concentration during that time period. These raw data, due to differences in their dimensions and ranges, need to be standardized to transform them into comparable values. The standardization of the total number of interactions references the system's preset maximum potential interaction limit, comparing the actual number of interactions with this limit to generate a percentage value representing the level of interactive activity. The answer accuracy rate exists as a natural ratio and is usually directly used as a usable value after standardization. The raw measurement of attention concentration is calculated based on the device's data acquisition principle, yielding a percentage value representing the level of attention concentration per unit time.

[0020] The system assigns adjustable contribution weight coefficients to these three standardized dimensions. The final student behavior indicator value for each time period is determined by the sum of these three weighted factors. The contribution weight allocation strategy is based on pre-set teaching analysis objectives. After calculating the student behavior indicator for each time period, the system immediately compares it with preset behavioral indicator standard values ​​stored in the system. The preset behavioral indicator standard values ​​represent the baseline learning behavior level expected to be achieved in that time period. The comparison process calculates the degree of deviation between the two values, and the magnitude of the deviation is mapped and converted into a learning status evaluation value within a predetermined range. This value directly reflects the degree of conformity between the student's behavior and the expected target state in that time period; the greater the deviation, the lower the evaluation value.

[0021] For each time period, a corresponding learning state evaluation value is generated, forming a chronological sequence. The system performs a summation calculation on this sequence using a time-weighted factor. The allocation strategy for the time weight factor emphasizes temporal proximity, assigning a larger weight to time periods closer to the current processing time. This weighting mechanism makes the system pay more attention to the state influence of the most recent time period, ensuring the evaluation results are sensitive to the current learning state. The comprehensive value generated by the weighted summation is the learning state evaluation result output for that round.

[0022] At the system operation level, the key to this embodiment lies in maintaining the temporal integrity of data and the real-time nature of processing. After student behavior data flows into the system, it is received and buffered through a distributed message queue. The background computing process subscribes to queue messages and performs segmentation according to time windows. Data blocks for each time period are distributed to parallel computing nodes for standardization, indicator calculation, and state evaluation value transformation. Calculation results are written to a high-speed cache immediately. When all current time periods have been processed, a weighted summation task is triggered, retrieving sequence values ​​from the cache for comprehensive calculation. The time weight sequence is loaded from the configuration file. After the calculation results are generated, they are pushed to the downstream comparison module interface. The entire processing chain relies on a microservice architecture for coordination, ensuring that the current round of calculation is completed before the next time period window closes, avoiding data backlog. An error detection mechanism monitors the status of each stage, initiating data supplementation or weight reloading processes when an anomaly occurs to maintain the continuity of system operation. Data storage uses a time-series database to efficiently record the original behavior flow, segmentation markers, intermediate indicators, and final state evaluation results, facilitating possible backtracking analysis. The system's dynamic configuration capability allows for adjustments to segmentation intervals, standardization parameters, dimension weights, or time weight sequence configurations without interrupting operation.

[0023] Example 2: See Figure 3 This embodiment focuses on processing data related to teaching content and constructing a knowledge network. When a signal indicating an adjustment to the teaching content is triggered, the system automatically initiates the retrieval process for related data. The related data mainly includes two categories of datasets: one is pre-stored data on relationships between knowledge points, and the other is historical learning record data extracted from student files. The knowledge point relationship data originates from a structured description file in the teaching resource database, recording knowledge point identifiers, the types of logical relationships between knowledge points, and quantified relationship strength values. Student historical learning data is dynamically linked to the database interface of the learning management system, extracting complete learning trajectory records based on student identity, including but not limited to practice scores, assessment levels, or mastery markers for each knowledge point.

[0024] The knowledge point association network is constructed using a graph structure model. The system initializes each independent knowledge point in the knowledge system as a network node. The node data structure includes a unique code, standard name, and content summary field for the knowledge point. After node creation, the system loads predefined association strength matrix data. This matrix stores knowledge point pairs and their corresponding association strength values ​​in list form. Each strength value in the matrix is ​​set according to teaching logic rules, typically within a closed interval of 0 to 1. The system traverses all knowledge point pair records in the matrix, determining whether the association strength exceeds a preset activation threshold (e.g., 0.5). Knowledge point pairs that meet the condition will automatically establish connections between their corresponding nodes. Connection edges exist as objects in the system, storing association strength values ​​and association type attributes (e.g., "prerequisites" or "extensions"). The connection edge establishment process is executed cyclically until the matrix traversal is complete, ultimately forming an interconnected network topology. The network construction process calls the in-memory graph database processing engine to perform node instantiation, edge object generation, and graph structure verification.

[0025] Based on the loaded student historical learning data, the system parses out students' knowledge point mastery information. This information exists in the form of structured records, with each record associated with a specific knowledge point identifier and a corresponding quantitative value of mastery (e.g., a score from 0 to 100). The system reads each record one by one according to the knowledge point identifier, and compares the mastery value of each knowledge point with the mastery threshold configured by the system. The threshold parameter is set to a fixed value or a dynamic range value (e.g., 70 points) according to teaching requirements. The comparison result triggers a marking logic: when the mastery value of a knowledge point is lower than the configured mastery threshold, the system adds a weak status mark to the knowledge point node, generating a list data structure containing all marked knowledge points.

[0026] Within the constructed knowledge point association network environment, the system performs a network traversal operation based on the aforementioned list of weak knowledge points. The traversal process employs an adjacency node search algorithm: the system sequentially visits each marked knowledge point node in the list, using it as a starting point to scan the entire network, searching for all associated knowledge point nodes with direct connections to it. The search process is limited to one-hop connections to avoid over-spreading. All located marked knowledge point nodes and their directly associated nodes are merged and output, forming a candidate knowledge point set required for reconstructing the teaching content. This set is cached as an ordered list, with each list item containing a knowledge point identifier, node attributes, and connection relationship attributes.

[0027] At the system implementation level, this embodiment emphasizes dynamic data integration and robust network processing. Knowledge point relationship matrix data is hot-loaded via configuration file interface or database connector. The network construction engine adopts an asynchronous execution mode, ensuring that node and edge generation does not block the main process. Student historical learning data is transmitted through an encrypted channel, with validity checks performed during data parsing to exclude abnormal records. The network traversal algorithm employs a depth-first search strategy, setting a maximum search depth limit to prevent uncontrolled resource consumption. The output data of the candidate set is encapsulated in a standardized JSON format, containing a dual structure of node arrays and relationship edge arrays. Anomaly monitoring is embedded throughout the processing flow; if related data is missing, a default value rollback strategy is triggered. The system resource management module dynamically allocates computing nodes, automatically initiating a job splitting and parallel processing mechanism when the network scale is large. Data processing results are written to a temporary storage area in real time, with lifecycle tags set to meet the timeliness requirements of subsequent modules. All operation logs record runtime metrics such as the number of nodes generated, the number of edge connections, and traversal time.

[0028] Example 3: See Figure 4 This embodiment focuses on judging the degree of mastery of knowledge points and locating them in the association network. After acquiring student knowledge point mastery data, the system stores this data in the form of a set, where each entry contains a unique identifier for a specific knowledge point and a corresponding quantitative evaluation value for mastery. The system also loads preset knowledge point mastery threshold parameters. The system sequentially traverses each knowledge point entry in the dataset, performing the following judgment operation: extracting the quantitative evaluation value for mastery from the entry and comparing it with the corresponding threshold stored in the system. The threshold parameters are configurable; a globally uniform threshold may be used for all knowledge points, or differentiated thresholds may be assigned based on the teaching difficulty or importance of the knowledge point. The threshold type can be a fixed value or a preset floating range based on knowledge classification.

[0029] The judgment logic is based on numerical comparison results: when the quantitative evaluation value of the mastery of a specific knowledge point is less than the association threshold of that knowledge point, the system automatically adds a special mark to the metadata record of that knowledge point. This mark indicates that the knowledge point is in a weak state. The marking operation modifies the status attribute field of the knowledge point entry, and simultaneously records the reason for marking (i.e., below the threshold) and the marking timestamp. After all entries have been traversed, the system aggregates all marked knowledge point entries to generate an ordered list of weak knowledge points. The list is sorted by the knowledge point identifier index for easy and quick location later.

[0030] In the constructed knowledge point association network model environment, the system uses the aforementioned list of weak knowledge points as input. The network model exists in computer memory as a graph data structure, where node objects represent knowledge points, and the edge objects between nodes store association attributes. The system initiates a network traversal subprocess: sequentially accessing each weak knowledge point identifier in the list and locating the corresponding network node through the graph structure query interface. Using this located node as the origin, the system performs an adjacency node search operation: extracting the set of all adjacent nodes directly connected to the origin node. The search logic is explicitly limited to first-degree adjacency relationships, that is, only obtaining one-hop neighbor nodes that are directly connected to the origin. The search process may access multiple connecting edges pointing to different directions or types, but the depth is controlled to only process the first-level adjacency relationships.

[0031] The acquired set of direct neighbor nodes is merged with the original weak knowledge point node itself to form a temporary node set. For each node in the set, the system extracts its complete knowledge point attribute information (such as identifier, name, and content summary) and its connection edge attribute information with the origin (such as association strength and relationship type). The system performs deduplication verification on this temporary node set: if the same knowledge point is repeatedly included because it is a neighbor of different origins, the system will merge it into a single record.

[0032] After the above processing, the system finally generates a set of candidate knowledge points for reconstructing teaching content. This set exists in the form of a structured list of records, with each record fully describing a knowledge point and its position and background information in the network. It should be noted that, to quantify the potential influence of a node in the network, the system defines a knowledge point matching degree parameter M(k) based on adjacency relationships. Its calculation logic is based on the mastery deviation of the current node from its neighbor connection strength: ; The characters in the formula have the following meanings: k: represents the identifier of the knowledge point being processed; S(k): the quantitative evaluation value of the mastery level of this knowledge point obtained from the student data; T(k): the mastery level threshold parameter corresponding to this knowledge point preset by the system; N(k): the set of all neighboring nodes in the network that have a direct adjacency relationship (one-hop connection) with knowledge point k; n: represents a single neighboring node in the set *N(k)*; w(n): represents the attribute strength value of the connection edge between knowledge point k and neighboring node n (in the range of 0 to 1); λ: represents the network influence adjustment coefficient (preset value between 0 and 1), used to control the weight ratio of the influence of neighboring nodes in the calculation.

[0033] At the technical implementation level, this embodiment emphasizes the accuracy of data mapping and the efficiency of network traversal. The knowledge point mastery dataset is loaded through a distributed cache, and threshold queries are service-oriented to avoid repeated database reads. Weakness markers employ a lock-free state-based marking mechanism, supporting multi-threaded concurrent marking operations. The graph-structured network model is accessed through a dedicated graph computing engine (such as JanusGraph), and adjacency search uses index-optimized queries, limiting the maximum number of neighbors retrieved to prevent result set expansion. Deduplication is based on hash comparison using unique knowledge point IDs. The candidate set adopts a data contract design, specifying JSONSchema for reliable transmission between modules. The monitoring system collects operational metrics in real time, such as processing time, number of marked points, and average degree of neighboring nodes. Traversal anomalies (such as non-existent nodes) trigger retry or skip strategies and generate alerts. The entire execution process is instantiated in a containerized environment, and resource configuration scales elastically according to the network node scale. Debug mode supports outputting intermediate network subgraph snapshots for logic verification.

[0034] Example 4: See Figure 5 This embodiment focuses on the assembly of interactive teaching content and the establishment of a real-time feedback mechanism. After acquiring the reconstructed teaching content, the system initiates the content assembly process. This process, based on a preset interactive template structure, transforms knowledge points into interactive teaching modules. Taking the junior high school mathematics "quadratic equations" knowledge unit as an example, the system extracts three core knowledge points from the reconstructed content: derivation of the quadratic formula, application of the discriminant, and modeling of practical problems. These knowledge points are assigned to different interactive stages according to the template specifications.

[0035] The question-and-answer session is constructed by automatically generating a sequence of questions that match the knowledge points. The system retrieves questions related to the knowledge points from the question bank and arranges them according to difficulty levels. For example, for the discriminant application knowledge point, the system generates the following question sequence: basic questions require calculating the discriminant value of a given equation; advanced questions require determining the root case based on the discriminant result; and comprehensive questions analyze the rationality of the solution in conjunction with a practical problem. The questions are presented in multiple-choice, fill-in-the-blank, and drag-and-drop sorting formats, with each question type corresponding to different interactive controls.

[0036] The simulation operation section designs a virtual experimental environment for knowledge points that require hands-on practice. Taking formula derivation as an example, the system generates an interactive simulator of formula transformation steps. Students can complete the derivation process from the standard form to the root formula by dragging and dropping algebraic terms and selecting transformation rules. During the operation, the system monitors the correctness of the steps in real time, but does not immediately interrupt erroneous operations; instead, it records the type of deviation for subsequent feedback.

[0037] The real-time feedback process includes multiple pre-set detection nodes within the content, referred to as feedback points. The location of these feedback points is determined based on the logical structure of the knowledge points. Taking the knowledge point of practical problem modeling as an example, the system sets feedback points at the following locations: at the end of the problem understanding stage, when variable setting is complete, after equation establishment, and before solution verification. Each feedback point is associated with learning behavior detection indicators, and the system continuously collects student operation data at these nodes through an interface.

[0038] Student attention data is collected through a visual analysis module. The system connects to smart cameras deployed in the classroom, capturing student facial images at a rate of 3 frames per second. The images are transmitted to an analysis server, where the following processing steps are executed: first, key facial points are located, with particular attention to the eye area; then, the eye opening / closing state, pupil center position, and gaze direction are calculated; finally, these parameters are combined to generate an attention score. The score is expressed as a percentage, and an average score is output every minute.

[0039] The table below shows a segment of a student's attention data recorded by the system during the teaching of the "Quadratic Equations in One Variable" unit. See Table 1.

[0040] Table 1: Data snippet on a student's attention.

[0041] The system assembles the components generated in each of the above stages according to the logical order of the teaching process. The Q&A session is placed in the knowledge introduction stage, the simulated operation session is arranged after the concept explanation, and feedback points are evenly distributed at various key transition points. The assembled interactive content package adopts a standard encapsulation format, including the HTML5 content body, interactive scripts, media resources, and metadata description files.

[0042] When students access the content via their terminal devices, the system initiates a real-time monitoring process. The student's choices during the question-and-answer session, their actions during the simulation, and their status data at each feedback point are all recorded. This data is compared with a pre-set behavioral pattern database to identify situations requiring intervention. For example, when three consecutive questions are answered incorrectly, or a specific type of error sequence occurs during the simulation, the system triggers the generation of feedback content.

[0043] The feedback content generation module selects an appropriate guidance strategy based on the type of error. For conceptual errors, the system extracts alternative explanatory text and examples; for operational mistakes, it provides step-by-step guidance; for distractions, it adjusts the pace of content presentation or inserts brief interactive activities. The generated feedback content is inserted into the original content flow as an overlay to maintain contextual coherence.

[0044] The visual analysis module runs continuously, and its data processing flow includes a quality inspection step. When changes in ambient light cause a decrease in image quality, the system automatically adjusts camera parameters or switches to a backup infrared mode. The analysis algorithm is tolerant of occlusion; when part of the face is obscured, attention trends can be estimated based on remaining visible features. The system periodically calibrates baseline values ​​to eliminate measurement biases caused by device differences.

[0045] The content delivery system employs adaptive streaming technology, dynamically adjusting media quality based on network conditions. Interactive operation data is transmitted back via a differential update mechanism, reducing bandwidth consumption. A lightweight analytics module on the terminal device preprocesses basic behavioral data, uploading only key metrics to the server to reduce system latency. All data transmission uses encrypted channels, and student privacy data is anonymized at the collection point.

[0046] The system maintains a complete log of the interaction process, recording the triggering circumstances of each feedback point, the type of feedback content generated, and the student's subsequent response. These logs are used to periodically optimize feedback strategies and content organization. The teacher management interface provides a visual dashboard displaying the attention change curves of all students in the class and feedback point activation heatmaps to assist in teaching decisions. The system supports adjusting feedback sensitivity and intervention intensity through a configuration interface to adapt to the needs of different teaching styles.

[0047] Example 5: This example is based on the previously generated interactive teaching content. The system initiates the content distribution process. The content is encapsulated into a structured data packet, including a metadata header and a payload. The metadata header records the content identifier, target student identifier, generation timestamp, and priority flag. The system queries the list of registered student terminal devices to match the target student's currently active device type and network address. The distribution protocol is adaptively selected based on the device type: TCP direct transmission is used for fixed terminals within the local area network; HTTPS secure channel is used for mobile devices. During transmission, packet verification is performed, with a checksum appended to each data block. The receiving terminal verifies the integrity of each block and triggers reassembly. A confirmation signal for successful reassembly is returned to the system log module.

[0048] Student terminals run a dedicated application or browser plugin to receive content packets. The application's unpacking module parses the metadata and content structure, and launches the local rendering engine to display the interactive interface. All student actions on the interface are captured by the monitoring subsystem; these actions include content block navigation, question option selection, simulated step operations, and feedback area responses. For each event, the terminal automatically appends a precise timestamp, the current content location identifier, and operation description parameters. Event data is temporarily stored in a local buffer in a lightweight binary format.

[0049] The terminal system is configured with a background data transmission agent that scans the local buffer at fixed time intervals. When new event data accumulation or the interval period is reached, the agent initiates a compressed and encrypted transmission process. Event data is first serialized into a protocol buffer format, compressed using the LZ4 algorithm, and then encrypted using the AES-GCM algorithm. The encrypted data is sent to the central processing system via a message queue middleware. Message queue topics are partitioned by student identifiers to ensure on-demand routing. The transmission process supports a breakpoint resumption mechanism, automatically buffering data to be transmitted when the network is interrupted.

[0050] The central processing system's receive port listens to the message queue, and the distribution service routes data to the designated processing unit based on the student identifier. Upon arrival at the processing unit, the data is first decrypted and decompressed to restore the original event stream. The deserialization module converts the binary stream into structured event records. These records then undergo a timestamp alignment phase to correct for minor deviations from different clock sources. The processing unit performs necessary data augmentation operations: mapping content location identifiers to standard knowledge point codes and categorizing operation description parameters into preset behavior categories. All transformed fields are written to a unified event schema.

[0051] The data fusion module receives the transformed event stream and synchronizes it with the system's internal clock signal. The module then retrieves the student's existing historical behavior database, extracting relevant records from the most recent few class periods. The old and new data are merged along a timeline: duplicate records are overwritten and updated, and conflicting timestamps are matched using manual rules, ultimately generating a continuous time series of behavioral events. This time series is restructured into a system-defined real-time teaching data format, where the student behavior data portion contains complete three-dimensional relationship data between time, operation, and knowledge point.

[0052] After integration, the system pushes the newly generated teaching data record to the input endpoint of the initial process. Once the endpoint service verifies the integrity of the data structure, it inserts it into the initial cache queue of the real-time processing pipeline. The cache queue adopts a first-in, first-out (FIFO) strategy, and the front-end processing program continuously consumes the queue's content. At this point, the data completes a closed-loop flow, forming a continuous cycle from content generation to interactive operations and then to new data collection.

[0053] The system is configured with a monitoring system for the entire process described above. During the distribution phase, transmission latency, packet loss rate, and terminal offline times are recorded. The terminal acquisition layer reports event loss counts, local buffer duration, and transmission failure types. The central processing unit tracks event parsing error rate, data alignment deviation, and fusion conflict handling flags. When these indicators exceed preset operating ranges, a hierarchical alarm mechanism is triggered: the system automatically attempts self-recovery operations according to priority, including switching transmission paths, clearing dirty data, and resetting processing units; if the recovery operation fails, manual intervention is notified, and a snapshot of the fault scene is preserved. All operation logs are appended with a unified tracking identifier, supporting end-to-end behavior auditing. During continuous operation, latency metrics for each cycle are recorded for system optimization reference. The maintenance interface supports dynamic adjustment of operating parameters such as acquisition frequency, transmission interval, and buffer size to adapt to different teaching scenario load requirements.

[0054] Example 6: In the WorldSkills Competition preparation scenario for Mechatronics on the "WorldSkills Connect" education platform, the system initiates a real-time teaching data acquisition process. Real-time teaching data comprises two parts: student behavior data and teaching content data. Student behavior data originates from multi-terminal interaction records on the platform, specifically including the number of times students search for mechatronics technical documents in the "WorldSkills Research" module on their desktops, their Q&A interaction records when participating in AI-assisted practice on their mobile phones (e.g., the number of questions asked and response time for "mechanical assembly precision control"), their answers to the platform's built-in provincial competition simulation questions (including the answering time and accuracy rate for each question), and facial image data collected by the terminal camera authorized by the platform. Teaching content data covers the WorldSkills Mechatronics project course materials currently pushed by the platform (e.g., 2024 WorldSkills Mechatronics project technical specification interpretation videos, national competition past exam question analysis PPTs), and personalized training tasks generated by AI-assisted practice (e.g., PLC programming special exercises). This data is aggregated to the data processing center through the platform's real-time data transmission channel, employing a segmented verification mechanism during transmission to ensure the integrity and timeliness of each batch of data.

[0055] The system divides the collected student behavior data into multiple time periods at preset 15-minute intervals, with each time period serving as an independent analysis unit for learning status assessment. For each time period, key behavioral data is first extracted: the number of interactions between the student and the AI ​​tutor (e.g., the number of times the student proactively asks questions or responds to tutoring tasks), the accuracy rate of completing simulated test questions (the ratio of correct answers to the total number of questions), and attention concentration (obtained through facial image data analysis; the platform's visual analysis module identifies eye states in the images and calculates the percentage of time the eyes are open and focused on the terminal screen per unit time, serving as a quantitative value for attention concentration). These three data points are then standardized: the number of interactions is calculated as the percentage of actual interactions based on the platform's maximum responsiveness for the AI ​​tutor within that time period; the accuracy rate is retained as a percentage; and attention concentration is retained as a percentage. All three are converted to standardized values ​​between 0 and 1. Next, based on the key requirements of the WorldSkills Competition preparation scenario, weights were assigned to three standardized data points: AI tutoring interaction frequency (0.3), answer accuracy (0.5), and attention concentration (0.2). These weighted sums were then used to calculate student behavior indicators for each time period. Subsequently, the behavior indicators for each time period were compared with the platform's preset standard values ​​(set according to the preparation requirements of different stages of the provincial, national, and WorldSkills competitions; for example, the standard value for the WorldSkills preparation stage is 0.8). The deviation between the indicators and the standard values ​​was calculated, and this deviation was mapped to a learning status evaluation score of 0-100 points; the smaller the deviation, the higher the evaluation score. Finally, the evaluation scores for all time periods were weighted according to their temporal proximity, with the two most recent 30-minute time periods each having a weight of 0.3, and the weights of the preceding time periods decreasing sequentially. The weighted sums were then used to obtain the final learning status evaluation result.

[0056] The system compares the learning status assessment results with a preset status threshold, which is dynamically adjusted according to the current preparation stage. For example, the threshold for the national competition sprint stage is set at 85 points. If a student's assessment result is 72 points, which is lower than the threshold, the system automatically generates a teaching content adjustment signal. The signal contains key information such as the student's identifier, the current preparation stage (national competition sprint), and weak behavioral dimensions (e.g., a correct answer rate of only 0.6 and an attention concentration of 0.75), which is used for accurate retrieval of subsequent related data.

[0057] Based on the teaching content adjustment signals, the platform's related data retrieval module obtains related data from the three-level knowledge base of "WorldSkills Connect" and the student historical database. Among them, the knowledge point (skill point) related data comes from the "WorldSkills Project" module of the knowledge base. This module stores the relationship between various skill points in the mechatronics project, such as the correlation strength between "PLC programming" and "sensor debugging" (set to 0.85 based on the frequency of their co-occurrence in WorldSkills real questions), and the correlation strength between "mechanical assembly" and "precision testing" (0.9). The student historical learning data is extracted from the student's profile on the platform, including the student's preparation records for the past 3 months (such as the average accuracy rate of PLC programming questions in the provincial competition simulation, the time taken to complete the sensor debugging task), and the weak skill items marked by the coach (such as "complex circuit wiring"). Subsequently, a knowledge point association network is constructed: each skill point in the mechatronics project (such as PLC programming, sensor debugging, and mechanical assembly) is used as a network node. The association strength in the knowledge point association data is extracted. If the association strength is greater than a preset threshold of 0.7, a connection edge is established between the corresponding two nodes to form a skill point association network. Each edge in the network is labeled with the association strength and association type (such as "operational collaboration" or "knowledge dependence"). Based on students' historical learning data, the mastery level data of each skill point is extracted. The average accuracy rate of completing related tasks for that skill point in the past is used as the quantification value of mastery level (such as 0.65 for PLC programming and 0.82 for sensor debugging). The mastery level data is compared with the platform's preset knowledge point mastery threshold (such as 0.8). If the mastery level of a certain skill point is lower than the threshold (such as 0.65 < 0.8 for PLC programming), it is marked as a weak skill point. The weak skill point (PLC programming) and its directly related nodes (such as sensor debugging and HMI interface design, with association strengths of 0.85 and 0.78, respectively) are extracted from the association network. The skill points corresponding to these nodes are used as candidate skill points for reconstructing the teaching content. Finally, based on the candidate skill points, relevant teaching resources are retrieved from the knowledge base, such as the latest WorldSkills Competition rules interpretation videos for PLC programming, case tutorials on the collaborative operation of sensor debugging and PLC programming, and interactive practice question banks for HMI interface design. These resources are integrated to form reconstructed teaching content, with the content structure organized according to the logic of "basic review - difficult point analysis - practical exercises" to meet the training needs of the national competition sprint stage.

[0058] The reconstructed teaching content is assembled according to a pre-set interactive template on the platform, which includes three core components: a Q&A session, a simulated operation session, and a real-time feedback session. The Q&A session designs a sequence of questions around candidate skill points. For example, for PLC programming, questions such as "What are some common instruction errors in PLC programming at the WorldSkills Competition?" and "How can I achieve real-time acquisition of sensor data through a PLC program?" are generated. After students answer, the platform automatically verifies their answers by accessing the "WorldSkills Research" module in the knowledge base and displays extended explanations (such as links to relevant technical documents). The simulated operation session utilizes the platform's virtual training function to construct a collaborative operation scenario for PLC programming and sensor debugging. Students can drag and drop programming modules and connect sensor circuits in the virtual environment. During the operation, the platform records the operation data at each step (such as the order of instruction input and the connection nodes of the circuits). The real-time feedback session includes three feedback points: after the Q&A session, at key steps of the simulated operation (such as before program download), and after the simulated operation is completed. When a student reaches a feedback point, the platform acquires their learning behavior data at that point (such as the accuracy of their answers and the completeness of their operational steps), and compares it with preset feedback behavior standards (such as an accuracy rate of ≥0.9 and no omissions in operational steps). If the accuracy rate is only 0.7, the platform generates real-time feedback, pointing out the knowledge gaps corresponding to the incorrect answers (such as insufficient understanding of PLC interrupt instructions), and recommends AI-assisted practice tasks related to "PLC interrupt instruction application" in the knowledge base. It also pushes similar incorrect questions with coach annotations. If the sensor calibration step is omitted, the feedback displays a video clip of the calibration steps and provides calibration operation guidance in a virtual environment. Throughout the interactive content, students can trigger the knowledge retrieval function at any time, asking questions in natural language (such as "What is the time limit for PLC programming in the WorldSkills Competition's electromechanical project?"). The platform searches the three-level knowledge base in real time, returning text, images, or video results. The AI ​​tutor simultaneously provides multi-round dialogue answers. For example, if a student asks, "How can I optimize the programming steps within the specified time?", the tutor provides specific suggestions based on WorldSkills practical experience (such as prioritizing the use of subroutines to simplify code structure).

[0059] The generated interactive teaching content is distributed to students' desktops and mobile devices via the platform's terminal distribution function. During distribution, the content format is adapted according to the terminal type (e.g., optimized for portrait mode on mobile devices and simplified operation buttons). During the learning process, the platform continuously receives real-time learning data from students (e.g., new answers in the Q&A session, modified steps in simulated operations, and keywords for knowledge retrieval). This data, after format conversion, is returned as new student behavior data to the "obtain real-time teaching data" step and incorporated into the next round of learning status assessment and analysis, forming a data loop. This data is synchronized to the platform's skills education research module to analyze the distribution of students' skill weaknesses and the learning effectiveness of interactive content during the national competition preparation stage. Subsequently, the platform will update the knowledge base content based on these analysis results (e.g., supplementing high-frequency error cases in PLC programming), optimize talent training standards (e.g., adjusting the proportion of skills training in national competition preparation), and apply the updated standards and content to school skills talent evaluation and teacher teaching and research work, ensuring that competition preparation teaching practices are synchronized with the forefront standards of the WorldSkills Competition and industry technical needs. During this process, the platform will also use AI to analyze the latest WorldSkills Competition rules documents and competition question trends, continuously enriching the materials in the "WorldSkills Projects" and "WorldSkills Research" modules, shortening the cycle of introducing WorldSkills standards into teaching practice, and generating course design drafts and evaluation scale templates for coaches and teachers, reducing the burden of document writing and allowing teachers to focus more on personalized guidance; while students can obtain targeted training suggestions at any time through 24 / 7 online AI coaching, efficiently advancing the competition preparation process.

[0060] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0061] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for generating interactive teaching content that supports real-time data processing, characterized in that, Including the following steps: Acquire real-time teaching data, which includes student behavior data and teaching content data; The student behavior data is analyzed to assess and obtain the learning status assessment results. The learning status assessment result is compared with a preset status threshold. When the learning status assessment result does not meet the preset status threshold, a teaching content adjustment signal is generated. Based on the teaching content adjustment signal, obtain the associated data of the teaching content, perform content reconstruction analysis on the associated data, and obtain the reconstructed teaching content; Based on the reconstructed teaching content, interactive teaching content is generated.

2. The interactive teaching content generation method supporting real-time data processing according to claim 1, characterized in that, The learning status assessment and analysis of the student behavior data includes: The student behavior data is divided into multiple time periods according to a preset time interval; Acquire student behavior data for each time period and calculate student behavior indicators for each time period; By comparing the student behavior indicators for each time period with the preset behavior indicator standard values, the learning status assessment value for each time period is obtained. The learning status assessment values ​​for each time period are weighted and summed to obtain the learning status assessment result.

3. The interactive teaching content generation method supporting real-time data processing according to claim 2, characterized in that, The calculation of student behavior indicators for each time period includes: Obtain the number of student interactions, the accuracy rate of student answers, and the student concentration within each time period; The number of student interactions, the accuracy rate of student answers, and the student concentration are standardized to obtain standardized student interaction counts, standardized student answer accuracy rates, and standardized student concentration. The standardized student interaction frequency, standardized student answer accuracy, and standardized student attention concentration are weighted and summed to obtain the student behavior indicators for each time period.

4. The interactive teaching content generation method supporting real-time data processing according to claim 1, characterized in that, The content reconstruction analysis of the associated data includes: Obtain the knowledge point association data and student historical learning data from the associated data; Based on the knowledge point association data, a knowledge point association network is constructed; Based on the students' historical learning data, obtain data on the students' mastery of knowledge points; Based on the data on students' mastery of knowledge points, weak knowledge points are identified in the knowledge point association network. Based on the identified weak knowledge points, restructured teaching content is generated.

5. The interactive teaching content generation method supporting real-time data processing according to claim 4, characterized in that, The construction of a knowledge point association network based on the knowledge point association data includes: Treat each knowledge point as a network node; The association strength between knowledge points is obtained, and a connection edge is established between two knowledge points whose association strength is greater than a preset association strength threshold to form a knowledge point association network.

6. The interactive teaching content generation method supporting real-time data processing according to claim 4, characterized in that, The step of identifying weak knowledge points in the knowledge point association network based on the students' knowledge point mastery data includes: The student's knowledge point mastery data is compared with the knowledge point mastery threshold. If the student's knowledge point mastery data is lower than the knowledge point mastery threshold, it is marked as a weak knowledge point. The weak knowledge points and their directly related knowledge points are extracted from the knowledge point association network and used as candidate knowledge points for reconstructing the teaching content.

7. The interactive teaching content generation method supporting real-time data processing according to claim 1, characterized in that, The generation of interactive teaching content includes: The reconstructed teaching content is assembled according to a preset interactive template to obtain interactive teaching content; The interactive template includes a question-and-answer section, a simulated operation section, and a real-time feedback section. The generation of the real-time feedback process includes: Set up feedback points in interactive teaching content; When a student arrives at a feedback point, their learning behavior data at that feedback point is collected. The learning behavior data is compared with preset feedback behavior standards to generate real-time feedback content.

8. The interactive teaching content generation method supporting real-time data processing according to claim 3, characterized in that, The acquisition of student attention span includes: Acquire students' facial image data via camera; Eye state data was obtained by analyzing the student's facial image data. The student's attention span is calculated based on eye condition data.

9. The interactive teaching content generation method supporting real-time data processing according to claim 1, characterized in that, After generating interactive teaching content, it also includes: The interactive teaching content is sent to the student's terminal; Receive real-time learning data from students during the interactive teaching process, and return the real-time learning data as new student behavior data to the step of obtaining real-time teaching data.

10. An interactive teaching content generation system supporting real-time data processing, used to implement the interactive teaching content generation method supporting real-time data processing as described in any one of claims 1 to 9, characterized in that, include: The real-time data acquisition module is used to acquire real-time teaching data, which includes student behavior data and teaching content data. The learning status assessment module is connected to the real-time data acquisition module and is used to perform learning status assessment and analysis on the student behavior data to obtain learning status assessment results. The state comparison and signal generation module is connected to the learning state evaluation module and is used to compare the learning state evaluation result with a preset state threshold. When the learning state evaluation result does not meet the preset state threshold, a teaching content adjustment signal is generated. The associated data acquisition and content reconstruction module is connected to the state comparison and signal generation module. It is used to adjust the signal based on the teaching content, acquire the associated data of the teaching content, and perform content reconstruction analysis on the associated data to obtain the reconstructed teaching content. An interactive teaching content generation module, connected to the associated data acquisition and content reconstruction module, is used to generate interactive teaching content based on the reconstructed teaching content.

Images