Multi-scenario integrated learning tutoring systems and methods, storage media, and program products

By collecting and integrating learning context data across multiple scenarios, a digital twin model is constructed to generate personalized tutoring content, addressing the shortcomings of existing systems in providing individualized instruction and improving learning efficiency and effectiveness.

CN121685205BActive Publication Date: 2026-11-17BEIJING INSTITUTE OF EDUCATIONAL SCIENCES
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202511559552.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-10-29
Publication Date
2026-11-17
Estimated Expiration
2045-10-29

AI Technical Summary

Technical Problem

Existing online tutoring systems fail to provide individualized instruction, resulting in low learning efficiency and effectiveness for individual students and failing to meet the personalized needs of different learners.

Method used

By collecting multimodal learning context data in schools, schools, homes, and wearable devices, and using the data fusion module for spatiotemporal alignment and correlation information mining, a digital twin model of the target learner is constructed to generate personalized tutoring content.

Benefits of technology

It achieves spatiotemporal fusion of cross-scenario, multimodal data, dynamically maps learning behavior, generates highly targeted tutoring content, improves learning efficiency and knowledge mastery, and meets the individualized needs of different learners.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121685205B_ABST
    Figure CN121685205B_ABST
Patent Text Reader

Abstract

This application provides a multi-scenario integrated learning tutoring system and method, storage medium, and program product, relating to the field of computer program control technology. The system includes: a data acquisition module for collecting learning context data related to the learning behavior of a target learner in at least one data acquisition scenario; the at least one data acquisition scenario includes at least one of the following: in-school, out-of-school, home, and wearable devices; a data fusion module for dividing the learning context data into multiple behavioral sequences according to the time dimension, and mining the correlation information between the behavioral sequences based on the feature information of each behavioral sequence; a digital modeling module for deploying a digital twin model of the target learner based on each behavioral sequence and the correlation information; and a personalized content generation module for generating tutoring content information tailored to the learning behavior of the target learner based on the digital twin model. This application improves the learning efficiency and learning effectiveness of individual target learners.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of computer program control technology, and in particular to a multi-scenario integrated learning tutoring system and method, storage medium, and program product. Background Technology

[0002] Currently, online tutoring systems typically provide a uniform tutoring model to different learners, allowing them to choose specific tutoring content independently. For example, the platform may push the same set of tutoring videos to all students of the same grade, and students can only watch them in sequence or manually skip to the next episode. The platform does not simplify the tutoring content based on the knowledge points that students have already mastered, nor does it focus on key knowledge points that students are not familiar with. It also does not consider the individual characteristics of students, which affects the achievement of the goal of "teaching according to aptitude" and reduces the individual learning efficiency and effectiveness of each student. Summary of the Invention

[0003] This application provides a multi-scenario integrated learning tutoring system and method, storage medium, and program product to alleviate or solve one or more technical problems existing in the prior art.

[0004] In a first aspect, embodiments of this application provide a multi-scenario integrated learning tutoring system, including: The data acquisition module is used to collect learning context data related to the learning behavior of the target learner in at least one data acquisition scenario; the at least one data acquisition scenario includes at least one of the following: on-campus, off-campus, home, and wearable devices; The data fusion module is used to divide the learning context data into multiple behavior sequences according to the time dimension, and to mine the correlation information between the behavior sequences based on the feature information of each behavior sequence. A digital modeling module is used to deploy a digital twin model of the target learner based on each of the behavioral sequences and the associated information; the digital twin model is used to dynamically map the learning behavior of the target learner. The personalized content generation module is used to generate tutoring content information tailored to the learning behavior of the target learner based on the digital twin model.

[0005] In some implementations, in the on-campus data collection scenario, at least one of the following authorized data collection modules is provided: classroom camera, classroom microphone, and academic affairs system data interface; In the off-campus data collection scenario, at least one of the following authorized data collection modules is provided: research and study platform data interface, tutoring institution camera, tutoring institution microphone, and tutoring institution system data structure; In the data collection scenario of the home, at least one of the following authorized data collection modules is provided: home camera, home microphone, student smart learning device, parent smart device application data interface, home light intensity sensor, and decibel meter; In the wearable data collection scenario, at least one of the following authorized data collection modules is provided: smart bracelet, smartwatch, and smart glasses.

[0006] In some implementations, the learning context data includes at least one of the following data acquired by the data acquisition module through image and / or video acquisition: Eye movement data, writing trace data, answer trajectory data, correction trace data, facial expression data, sitting posture data, heart rate data, experimental operation process data, learning area lighting data, learning area noise data, and guardian guidance interaction data.

[0007] In some implementations, the data fusion module is also used to specifically perform: Based on the time and data collection scenario of each behavior sequence, determine the learning content label corresponding to each behavior sequence; Add the corresponding learning content label to each of the aforementioned behavior sequences; For at least one behavior sequence with the same learning behavior label, association information is generated based on the feature information of each behavior sequence to describe the association strength between different behavior sequences.

[0008] In some implementations, the digital modeling module is also used to specifically perform: Every first preset time interval, the current state parameters in the digital twin model are updated based on the association information mined from the learning context data obtained within the first preset time interval. Every second preset time interval, based on the association information mined from the learning context data obtained within the second preset time interval, the learning behavior parameters of multiple dimensions in the digital twin model are updated. The multiple learning behavior parameters include: cognitive level parameters, learning style parameters, emotional state parameters, and environmental support parameters; the second preset time interval is longer than the first preset time interval.

[0009] In some implementations, the personalized content generation module is also used to specifically perform: Based on the learning behavior parameters of the multiple dimensions, determine the display method of the tutoring information; Based on the aforementioned display method, tutoring information for the current learning content is generated according to the learning behavior parameters of the multiple dimensions. The tutoring information is displayed.

[0010] In some implementations, the personalized content generation module is also used to specifically perform: Based on the degree of matching between the learning context data and the digital twin model, optimization suggestions are determined for at least one data acquisition scenario; The optimization suggestion information is pushed to the target learner and / or the target learner's guardian.

[0011] Secondly, embodiments of this application provide a multi-scenario integrated learning tutoring method, which can be applied to the multi-scenario integrated learning tutoring system provided in embodiments of this application. The method includes: In at least one data acquisition scenario, learning context data related to the learning behavior of the target learner is collected through a data acquisition module; the at least one data acquisition scenario includes at least one of the following: on-campus, off-campus, home, and wearable devices; The learning context data is divided into multiple behavior sequences according to the time dimension, and the data fusion module is used to mine the correlation information between the behavior sequences based on the feature information of each behavior sequence. Using a digital modeling module, a digital twin model of the target learner is deployed based on each of the behavioral sequences and the associated information; the digital twin model is used to dynamically map the learning behavior of the target learner. Using the personalized content generation module, tutoring content information is generated based on the digital twin model to target the learning behavior of the learner.

[0012] Thirdly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the method of any one of the embodiments of this application.

[0013] Fourthly, embodiments of this application provide a computer program product, including a computer program, which, when executed by a processor, implements any of the methods described in the embodiments of this application.

[0014] Based on the aforementioned multi-scenario integrated learning tutoring system and method, storage medium, and program product, this application has at least the following beneficial effects or advantages: By collecting learning context data related to the learning behavior of target learners in at least one data collection scenario—in-school, out-of-school, at home, or wearable devices—using the data acquisition module, multimodal data related to student behavior across scenarios can be obtained. The data fusion module divides the learning context data into multiple behavioral sequences according to the time dimension. Based on the feature information of each behavioral sequence, the correlation information between these sequences is mined, enabling spatiotemporal fusion of cross-scenario, multimodal data. The digital modeling module deploys a digital twin model of the target learner based on each behavioral sequence and correlation information, dynamically mapping the learner's learning behavior. The fusion of data from different times and spaces more accurately depicts individual learning states. Furthermore, the personalized content generation module generates tutoring content information tailored to the learner's learning behavior based on the digital twin model. This allows for the generation of targeted tutoring content by combining the learner's individual behavioral profile, meeting the differentiated learning needs of different learners and improving their learning efficiency and knowledge mastery.

[0015] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application, it can be implemented according to the contents of the specification. In order to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below. Attached Figure Description

[0016] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the various drawings denote the same or similar parts or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings depict only some embodiments according to this application and should not be construed as limiting the scope of this application.

[0017] Figure 1 This illustration shows a schematic block diagram of a multi-scenario integrated learning tutoring system provided in an embodiment of this application. Figure 1 ; Figure 2 This application illustrates the flowchart of a multi-scenario integrated learning tutoring method provided in an embodiment. Figure 1 ; Figure 3 This illustration shows a schematic block diagram of a multi-scenario integrated learning tutoring system provided in an embodiment of this application. Figure 2 ; Figure 4 This illustration shows a portion of the process in a multi-scenario integrated learning tutoring method provided in an embodiment of this application. Figure 2 ; Figure 5 A block diagram of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0018] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the concept or scope of this application. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0019] To facilitate understanding of the technical solutions of the embodiments of this application, the relevant technologies of the embodiments of this application are described below. The following related technologies are optional solutions and can be arbitrarily combined with the technical solutions of the embodiments of this application, all of which fall within the protection scope of the embodiments of this application. It should be noted that the application scenarios or application examples provided in this application are for ease of understanding, and the embodiments of this application do not specifically limit the application of the technical solutions.

[0020] Furthermore, the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. The collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation portals are provided for users to choose to authorize or refuse.

[0021] This application's embodiments involve the intersection of educational information technology and multimodal data processing, and can be applied to the collection of multi-source data and cross-scenario (in-school classroom, home tutoring, and off-campus study tour) fusion modeling of learners in the K-12 stage (grades 1 to 12). Combined with virtual simulation and multimodal content generation technology, it forms a set of adaptive learning systems and methods, which are applicable to K-12 education scenarios such as online education in primary and secondary schools, smart classrooms, after-school tutoring, and off-campus study tours.

[0022] Currently, online education and personalized learning systems for the K-12 stage are insufficient to fully meet the actual needs of "teaching according to aptitude," especially in areas such as cross-scenario data collaboration, real-time status awareness, and generation of content adapted to different educational stages, where there are significant technical limitations. 1. Data fragmentation and lack of collaboration across K12 scenarios: K12 learner data is scattered across independent scenarios such as school, home, and off-campus. Related technologies lack standardized data collaboration mechanisms between these scenarios, resulting in fragmentation of personalized support in school learning, after-school reinforcement, and off-campus practice. The embodiments of this application can integrate the data from the above learning scenarios.

[0023] 2. Limited Multimodal Data Dimensions and Insufficient Adaptability: The "multimodal" technologies mentioned in related technical solutions are mostly limited to conventional academic data such as text, audio, and video, and do not include detailed behavioral data specific to K-12 learners (such as the correction trajectory of primary school students' homework and the experimental operation steps of junior high school students), physiological state data (such as heart rate changes after sitting for a long time in class and eye movement trajectory when concentrating on homework), and family environment data (such as the lighting in the learning area and the frequency of tutoring by guardians). This results in a lack of stage-specific judgment on the learner's state. However, the embodiments of this application can collect the above-mentioned learning context data, and can timely monitor a large amount of data lake information on the individual learning behavior of learners, which greatly improves the diversity and complexity of data sources.

[0024] 3. Static learner models and delayed intervention: In related technologies, the models designed for learners are mostly updated weekly or monthly, which cannot match the rapid changes in cognition and emotion of K-12 learners (especially in the lower grades); intervention measures are mostly based on historical performance for "remedial" adjustments, lacking "preventive" guidance based on real-time status. However, the embodiments of this application, through multi-granularity update frequencies, can provide timely feedback on the generation of tutoring content based on real-time monitoring update frequencies, and intervene in learners' learning behaviors in a timely manner.

[0025] 4. Insufficient Adaptability of Personalized Content Generation: The current content generation solution does not fully integrate the cognitive characteristics of K-12 students, nor does it deeply connect with learners' interests. The content presentation format is rigid and fails to meet the needs of learners at different grade levels for "fun" and "easy to understand," resulting in low participation among lower grade students and a feeling that the content is not targeted enough among higher grade students.

[0026] 5. Delayed interactive feedback in virtual scenarios: In online learning scenarios, existing systems mostly obtain learner responses through delayed methods such as "post-class questionnaires" and "clicking on multiple-choice questions," lacking instant feedback that conforms to the interaction habits of K-12 learners. This results in a significant gap between the interactive experience of virtual scenarios and offline classrooms, making it difficult to enhance the learning immersion.

[0027] Because learners at different K-12 levels have significantly different data dimensions and cognitive needs, the algorithms for cross-level data fusion and content adaptation are highly complex. For example, primary school focuses on behavioral data, while secondary school focuses on cognitive data, requiring different algorithms for processing. Furthermore, the correlation modeling between real-time physiological signals, detailed behavioral data, and K-12 learning states lacks support from grade-specific labeled datasets, making it difficult to accurately establish the correspondence between data and states.

[0028] This application's embodiments can overcome the shortcomings of existing K12 online learning systems in terms of cross-scenario data collaboration, grade level adaptability, and real-time interaction. Through K12-specific multimodal data collection and perception, grade level learner modeling, and customized content generation, it can achieve immersive learning support in K12 scenarios that fits the characteristics of the grade level and matches the learner's real-time status. At the same time, through compliant data collection and curriculum standard calibration, it ensures the system's feasibility and educational reliability.

[0029] The technical solution of this application and how it solves the aforementioned technical problems are described in detail below with specific embodiments. The listed specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0030] See Figure 1 The diagram shown illustrates a multi-scenario integrated learning tutoring system. This system specifically includes: Data acquisition module 101 is used to collect learning context data related to the learning behavior of the target learner in at least one data acquisition scenario; at least one data acquisition scenario includes at least one of the following: on-campus, off-campus, home, wearable; Data fusion module 102 is used to divide the learning context data into multiple behavior sequences according to the time dimension, and to mine the correlation information between the behavior sequences based on the feature information of each behavior sequence; The digital modeling module 103 is used to deploy a digital twin model of the target learner based on each behavioral sequence and related information; the digital twin model is used to dynamically map the learning behavior of the target learner. The personalized content generation module 104 is used to generate tutoring content information tailored to the learning behavior of the target learner based on the digital twin model.

[0031] The data acquisition module 101 can integrate interfaces from various data sources, including those within the school (such as classroom cameras, microphones, and academic affairs system interfaces), at home (such as learning tablets and environmental sensors), outside the school (such as research and study platform APIs), and authorized physiological sensing devices (such as smart bracelets), to acquire learning context data. Optionally, the data acquisition module 101 may include the aforementioned monitoring module for monitoring the raw learning context data, or it may exclude the monitoring module and only include the data source interface for acquiring learning context data. A data acquisition scenario can collect multiple types of learning context data, while a data source interface can acquire at least one type of learning context data. The data acquisition module 101 can also process the raw data acquired based on the data source interface to obtain the required learning context data. For example, if heart rate mutation data is required, after acquiring the heart rate monitoring curve, the moment when the heart rate difference between two adjacent heart rate sampling time points exceeds a threshold can be found to obtain the heart rate mutation data. It is understood that all data acquisition activities are conducted under the compliance constraints of the privacy protection module.

[0032] The data fusion module 102 can perform precise spatiotemporal alignment of multi-source heterogeneous data and deeply mine the inherent correlations between different modalities of data. In some implementations, each type of learning scenario data can be divided into at least one behavioral sequence in the time dimension according to a preset period (e.g., 5 minutes). When mining correlation information, at least one feature (e.g., learning content, course type, etc.) can be used as the correlation relationship to find related behavioral sequences existing in different times and spaces. For example, during time alignment, the NTP (Network Time Protocol) protocol can be used, and artificial intelligence models such as the Transformer model can be used to deeply mine the inherent correlations between different modalities of data.

[0033] The digital modeling module 103 can create or update a digital twin model of the target learner based on various behavioral sequences and related information. The digital twin model represents the target learner's learning behavior and can be represented by multiple dimensions of learning behavior parameters. For example, these parameters can include indicators in dimensions such as cognition, emotion, behavior, and social environment. Each indicator can be calculated based on the changes in the behavioral sequence over time and / or the related information of the behavioral sequences under different data collection scenarios. After creation, the digital twin model can be dynamically updated based on real-time monitored learning context data to map the target learner's dynamic learning behavior. Changes in the target learner will be reflected in changes in at least one dimension of the digital twin model's learning behavior parameters.

[0034] In some implementations, the personalized content generation module 104 can be divided according to dimensions such as academic year, semester, subject, and knowledge point, and pre-train different content generation models. The digital twin model is then input into the content generation model to generate targeted tutoring content information for the target learner. In other implementations, tutoring content with different feature tags can be provided in advance. The tutoring content is adapted to the learner whose tagged features are used. Thus, based on the target learner's digital twin model, tutoring content with the corresponding tags can be selected. It is understood that the above implementations are merely examples and are not intended to limit the embodiments of this application. Tutoring content generation can also include interactive information, such as a virtual teacher avatar interacting with students through video, audio, and text. It can generate content presentation methods that match students' content presentation preferences, and can also combine knowledge points with the target learner's interests to generate tutoring content, and push suggestions for the current learning environment, such as generating background noise and suggesting adjustments to the ambient volume.

[0035] This application embodiment uses a data acquisition module to collect learning context data related to the learning behavior of the target learner in at least one data acquisition scenario, including on-campus, off-campus, home, and wearable devices. This allows for the acquisition of multimodal data related to student behavior across scenarios. A data fusion module divides the learning context data into multiple behavioral sequences according to the time dimension. Based on the feature information of each behavioral sequence, the correlation information between these sequences is mined, enabling spatiotemporal fusion of cross-scenario, multimodal data. A digital modeling module deploys a digital twin model of the target learner based on each behavioral sequence and correlation information, dynamically mapping the learner's learning behavior. By fusing data from different times and spaces, it more accurately depicts the individual's learning state. Furthermore, a personalized content generation module generates tutoring content information tailored to the learner's learning behavior based on the digital twin model. This allows for the generation of targeted tutoring content based on the learner's individual behavioral profile, meeting the differentiated learning needs of different learners and improving their learning efficiency and knowledge mastery.

[0036] In some implementations, in on-campus data collection scenarios, at least one of the following authorized data collection modules is provided: classroom camera, classroom microphone, and academic affairs system data interface; in off-campus data collection scenarios, at least one of the following authorized data collection modules is provided: research and study platform data interface, tutoring institution camera, tutoring institution microphone, and tutoring institution system data structure; in home data collection scenarios, at least one of the following authorized data collection modules is provided: home camera, home microphone, student smart learning device, parent smart device application data interface, home light intensity sensor, and decibel meter; in wearable data collection scenarios, at least one of the following authorized data collection modules is provided: smart bracelet, smartwatch, and smart glasses.

[0037] After authorization, classroom cameras can continuously record students' facial micro-expressions and head turning angles. The system provided in this application (hereinafter referred to as "this system") can acquire facial images of the target learner every 5 seconds through a data source interface, determine the proportion of the target learner's gaze deviating from the blackboard, and when this proportion exceeds 40% for 3 consecutive minutes, a light vibration reminder can be generated on the target learner's wristband to indicate that the target learner has lost focus and needs to pay attention. The academic affairs system interface can simultaneously output classroom quiz scores to this system, and automatically add the question numbers below a threshold to the personalized after-class practice list.

[0038] The off-campus research and learning platform's data source interface transmits back the GPS tracks, altitude readings, and group peer review texts uploaded by students in the "Measuring Mountain Height" project. The personalized content generation module 104 can use the altitude change curve to generate a sine function fitting task, replacing the abstract concept of amplitude with a profile of the hillside the students have just traversed. This allows for the contextualized reproduction of knowledge points based on the students' recent practice. With authorization from students and parents, the off-campus tutoring institution's camera and microphone can capture the frequency of frowning and pauses in speech during one-on-one Q&A sessions. If the frowning frame rate suddenly increases and the pause exceeds 2 seconds, the system can display a prompt on the teacher's end stating "The student may not have understood; it is recommended to go back to the previous step," and simultaneously push an example problem with the same knowledge point but a smaller numerical value to the student's tablet. Understandably, the system's display can provide different interfaces for different roles. Teachers, students, and parents can download different software applications to log in and obtain various tutoring content information provided by the personalized content generation module 104.

[0039] In a home setting, students' smart learning devices (such as learning tablets) can record the pressure sensitivity, pen lift intervals, and number of corrections in their handwriting. Combined with real-time monitoring of ambient noise by a decibel meter, if the noise level exceeds 55dB and the number of corrections increases significantly, the parent's mobile app will receive an intervention notification stating "The environment is noisy; it is recommended to lower the volume." The original 15-minute practice session will be automatically split into two 8-minute tasks, with an eye exercise animation inserted in between. If the light intensity sensor detects that the desktop illuminance is below 300lx, it will trigger the smart desk lamp to increase its brightness and mark this low-illuminance period as a potential risk of eye strain. This data will be used to adjust the font size and background contrast for the next practice session. On the wearable side, the smart bracelet collects heart rate and skin conductance at a frequency of 10Hz. When the heart rate variability continues to decrease at night, the system judges it as "stress accumulation" and pushes a suggestion to parents the next morning: "Reduce extra practice today and prioritize reviewing wrong questions." The smartwatch records cadence and VO2 max during physical education class. Data that is more than 20% lower than the average for the same age is synchronized to the physical education teacher's end, automatically generating a personalized exercise prescription that reduces intensity and increases aerobic jogging. The inward-facing camera of the smart glasses records the eye saccade trajectory when reading. If the eye return rate is found to be more than 15% higher than the normal value, a micro icon of "Recommended to rest for 20 seconds" immediately pops up on the edge of the lens, and the eye movement data is marked as "potential risk of reading difficulties" for subsequent intervention by the psychological counselor.

[0040] By deploying at least one authorized data acquisition module in different scenarios, this system adds a unified timestamp to all previously scattered and differently formatted micro-data and injects it into the same computing pipeline, forming a minute-level three-dimensional perception of students' cognition, emotions, environment, and physiological state. This drives real-time intervention in four dimensions: content, rhythm, environment, and reminders, significantly improving the precision, timeliness, and adaptability of personalized tutoring to different educational stages.

[0041] In some implementations, the learning context data includes at least one of the following data collected by the data acquisition module through image and / or video acquisition: eye-tracking data, writing trace data, answer trajectory data, correction trace data, facial expression data, sitting posture data, heart rate data, experimental operation process data, learning area illumination data, learning area noise data, and guardian guidance interaction data.

[0042] For example, after obtaining authorization from the guardian, student, and school, the classroom camera captures upper body video of the target learner at 30fps. Visual algorithms extract the coordinates of key points on the shoulders, elbows, and wrists to obtain the head tilt angle and back curvature angle, which are then converted into a posture score. The posture score is used to determine the learner's learning status. For example, a high number of times and prolonged periods of hunching over may indicate insufficient sleep, prompting a reminder to parents about abnormal posture and suggesting the student go to bed earlier. When the student logs into the learning tablet to view tutoring videos, the tablet can capture facial video of the student. This video is input into an expression classification model, selecting from three expression tags: "confused / focused / distracted." When the probability of a "confused" expression is >0.7 and lasts for 8 seconds, the system automatically reverts to the previous explanation and inserts a 30-second interactive question, allowing the student's doubts to be resolved promptly.

[0043] For another example, in a home setting, a smart learning device for students might have a built-in electromagnetic pen that records the strokes of a student's pen while writing homework. If there are many corrections, it indicates a lower level of understanding, and relevant video explanations of the related knowledge points can be pushed to the student to prevent accumulating doubts and to increase the frequency of similar questions in the future to reinforce learning. If the student's pen strokes are quick and there are few corrections, it indicates a higher level of understanding, and the frequency of similar questions can be reduced.

[0044] This implementation method acquires raw data, processes it into learning context data, divides it into behavior sequences according to time periods, and then uses the behavior sequences and the correlation information between different behavior sequences to generate targeted tutoring content information in a timely manner. This enables the system to complete a closed loop from learning state perception to learning behavior intervention execution, significantly improving the granularity, timeliness and health protection level of personalized tutoring.

[0045] In some implementations, the data fusion module 102 is also used to specifically perform: determining the learning content label corresponding to each behavior sequence based on the time and data acquisition scenario of each behavior sequence; adding the corresponding learning content label to each behavior sequence; and generating association information to describe the association strength between different behavior sequences based on the feature information of each behavior sequence for at least one behavior sequence with the same learning behavior label.

[0046] For example, in a fraction multiplication lesson, the classroom camera continuously records the learner's (student's) head movements (looking up, looking down, and frowning), while the learning tablet simultaneously captures the pen tip's markings and corrections. The academic system returns a score of 0.6 for that question. The data fusion module first aligns the three data streams according to a unified timestamp, then affixes the same learning content label: the first-level learning content label is "fraction multiplication," and the second-level learning content label is "(specific knowledge point involved in the third example question)." This system extracts the frequency of frowning, the number of corrections, and the score, calculating the correlation weight. This indicates that frowning plus corrections is highly correlated with incorrect answers. This weight is written into the digital twin model. The next time, when the learner exhibits the same behavioral sequence combination for other knowledge points, the teacher receives a prompt message indicating that the student has not fully mastered the knowledge point labeled with that learning content, allowing the teacher to further explain that knowledge point.

[0047] In other examples, during homework time, the learning tablet recorded students repeatedly watching a video lesson. The smart bracelet worn by the student reported an increased heart rate, and the decibel meter showed noise levels exceeding 55 decibels. These three behavioral sequences overlapped in time and shared the same learning content label, leading to their merging into a single learning content event. The system calculated that the association weights between noise and playback, as well as between noise and heart rate, both increased, confirming that environmental noise was a common factor contributing to distraction and irritability while watching videos. This information was written into the digital twin model, and parents immediately received a volume reduction reminder. The system automatically increased the subsequent audio gain by six decibels and added subtitles to improve student attention.

[0048] Through a dual alignment mechanism of time alignment and learning content tag alignment, learning behavior data that was originally scattered across different scenarios is integrated into a computable causal chain. The digital twin model obtains the correlation weights of behavior sequences across scenarios or across time, which can drive the dual adjustment of tutoring content and environment, significantly reducing the loss of learning efficiency caused by external interference or student questions.

[0049] In some implementations, the digital modeling module 103 is also used to specifically perform: every first preset time interval, updating the current state parameters in the digital twin model based on the association information mined from the learning context data obtained within the first preset time interval; every second preset time interval, updating the learning behavior parameters of multiple dimensions in the digital twin model based on the association information mined from the learning context data obtained within the second preset time interval, the multiple learning behavior parameters including: cognitive level parameters, learning style parameters, emotional state parameters, and environmental support parameters; the second preset time interval is longer than the first preset time interval.

[0050] For example, the initial preset duration could be 5 minutes. In real-time updates every 5 minutes, the system inputs newly captured frowning frame rate from the classroom camera, the number of pauses in handwriting on the tablet, and heart rate variability from the wristband into the digital twin model, updating the model's current state parameters: focus level decreases from 0.71 to 0.58, and confusion level increases to 0.65. This update triggers the personalized content generation module to lower the difficulty of subsequent example questions and insert a 30-second interactive question, allowing students to complete error correction within the same lesson and avoiding a backlog of problems.

[0051] The second preset duration can be one day. Within the daily update cycle at midnight, the system summarizes the relevant information accumulated over 24 hours: the cognitive level parameter shows that the mastery of "trigonometric function amplitude" has improved from 62 points to 78 points; the learning style parameter confirms that the visual type weight remains at 0.82; the emotional state parameter reflects that the average anxiety value has decreased by 0.15; and the environmental support parameter reveals that after the period of household noise shifted from 7 pm to 8 pm, the accuracy rate improved by 8%. Based on the update of these four-dimensional parameters, the system reschedules the students' assignments, shifting the focus of tomorrow night's homework from "amplitude" to "cycle," and pushes a 3-minute "cycle-clock swing" animation, achieving adaptive optimization of the learning path at the 24-hour level.

[0052] By linking the first and second preset durations in a dual-cycle manner, we can ensure zero lag in learning behavior intervention, prevent daily repetitive practice and difficulty jumps, improve overall learning efficiency, reduce ineffective homework time, and maintain students' appropriate challenge and emotional stability.

[0053] Furthermore, in some implementations, the personalized content generation module 104 is also used to specifically perform: determining the display method of tutoring information based on learning behavior parameters of multiple dimensions; generating tutoring information for the current learning content based on the display method and learning behavior parameters of multiple dimensions; and displaying the tutoring information.

[0054] For example, this system can tailor the presentation of the concept of "slope of a linear function" to "high visualization + low text density" based on students' learning styles (visual learners). It can then use a generative model to output a 2-minute hand-drawn animation: comparing the slope using the speed of a skateboard going downhill, with only key words in the voiceover and the subtitle font size increased by 20%. After the animation, three drag-and-drop image questions are presented. The system adjusts the presentation based on students' heart rate and answer trajectory, achieving a dual regulation of learning difficulty and student mood. By adjusting the presentation method, student attention loss due to mismatched presentation formats can be significantly reduced.

[0055] In some implementations, the personalized content generation module is also used to specifically perform: determining optimization suggestions for at least one data collection scenario based on the degree of matching between the learning context data and the digital twin model; and pushing the optimization suggestions to the target learner and / or the target learner's guardian.

[0056] For example, during off-campus study tours, if the smart bracelet's heart rate changes and GPS signal indicate that students have walked a long distance and expended too much energy, reaching the "high-load" threshold in the digital twin model, this system can send suggestions to the accompanying teachers and parents, such as "shorten the next mountain path by 200 meters and add a 10-minute rest." Based on real-time matching, proactively pushing scenario optimization suggestions can reduce distractions in the learning environment, improve students' learning state and interest, and help them maintain a moderate level of challenge and emotional stability.

[0057] The multi-scenario integrated learning tutoring system provided in some embodiments of this application can automatically collect multimodal learning data of K12 learners in different scenarios. It achieves spatiotemporal alignment of K12 in-school, home, and out-of-school scenario data through NTP time calibration. Combined with Transformer / RNN models, it mines the correlations between data from different scenarios, forming a unified data view covering the K12 stage, solving the problem of data fragmentation across scenarios. Based on the integrated data, it constructs and updates the digital twin model of the target learner through a multi-granularity update mechanism. This "real-time + daily + milestone" multi-granularity update mechanism matches the rapidly changing state of K12 learners, avoiding model staticization. The digital twin model is configured with a four-dimensional indicator system of "cognition – emotion – behavior – social environment," and the indicators are adapted to the characteristics of the K12 stage. Therefore, based on the current state of the digital twin model and the current curriculum standards, it can generate and output personalized learning content and feedback in a targeted manner. Based on a personalized content generation model developed for the cognitive characteristics of different K12 stages, it dynamically generates personalized content and virtual interactive feedback that conform to the learner's cognitive characteristics and interests, enhancing learning immersion and relevance.

[0058] This application also provides a multi-scenario integrated learning tutoring method, which can be applied to the multi-scenario integrated learning tutoring system provided in this application embodiment. (See reference...) Figure 2 The method includes the following steps: Step 201: In at least one data acquisition scenario, collect learning context data related to the learning behavior of the target learner through the data acquisition module; at least one data acquisition scenario includes at least one of the following: on-campus, off-campus, home, wearable; Step 202: Divide the learning context data into multiple behavior sequences according to the time dimension, and use the data fusion module to mine the correlation information between the behavior sequences based on the feature information of each behavior sequence; Step 203: Using the digital modeling module, deploy a digital twin model of the target learner based on each behavioral sequence and related information; the digital twin model is used to dynamically map the learning behavior of the target learner. Step 204: Using the personalized content generation module, generate tutoring content information tailored to the learning behavior of the target learner based on the digital twin model.

[0059] It is understood that the specific implementation methods of each step in the method provided in this application embodiment can be found in the corresponding description in the above system, and have corresponding beneficial effects, which will not be repeated here.

[0060] The following is combined Figure 3 and Figure 4 This application provides an example of how the embodiments of this application are applied in some specific implementations. Taking a second-year junior high school student A as an example, the subject is mathematics, the knowledge point is trigonometric function graphs; the learner's characteristics are a visual learning style, likes mountain climbing, and has a weak foundation in trigonometric function concepts.

[0061] refer to Figure 3 This application provides an architecture diagram of a multi-scenario integrated learning tutoring system based on some embodiments, specifically demonstrating the four core layers and module composition of the system: The multimodal data acquisition layer (corresponding to data acquisition module 101) integrates interfaces from various data sources, including those within the school (such as classroom cameras, microphones, and academic affairs system interfaces), homes (such as learning tablets and environmental sensors), external sources (such as research and study platform APIs), and authorized physiological sensing devices (such as smart bracelets). All data acquisition activities are conducted under the compliance constraints of the privacy protection module and after authorization by the authorization module.

[0062] The cross-scenario data fusion management layer (corresponding to data fusion module 102) consists of the NTP time calibration module and the Transformer fusion analysis module. The former is responsible for accurate spatiotemporal alignment of multi-source heterogeneous data, while the latter is responsible for in-depth mining of the intrinsic correlations between different modalities of data.

[0063] The digital twin modeling layer (corresponding to digital modeling module 103) is responsible for constructing and dynamically updating the learner's digital twin. This twin includes indicators across four dimensions: cognition, emotion, behavior, and social environment. It employs a multi-granularity update mechanism (real-time, daily, milestone) to match the rapid changes in the K-12 learner's status. Real-time updates can be configured to smaller granularities such as 5 minutes or 15 minutes, daily updates are performed daily, and milestone updates are used to record significant change points.

[0064] The multimodal content generation layer (corresponding to personalized content generation module 104) includes a core content generation model customized for K-12 education, a curriculum standard calibration module, and a virtual tutor interaction module. These components work together to generate content and interactions that meet both teaching requirements and the learner's individual needs, based on the virtual twin's status.

[0065] refer to Figure 4 This demonstrates a portion of the workflow of a multi-scenario integrated learning tutoring method. This example method applies... Figure 3 The multi-scenario integrated learning and tutoring system shown.

[0066] Step 1: After starting, proceed with K12 multi-source data parallel acquisition.

[0067] The process begins with the parallel collection of data from "on-campus," "home," and "off-campus" scenarios, as well as "physiological data." NTP time calibration ensures the temporal consistency of all data. Transformer fusion analysis performs in-depth processing on the aligned data, outputting quantitative conclusions such as "home environmental noise is positively correlated with homework hesitation behavior, and further correlated with classroom confusion expressions," providing direct evidence for subsequent modeling.

[0068] Internal school data: Through intelligent sensing devices in the classroom, learner A's behavioral data in the trigonometric function class is automatically collected, including but not limited to: analyzing his facial expressions and the duration of his gaze deviating from the screen through the camera; recording the changes in his speaking speed and the frequency of his pauses when answering questions through the microphone; and automatically obtaining his answer trajectory and accuracy rate of in-class tests from the school's academic affairs system interface.

[0069] Home Data: Through the smart learning tablet used by learner A, detailed behavioral data of completing trigonometric function homework is automatically recorded, such as time spent on each question, correction marks, and number of pauses; the light intensity and noise level in the learning area are automatically collected through home environment sensors; the system automatically analyzes and records tutoring interaction patterns through the platform data interface, eliminating the need for manual uploading.

[0070] External data: By connecting with the external study tour platform through the system API interface, detailed observation records and results data of learner A's recent "outdoor measurement" study tour activities can be automatically obtained.

[0071] Physiological data: After obtaining explicit authorization from learner A and their guardian through the system's privacy authorization module, the physiological signal changes during the problem-solving process, such as heart rate fluctuations and the frequency of posture adjustments, are collected through the smart bracelet worn by learner A.

[0072] Step 2: Pass the privacy protection compliance check. If the authorization fails, the data collection process will terminate; otherwise, if the authorization is successful, the process will continue.

[0073] Step 3: Perform data preprocessing on the collected data, standardize the format, and handle outliers, such as smoothing or removing them.

[0074] Step 4: Perform NTP spatiotemporal alignment on the collected data. Use an NTP server to calibrate the timestamps of all data, unifying classroom data, homework data, and study tour data to the UTC+8 time zone. Simultaneously, add a "Trigonometric Function Learning" tag to all data to ensure a strong correlation between the data and the target knowledge points, avoiding interference from irrelevant data.

[0075] Step 5: Divide the time-aligned collected data into small blocks according to time, process them to obtain multiple sequences, and use the feature extraction module to extract feature vectors. For example, the extracted feature vectors may include: facial expression features (128 dimensions), behavioral features (64 dimensions), physiological features (32 dimensions), and environmental features (16 dimensions). Step 6: The feature vectors are fed into the Transformer fusion analysis module. The input dimensions are facial expression features (128 dimensions), behavioral features (64 dimensions), physiological features (32 dimensions), and environmental features (16 dimensions). Through deep learning, a "state association matrix" (10×10) is output.

[0076] This process can uncover meaningful associations, such as: "When the noise level in the home learning environment reaches 55 decibels, the number of times homework is paused increases significantly (association weight 0.75). This behavioral pattern is strongly associated with frowning expressions (association weight 0.82) and increased heart rate (association weight 0.78) observed in the classroom."

[0077] Step 7: Control data quality. This can be done using a threshold. If the data exceeds the threshold, it is considered unqualified. Repair the data and return to Step 3, or return to Step 1 (connected to Step 1). Figure 4 (Not shown in the image) Data re-acquisition.

[0078] Step 8: If the quality of Step 7 is satisfactory, a unified data view of K12 can be generated. K12 can use the same view to display all or part of the data obtained after the above collection and / or calculation.

[0079] Step 9: Call the downstream application interface. The downstream application obtains the alignment data and association matrix, and updates the digital twin model. The update frequency can include real-time updates and daily updates. Real-time updates utilize an RNN model to process real-time streaming data from the classroom, updating the twin's instantaneous state every 5 seconds, for example, labeled as: "Current emotional state: Anxiety (confidence 0.65); Cognitive state: Confused about the concept of trigonometric function amplitude (confidence 0.8)". Daily updates are triggered daily at set times using Transformer-based aggregation analysis to update the four-dimensional indicators. For example: Cognitive dimension (trigonometric function mastery score 35 / 100, weakness concentrated in "the relationship between amplitude and image"); Behavioral dimension (learning style is visual, matching degree 80%; lack of derivation process records in homework habits, score 40 / 100); Social environment dimension (family learning environment support score, based on automatically collected environmental data and interaction pattern analysis).

[0080] After updating the digital twin model, Figure 3 The multimodal content generation layer of the middle system can generate an input set containing 18 key parameters based on the real-time state of the digital twin produced in the above steps, combined with the requirements for trigonometric function graphs in the second-year junior high school mathematics curriculum standards. It then calls the junior high school multimodal content generation model, specifically achieving the following output: Personalized content generation: Given learner A's interest in mountain climbing and visual learning style, the model draws an analogy between the amplitude concept of the sinx graph and the slope of a mountain climb. For example, a 2-minute and 30-second immersive animation is generated, using photos of steep slopes taken by A during their study tour, corresponding to y=2sinx (large amplitude, high peak), and photos of gentle slopes, corresponding to y=0.5sinx (small amplitude, gentle peak). The animation explains the steps of "amplitude definition → slope analogy → graph changes".

[0081] Environmental intervention and virtual interaction: The system can automatically push soothing background music and send environmental optimization suggestions to the guardian's terminal (such as "Current noise level is 55 decibels, it is recommended to adjust it to below 50 decibels"). The virtual tutor appears with a "youthful outdoor style". When A asks "How do slope and amplitude correspond?" via voice, the system generates and presents a voice and text response based on the user's personal experience within 1.5 seconds, such as "The steep peaks you climbed are like high waves, and the gentle slopes are like low waves", and highlights the key formula y=Asinx.

[0082] Tiered practice: The system first generates basic matching questions. When the accuracy rate of A reaches the standard (≥80%), it automatically pushes advanced application questions, such as "Draw the corresponding sine function graph based on the slope change diagram of a mountain climbing route", so as to deepen understanding in an interesting context.

[0083] Feedback and System Optimization: The system continuously monitored A's learning process, recording that A's single learning session duration increased to 45 minutes, A asked questions proactively twice, and anxiety was reduced according to frowning facial expressions and heart rate indicators. The accuracy rates of basic and advanced exercises improved to 85% and 78%, respectively. At the same time, the system recorded that the noise level in the home learning environment decreased to 45 decibels after intervention.

[0084] Based on this positive feedback data, the system optimizes the parameters of the relevant models. For example, it increases the weights of the parameters "interest embedding depth" and "visualization level" in the content generation model, making the generated content more aligned with A's preferences.

[0085] The successful "Mountain Climbing Interest - Amplitude Concept" teaching strategy has been stored in the K12 teaching strategy library for intelligent recommendations when encountering learners with similar characteristics in the future, enabling the reuse of experience and the continuous evolution of the system.

[0086] For the above example, the following alternative technical solutions are provided: Alternative data acquisition solutions: If smart bracelets are unavailable, the front-facing camera of a learning tablet can be used to analyze facial micro-expressions using computer vision technology (such as the OpenCV framework) to infer anxiety levels. If environmental sensors are missing, the quality of the learning environment can be indirectly assessed by analyzing task behavior data.

[0087] Alternative model fusion solution: In scenarios with limited computing resources, the more computationally efficient LSTM model can be used to replace the Transformer for data fusion. Although the accuracy of association mining may decrease slightly, the number of model parameters and memory usage can be significantly reduced, making it more suitable for resource-constrained terminal devices.

[0088] Personalized content generation alternative: If real-time rendering of high-precision animations is difficult, a "pre-made template + parametric adjustment" approach can be adopted. A core animation template is pre-made, and then materials and text annotations are replaced according to the learner's specific interests. This can significantly shorten response time and improve the user experience in low-bandwidth environments.

[0089] Virtual Interaction Alternative: In environments where voice interaction is unavailable or network latency is high, a text-based interactive interface can be enabled. The system generates illustrated answers with detailed explanations, diagrams, and formula annotations to ensure the availability of core interactive functions.

[0090] This application provides a multi-scenario integrated learning tutoring system in some embodiments, covering multiple scenarios in and out of school during the K-12 stage, automatically collecting multi-dimensional data tailored to the characteristics of each educational stage. In the school scenario, it collects classroom facial expressions, hand-raising frequency, test answer trajectories, and experimental operation videos; in the home scenario, it automatically collects homework writing traces, guardian tutoring interaction records, and light / noise levels in the learning area via a smart learning terminal; in the off-campus scenario, it collects observation records of study tours and incorrect answers from training institutions. It also incorporates stage-specific physiological data (such as changes in classroom posture for lower-grade students and eye movement trajectories for higher-grade students when focusing on homework). Through the fusion and analysis of multimodal data using deep learning models such as Transformer and RNN, it infers the learner's stage-appropriate fit in real time and constructs a learner profile model that can be updated as the learning progresses based on continuous K-12 stage data. Based on learners' real-time status and K-12 profiles, the system utilizes a multimodal content generation model customized for educational scenarios to generate content tailored to different grade levels. For example, it generates an "animated character shopping" scenario for elementary school students learning multiplication, and an "experiment step breakdown" animation for middle school physics. Simultaneously, it reconstructs content by embedding interest, such as integrating physics formulas into "home experiments" for students who enjoy science experiments. It also generates K-12 exclusive interactive elements, such as "voice Q&A + cartoon feedback" for elementary school and "experiment questions + step demonstrations" for middle school. To ensure educational reliability, the system calibrates the generated content against K-12 curriculum standards to avoid exceeding the curriculum or being insufficiently challenging. It also optimizes the generation strategy based on teacher feedback, balancing "fun" and "knowledge accuracy." The system continuously tracks K-12 learners' interaction with the generated content and their learning outcomes, re-inputting multimodal feedback data into the model to optimize content adaptability. It also updates continuous learning profiles across the K-12 stage, supporting personalized support for the "in-school-home-out-of-school" transition.

[0091] Figure 5 This is a block diagram of an electronic device used to implement the embodiments of this application. The electronic device can run the multi-scenario fusion learning tutoring method provided in the embodiments of this application, and can be applied to the multi-scenario fusion learning tutoring system provided in the embodiments of this application. Figure 5 As shown, the electronic device includes a memory 501 and a processor 502. The memory 501 stores a computer program that can run on the processor 502. When the processor 502 executes the computer program, it implements the method described in the above embodiments. The number of memories 501 and processors 502 can be one or more. In a specific implementation, the electronic device may also include a communication interface 503 for communicating with external devices and exchanging data.

[0092] In practical implementation, if the memory 501, processor 502, and communication interface 503 are implemented independently, they can be interconnected via a bus to communicate with each other. This bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. This bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0093] Optionally, in a specific implementation, if the memory 501, processor 502 and communication interface 503 are integrated on a single chip, the memory 501, processor 502 and communication interface 503 can communicate with each other through an internal interface.

[0094] This application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method provided in this application.

[0095] This application provides a computer program product, including a computer program that, when executed by a processor, implements the method provided in this application.

[0096] This application also provides a chip including a processor for calling and executing instructions stored in a memory, causing a communication device with the chip installed to perform the method provided in this application.

[0097] This application also provides a chip, including: an input interface, an output interface, a processor, and a memory. The input interface, output interface, processor, and memory are connected through an internal connection path. The processor is used to execute code in the memory. When the code is executed, the processor is used to execute the method provided in the application embodiment.

[0098] It should be understood that the aforementioned processor can be a CPU (Central Processing Unit), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. General-purpose processors can be microprocessors or any conventional processor. It is worth noting that the processor can be a processor supporting Advanced Reduced Instruction Set Machines (ARM) architecture.

[0099] Further, optionally, the aforementioned memory may include read-only memory and random access memory. The memory may be volatile memory or non-volatile memory, or may include both. Non-volatile memory may include read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory may include random access memory (RAM), which serves as an external cache. By way of example, but not limitation, many forms of RAM are available. Examples include Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced Synchronous DRAM (ESDRAM), Sync Link DRAM (SLDRAM), and Direct Rambus RAM (DR RAM).

[0100] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions according to this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another.

[0101] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0102] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "a plurality of" means two or more, unless otherwise explicitly specified.

[0103] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process. Furthermore, the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functionality involved.

[0104] The logic and / or steps described in the flowchart or otherwise herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus or device (such as a computer-based system, a processor-included system or other system that can fetch and execute instructions from, an instruction execution system, apparatus or device).

[0105] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. All or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware, the program being stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiments.

[0106] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. This storage medium can be a read-only memory, a disk, or an optical disk, etc.

[0107] The above description is merely an exemplary embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope described in this application, and these should all be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A multi-scenario fusion learning tutoring system, characterized in that, include: The data acquisition module is used to collect learning context data related to the learning behavior of the target learner in at least one data acquisition scenario; The at least one data collection scenario includes at least one of the following: on campus, off campus, home, and wearable devices; The data fusion module is used to divide the learning context data into multiple behavior sequences according to the time dimension, and to mine the correlation information between the behavior sequences based on the feature information of each behavior sequence. The data fusion module is also used to specifically perform: determining the learning content label corresponding to each behavior sequence according to the time and data collection scenario of each behavior sequence; adding the corresponding learning content label to each behavior sequence, including aligning according to a unified timestamp and then affixing the same learning content label, and integrating the originally scattered learning behavior data into a computable causal chain through a dual alignment mechanism of time alignment and learning content label alignment; for at least one behavior sequence with the same learning content label, generating the correlation information describing the correlation strength between different behavior sequences based on the feature information of each behavior sequence. A digital modeling module is used to deploy a digital twin model of the target learner based on each behavioral sequence and the associated information. The digital twin model is used to dynamically map the learning behavior of the target learner and obtain the association weights of cross-scenario or cross-time behavioral sequences. The digital modeling module is also used to specifically perform the following: every first preset duration, updating the current state parameters in the digital twin model based on the associated information mined from the learning context data obtained within the first preset duration; every second preset duration, updating multiple dimensions of learning behavior parameters in the digital twin model based on the associated information mined from the learning context data obtained within the second preset duration, the multiple dimensions of learning behavior parameters including: cognitive level parameters, learning style parameters, emotional state parameters, and environmental support parameters; the second preset duration is longer than the first preset duration. The personalized content generation module is used to generate tutoring content information tailored to the learning behavior of the target learner based on the digital twin model.

2. The system of claim 1, wherein, In the aforementioned on-campus data collection scenario, at least one of the following authorized data collection modules is provided: classroom camera, classroom microphone, and academic affairs system data interface; In the off-campus data collection scenario, at least one of the following authorized data collection modules is provided: research and study platform data interface, tutoring institution camera, tutoring institution microphone, and tutoring institution system data structure; In the data collection scenario of the home, at least one of the following authorized data collection modules is provided: home camera, home microphone, student smart learning device, parent smart device application data interface, home light intensity sensor, and decibel meter; In the wearable data collection scenario, at least one of the following authorized data collection modules is provided: smart bracelet, smartwatch, and smart glasses.

3. The system of claim 1, wherein, The learning context data includes at least one of the following data collected by the data acquisition module through image and / or video acquisition: Eye movement data, writing trace data, answer trajectory data, correction trace data, facial expression data, sitting posture data, heart rate data, experimental operation process data, learning area lighting data, learning area noise data, and guardian guidance interaction data.

4. The system of claim 1, wherein, The personalized content generation module is also used for specific execution: Based on the learning behavior parameters of the multiple dimensions, determine the display method of the tutoring information; Based on the aforementioned display method, tutoring information for the current learning content is generated according to the learning behavior parameters of the multiple dimensions. The tutoring information is displayed.

5. The system of claim 1, wherein, The personalized content generation module is also used for specific execution: Based on the degree of matching between the learning context data and the digital twin model, optimization suggestions are determined for at least one data acquisition scenario; The optimization suggestion information is pushed to the target learner and / or the target learner's guardian. 6.A multi-scene fusion learning tutoring method, characterized in that, The method is applied to the multi-scenario integrated learning tutoring system according to any one of claims 1-5, and the method includes: In at least one data acquisition scenario, learning context data related to the learning behavior of the target learner is collected through a data acquisition module; the at least one data acquisition scenario includes at least one of the following: on-campus, off-campus, home, and wearable devices; The learning context data is divided into multiple behavior sequences according to the time dimension. A data fusion module is used to mine the correlation information between the behavior sequences based on the feature information of each behavior sequence. Specifically, this includes: determining the learning content label corresponding to each behavior sequence according to the time and data collection scenario of each behavior sequence; adding the corresponding learning content label to each behavior sequence, including aligning according to a unified timestamp and then affixing the same learning content label. Through a dual alignment mechanism of time alignment and learning content label alignment, the learning behavior data in the originally scattered scenarios are integrated into a computable causal chain; for at least one behavior sequence with the same learning content label, the correlation information describing the correlation strength between different behavior sequences is generated based on the feature information of each behavior sequence. Using a digital modeling module, a digital twin model of the target learner is deployed based on each behavioral sequence and the associated information. The digital twin model dynamically maps the learning behavior of the target learner and obtains associated weights for cross-scenario or cross-time behavioral sequences. The digital modeling module is further configured to: update the current state parameters in the digital twin model every first preset duration based on the associated information mined from the learning context data obtained within the first preset duration; and update multiple dimensions of learning behavior parameters in the digital twin model every second preset duration based on the associated information mined from the learning context data obtained within the second preset duration. These multiple dimensions of learning behavior parameters include: cognitive level parameters, learning style parameters, emotional state parameters, and environmental support parameters. The second preset duration is longer than the first preset duration. Using the personalized content generation module, tutoring content information is generated based on the digital twin model to target the learning behavior of the learner.

7. A computer readable storage medium characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method of claim 6.

8. A computer program product, characterised in that, Includes a computer program that, when executed by a processor, implements the method according to claim 6.

Citation Information

Patent Citations

  • Learner model dynamic construction method based on education large data

    CN108664614A

  • Multi-space fusion learning environment construction method and device

    CN110598770A

  • Homework data feedback method based on intelligent question bank and learning evaluation system

    CN117557426A

  • Intelligent learning behavior monitoring and abnormity early warning method and device

    CN119723639A

  • Teaching information processing method and system based on digital twinning

    CN120146367A