Student learning track recording and personalized service system based on scene memory
By using a learning trajectory recording system based on contextual memory, the system enables the associated storage and deep understanding of learning trajectories across multiple scenarios, generating precise personalized services. This solves the problem that existing technologies struggle to reflect learning correlation patterns and provide insufficient service accuracy, thereby improving learning efficiency.
Patent Information
- Application Number
- CN202511708027.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-20
- Publication Date
- 2026-03-06
AI Technical Summary
Existing learning trajectory recording systems are unable to reflect the learning correlation patterns of students in different scenarios, and personalized services lack in-depth exploration and have limited service accuracy.
A student learning trajectory recording system based on contextual memory is adopted, which realizes multi-scenario trajectory association recording and accurate service through status acquisition devices, feature extraction devices, and service generation devices. The status acquisition devices receive learning data in real time and extract learning status data, the feature extraction devices perform three-dimensional target feature extraction, and the service generation devices generate personalized service suggestions.
It achieves a complete reconstruction of the student's learning landscape, provides a three-dimensional data foundation, generates personalized services that meet the actual needs of students, and improves learning efficiency.
Smart Images
Figure CN121614770A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of interdisciplinary technology of artificial intelligence and education, and more specifically, to a student learning trajectory recording and personalized service system based on contextual memory. Background Technology
[0002] A learning trajectory refers to the changing pattern of a learner's (usually a student's) learning state over time. Because learning states are multidimensional, a learning trajectory is essentially a multidimensional vector function of time. Learning states include learning scenarios, learning strategies, learning environments, and learning content. Learning trajectories are crucial for assessing learners' learning abilities, personalities, and habits, and are essential for personalized learning progress diagnosis and ability improvement. Therefore, existing technologies have disclosed various student learning trajectory recording systems built specifically for learning trajectories.
[0003] A student learning trajectory recording system is an educational management tool that uses digital means to systematically record students' learning process, behavior, and growth path. When such a system is running, it first collects students' learning process data to build student growth profiles. Then, it conducts cognitive diagnosis on students based on the content of the constructed growth profiles and visualizes the diagnostic results to achieve personalized feedback.
[0004] However, existing student learning trajectory recording systems often focus on single scenarios (e.g., homework systems only record homework data, and classroom interaction systems only record answer responses), resulting in fragmented trajectory data that fails to reflect the correlation patterns in students' learning across different scenarios (e.g., the impact of classroom learning effectiveness on homework completion quality). Furthermore, personalized services are often based on the mastery of single knowledge points, lacking in-depth analysis of students' evolving learning abilities and scenario adaptation characteristics, thus limiting service accuracy. Therefore, a system that incorporates episodic memory theory to achieve multi-scenario trajectory correlation recording and precise service is needed. Summary of the Invention
[0005] The technical problem to be solved by the present invention is how to overcome the technical defects of existing learning trajectory recording systems, which are difficult to reflect the learning correlation patterns of students in different scenarios and whose service direction is difficult to meet the actual needs. In order to overcome this defect, the present invention provides a student learning trajectory recording and personalized service system based on contextual memory.
[0006] This invention provides a student learning trajectory recording and personalized service system based on contextual memory, including... A status acquisition device is used to receive student learning data output by the learning platform in real time and extract learning status data from the student learning data. The feature extraction device communicates with the state acquisition device and is used to extract three-dimensional target features from the learning state data based on a deep learning algorithm to obtain three-dimensional target features; the three-dimensional targets are knowledge and skills targets, process and method targets, and emotion, attitude and value targets; The service generation device communicates with the feature extraction device to extract the gap between the three-dimensional target features and the teacher's predetermined target features, and uses a large language model to generate personalized service suggestions adapted to the corresponding students based on the obtained gap; the personalized service suggestions include learning resource recommendations, ability diagnostic reports, and learning method suggestions.
[0007] The student learning trajectory recording and personalized service system based on contextual memory disclosed in this invention addresses the technical problems of this invention by setting up a state acquisition device, a feature extraction device, and a service generation device. The state acquisition device extracts learning state data, which includes learning scenarios, learning strategies, learning environments, and learning content, thus breaking through the limitations of single scenarios. This not only enables context-related storage, fully reconstructing the student's learning landscape in classrooms, assignments, and exams, but also provides a three-dimensional data foundation for personalized services. Simultaneously, the feature extraction device extracts three-dimensional target features, extracting features from multiple dimensions such as knowledge and skills, processes and methods, and emotions, attitudes, and values. This goes beyond the superficial analysis of traditional "knowledge point mastery" and achieves a deeper understanding of students' learning patterns. Finally, the service generation device combines learning preferences, scenario patterns, and ability characteristics from contextual memory to generate service content that better meets students' actual needs, improves learning efficiency, and provides teachers with ability diagnostic reports containing scenario details, assisting teachers in providing targeted tutoring. This overcomes the technical shortcomings of existing technologies, which struggle to reflect the learning correlation patterns of students in different scenarios and whose service directions fail to meet real-world needs.
[0008] In one possible implementation, the status acquisition device includes: The receiver is used to receive in real time learning process data, learning environment data, and learning content data paired with the learning process data and the learning environment data output by the learning platform. A policy recognizer communicates with the receiver and is used to retrieve learning policies from the policy pool that are compatible with the learning process in the learning process data to obtain learning policy data. The scene recognizer communicates with both the policy recognizer and the feature extraction device to perform scene recognition on the learning policy data, the learning environment data, and the learning content data to obtain the learning state data.
[0009] The state acquisition device with the above structure and functions can identify learning strategies through a strategy recognizer and mark scenes through a scene recognizer to obtain complete learning state data. This further breaks through the limitation of a single scene, helps to meet the requirements of context-related storage, fully restores the student's learning in classroom, homework, exam and other scenarios, realizes panoramic recording of learning trajectory, and provides a three-dimensional data foundation for personalized services.
[0010] In one possible implementation, the scene recognizer is configured to perform the following steps: A1: Obtain scene classification data by using the learning strategy data and the learning environment data through a classification model algorithm; A2: Integrate the scenario classification data, the learning strategy data, the learning environment data, and the learning content data into a learning state vector sequence, so that each learning state vector in the learning state vector sequence consists of a learning scenario part, a learning strategy part, a learning environment part, and a learning content part. Then, use the learning state vector sequence as the learning state data.
[0011] The scene recognizer that implements the above method constructs a learning trajectory by building a learning state vector, further breaking through the limitation of a single scene. This helps to meet the requirements of context-related storage and fully restore the student's learning situation in scenarios such as classroom, homework, and exams. In one possible implementation, the feature extraction device includes: The memory, which communicates with the scene recognizer, is used to store the learning state data and update the storage state in real time; The feature extraction device communicates with both the memory and the service generation device to extract the learning state data of a specified time length from the memory and perform 3D target feature extraction based on a deep learning algorithm to obtain 3D target features.
[0012] By setting up a memory, learning trajectories can be stored for short-term or long-term storage for easy retrieval. By setting up a feature extraction device, three-dimensional target features can be extracted, extracting features from multiple dimensions such as scene association, time series, and ability evolution, going beyond the superficial analysis of traditional "knowledge point mastery" to achieve a deeper understanding of students' learning patterns.
[0013] In one possible implementation, the feature extraction device includes: The target detection network, which communicates with both the memory and the service generation device, is used to execute a target detection network model algorithm with at least two detection heads to extract knowledge and skill target features from the learning state data for a specified time period. A Bayesian network, communicating simultaneously with the memory and the service generation device, is used to model the student's knowledge state and learning process to extract process and method target features from the learning state data over a specified time period. The sentiment analyzer, which communicates with both the memory and the service generation device, is used to extract sentiment information from the learning state data over a specified time period to obtain target features of sentiment, attitude, and values.
[0014] The feature extraction device with the above structure and functions can extract target features by setting three different network names, thereby achieving simultaneous extraction of three-dimensional target features and realizing deep capture of student learning features.
[0015] In one possible implementation, the target detection network optimizes its parameters using a loss function formed by a weighted sum of the mean squared error loss function and the cross-entropy loss function, in order to improve the feature extraction accuracy and efficiency of the network model.
[0016] In one possible implementation, the service generation device includes: The comparator communicates with the target detection network, the Bayesian network, and the sentiment analyzer to extract the gap between the three-dimensional target features and the teacher's predetermined target features, and obtain the corresponding student's learning deficit features based on the obtained gap. The Big Prophecy module, communicating with the comparator, is used to generate personalized service suggestions tailored to the corresponding student based on the learning deficit characteristics using a large language model.
[0017] In one possible implementation, the comparator obtains the difference between the three-dimensional target features and the teacher's predetermined target features by calculating the Euclidean distance between them, thereby reducing computational complexity and avoiding resource waste while maintaining controllable accuracy.
[0018] In one possible implementation, the comparator obtains the learning deficit features of the corresponding student based on the obtained gap by: retrieving the deficit feature from the deficit feature pool that has the highest matching degree with the obtained distance, and using this deficit feature as the learning deficit feature.
[0019] In one possible implementation, the large oracle module is characterized by comprising at least three large oracle models to generate learning resource recommendations, capability diagnostic reports, and learning method suggestions, respectively.
[0020] Service generation devices with the above structure and functions can generate suggestions by setting different large language models. They can combine learning preferences, scene patterns and ability characteristics in contextual memory to generate service content that is more in line with students' actual needs and improve learning efficiency. Attached Figure Description
[0021] Figure 1 This is a schematic diagram of the structure of a student learning trajectory recording and personalized service system based on contextual memory, as disclosed in an embodiment of this application. Detailed Implementation
[0022] First, those skilled in the art should understand that these embodiments are merely used to explain the technical principles of the embodiments of this application and are not intended to limit the scope of protection of the embodiments of this application. Those skilled in the art can make adjustments as needed to adapt to specific application scenarios.
[0023] In the description of the embodiments of this application, it should be noted that, unless otherwise explicitly specified and limited, the term "forming a communication link structure" means that the multiple communication elements or modules involved form a network structure or network link structure through communication connection. Communication or communication connection means that there is information transmission between the first feature and the second feature. This information transmission can be unidirectional or bidirectional. The way to realize the communication connection can be electrical connection of wires, radio connection, electrical connection of electromagnetic media (such as optical fiber, semiconductor), communication realized by special channels, etc. It can be direct signal transmission or signal transmission through intermediate devices.
[0024] The present application will now be described in further detail with reference to the accompanying drawings and specific embodiments.
[0025] See Figure 1 This application discloses a student learning trajectory recording and personalized service system based on contextual memory. Figure 1 This is a schematic diagram of the system's structure. In this embodiment, the system is applied to an algebra course, and it runs once after each chapter. The algebra course is conducted on a learning platform, through which behavioral data, content data, and environmental data can be obtained. Behavioral data includes the speed of submitting answers in class, the number of times assignments are modified, the frequency of exam pauses, and page navigation records during self-study. Content data includes draft versions of essays, the process of inputting mathematical formulas, the content of incorrect question annotations, and the knowledge point documents consulted. Environmental data includes the class number of the class, the time of assignment submission (such as evening self-study time), and the device used for self-study (such as a home computer / school computer lab). The system includes a status acquisition device, a feature extraction device, and a service generation device. The feature extraction device communicates with the status acquisition device, and the service generation device communicates with the feature extraction device.
[0026] See Figure 1 In this system, the status acquisition device is used to receive student learning data output by the learning platform in real time and extract learning status data from the student learning data. For example... Figure 1As shown, in this embodiment, the state acquisition device includes a receiver, a policy recognizer, and a scene recognizer. The policy recognizer communicates with the receiver, while the scene recognizer communicates with both the policy recognizer and the feature extraction device.
[0027] In the status acquisition device, the receiver is used to receive in real time the learning process data, learning environment data, and learning content data paired with the learning process data and learning environment data output by the learning platform. The policy recognizer is used to retrieve learning policies from the policy pool that are compatible with the learning process in the learning process data to obtain learning policy data. The scene recognizer is used to perform scene recognition on the learning policy data, learning environment data, and learning content data to obtain learning status data.
[0028] In this embodiment, the scene recognizer is configured to perform the following steps: A1: obtain scene classification data using a classification model algorithm with learning strategy data and learning environment data; A2: integrate the scene classification data, learning strategy data, learning environment data, and learning content data into a learning state vector sequence, such that each learning state vector in the learning state vector sequence consists of a learning scene part, a learning strategy part, a learning environment part, and a learning content part, and then use the learning state vector sequence as the learning state data.
[0029] See Figure 1 In this system, a feature extraction device is used to extract three-dimensional target features from the learning state data based on a deep learning algorithm to obtain three-dimensional target features; the three-dimensional targets are knowledge and skills targets, process and method targets, and emotion, attitude, and value targets. Figure 1 As shown, in this embodiment, the feature extraction device includes a memory and a feature extraction apparatus, wherein the memory communicates with the scene recognizer, and the feature extraction apparatus communicates with both the memory and the service generation device.
[0030] In the feature extraction device, a memory is used to store the learning state data and update the storage status in real time. Specifically, in this embodiment, the memory includes a short-term context memory, a medium-term context memory, and a long-term context memory. The short-term context memory (Redis database) stores data for the past 7 days, supporting fast querying; the medium-term context memory (MySQL database) stores data sequences for the past 3 months for trend analysis; and the long-term context memory (HDFS distributed storage) stores feature patterns over 3 months (such as "self-study time decreases every weekend"). This setup facilitates querying learning trajectory data and also helps in the hierarchical operation of feature extraction to obtain three-dimensional target features for different time periods.
[0031] In a feature extraction device, the feature extraction unit is used to retrieve learning state data of a specified time length from memory and perform 3D target feature extraction based on a deep learning algorithm to obtain 3D target features. For example... Figure 1 As shown, in this embodiment, the feature extraction device includes an object detection network, a Bayesian network, and a sentiment analyzer. The object detection network communicates with the memory and service generation device simultaneously, the Bayesian network communicates with the memory and service generation device simultaneously, and the sentiment analyzer communicates with the memory and service generation device simultaneously.
[0032] In the feature extraction device, the object detection network executes an object detection network model algorithm with at least two detection heads to extract knowledge and skill target features from the learning state data over a specified time period. In this embodiment, the object detection network uses the eighth version of the object detection network, with three detection heads to extract knowledge features, skill features, and association features between knowledge and skills, respectively, over a specified time period of seven days. The object detection network optimizes its parameters using a loss function formed by a weighted sum of the mean squared error loss function and the cross-entropy loss function to improve detection accuracy and efficiency.
[0033] In the feature extraction device, a Bayesian network is used to model students' knowledge states and learning processes to extract process and method target features from learning state data over a specified time period of 3 months. A sentiment analyzer is used to extract sentiment information from the learning state data over a specified time period to obtain sentiment, attitude, and value target features; the specified time period is longer than 3 months to accurately capture sentiment information.
[0034] Please continue reading Figure 1 In this system, the service generation device extracts the gap between the three-dimensional target features and the teacher's predetermined target features, and uses a large language model to generate personalized service suggestions tailored to the corresponding students based on the obtained gaps. These personalized service suggestions include learning resource recommendations, ability diagnostic reports, and learning style suggestions. Figure 1 As shown, in this embodiment, the service generation device includes a comparator and a big oracle module. The comparator communicates with the object detection network, the Bayesian network, and the sentiment analyzer simultaneously, while the big oracle module communicates with the comparator.
[0035] In the service generation device, a comparator is used to extract the gap between the 3D target features and the teacher's predetermined target features, and obtain the corresponding student's learning deficit features based on the obtained gap. In this embodiment, the comparator obtains the gap by calculating the Euclidean distance between the 3D target features and the teacher's predetermined target features. The comparator obtains the corresponding student's learning deficit features based on the obtained gap by retrieving the defect feature from the defect feature pool that has the highest matching degree with the obtained distance, and using this defect feature as the learning deficit feature.
[0036] In the service generation device, the big oracle module is used to generate personalized service suggestions tailored to the corresponding student based on learning deficit characteristics using a large language model. In this embodiment, the big oracle module consists of at least three big oracle models, all of which communicate with a comparator to generate learning resource recommendations, ability diagnostic reports, and learning method suggestions, respectively.
[0037] Learning resource recommendations refer to the learning resources most suitable for the student's current state. Specifically, this can be achieved by combining time series characteristics and pushing extension questions that match the student's current ability during their historically high-performing periods (such as 7-8 pm every evening) (e.g., upgrading from "basic application problems" to "comprehensive application problems"). The ability diagnostic report, based on ability evolution characteristics, generates assessments such as "Logical reasoning ability: B+ (improved from B- in the last 3 months, recommended to strengthen geometry proof practice)," with scenario examples (e.g., "First time independently adding auxiliary lines in the homework on October 8th"). Learning method suggestions are a personalized learning path, specifically planned according to scenario-related characteristics for students who "understand in class but easily make mistakes in homework," progressing through "class notes review → similar example practice → revisiting incorrect homework questions."
[0038] The student learning trajectory recording and personalized service system based on contextual memory disclosed in this embodiment addresses the technical problems of this invention by setting up a state acquisition device, a feature extraction device, and a service generation device. The state acquisition device extracts learning state data, which includes learning scenarios, learning strategies, learning environments, and learning content, thus breaking through the limitations of single scenarios. This not only enables context-related storage, fully reconstructing the student's learning landscape in classrooms, assignments, and exams, but also provides a three-dimensional data foundation for personalized services. Simultaneously, the feature extraction device extracts three-dimensional target features from multiple dimensions: knowledge and skills, processes and methods, and emotions, attitudes, and values. This goes beyond the superficial analysis of traditional "knowledge point mastery" and achieves a deeper understanding of students' learning patterns. Finally, the service generation device combines learning preferences, scenario patterns, and ability characteristics from contextual memory to generate service content that better meets students' actual needs, improves learning efficiency, and provides teachers with ability diagnostic reports containing scenario details, assisting teachers in providing targeted guidance. This overcomes the technical shortcomings of existing technologies, which struggle to reflect the learning correlation patterns of students in different scenarios and whose service directions fail to meet real-world needs.
[0039] In the description of the embodiments of this application, it should be noted that the terms "inner" and "outer" and other terms indicating direction or positional relationship are based on the direction or positional relationship shown in the drawings. This is only for the convenience of description and does not indicate or imply that the device or component must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this application.
[0040] In the description of this application, the references to terms such as "an embodiment," "some embodiments," "in this embodiment," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in a suitable manner in any 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 different embodiments or examples.
[0041] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should 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 situational memory-based student learning trajectory recording and personalized service system, characterized in that, The method comprises the following steps: a state acquisition device is configured to receive student learning data output by a learning platform in real time, and extract learning state data from the student learning data; a feature extraction device in communication with the state acquisition device is configured to perform three-dimensional target feature extraction on the learning state data based on a deep learning algorithm, to obtain three-dimensional target features; the three-dimensional target is a knowledge and skill target, a process and method target, and an emotion, attitude and value target; a service generation device in communication with the feature extraction device is configured to extract a gap between the three-dimensional target features and teacher predetermined target features, and generate individual service suggestions adapted to corresponding students by using the obtained gap through a large language model; the individual service suggestions include learning resource recommendations, ability diagnosis reports, and learning method suggestions. 2.The situational memory-based student learning trajectory recording and individualized service system according to claim 1, characterized in that, The state acquisition device comprises: a receiver configured to receive learning process data, learning environment data, and learning content data paired with the learning process data and the learning environment data output by a learning platform in real time; a strategy identifier in communication with the receiver is configured to retrieve a learning strategy adapted to a learning process in the learning process data from a strategy pool to obtain learning strategy data; a scene identifier in communication with the strategy identifier and the feature extraction device is configured to perform scene recognition on the learning strategy data, the learning environment data, and the learning content data to obtain the learning state data. 3.The situational memory based student learning trajectory recording and individualized service system according to claim 2, characterized in that, The scene identifier is configured to perform the following steps: A1: obtain scene classification data by using the learning strategy data and the learning environment data through a classification model algorithm; A2: integrate the scene classification data, the learning strategy data, the learning environment data, and the learning content data into a learning state vector sequence, so that each learning state vector in the learning state vector sequence is composed of a learning scene part, a learning strategy part, a learning environment part, and a learning content part, and then use the learning state vector sequence as the learning state data. 4.The situational memory-based student learning trajectory recording and individualized service system according to claim 2 or 3, characterized in that, The feature extraction device comprises: a memory in communication with the scene identifier is configured to store the learning state data and update the storage state in real time; a feature extraction apparatus in communication with the memory and the service generation device is configured to extract the learning state data of a specified time length from the memory, and perform three-dimensional target feature extraction based on a deep learning algorithm to obtain three-dimensional target features. 5.The situational memory-based student learning trajectory recording and individualized service system according to claim 4, characterized in that, The feature extraction apparatus comprises: a target detection network in communication with the memory and the service generation device is configured to perform a target detection network model algorithm with at least two detection heads to extract knowledge and skill target features from the learning state data of a specified time length; a Bayesian network in communication with the memory and the service generation device is configured to model a student's knowledge state and learning process to extract process and method target features from the learning state data of a specified time length; The sentiment analyzer, in communication with the memory and the service generation device, is configured to extract sentiment information from the learning state data of a specified time length to obtain emotion, attitude and value target features. 6.The situational memory-based student learning trajectory recording and individualized service system according to claim 5, characterized in that, The target detection network is optimized by a loss function formed by a weighted sum of a mean square error loss function and a cross-entropy loss function. 7.The situational memory-based student learning trajectory recording and individualized service system according to claim 5 or 6, characterized in that, The service generation device comprises: The comparator, in communication with the target detection network, the Bayesian network and the sentiment analyzer, is configured to extract a distance between the three-dimensional target features and the teacher-determined target features, and obtain learning deficiency features of the corresponding student based on the obtained distance; The big prophecy module, in communication with the comparator, is configured to generate personalized service suggestions adapted to the corresponding student according to the learning deficiency features by a big language model. 8.The situational memory-based student learning trajectory recording and individualized service system according to claim 7, characterized in that, The comparator obtains the distance between the three-dimensional target features and the teacher-determined target features by calculating the Euclidean distance between them. 9.The situational memory based student learning trajectory recording and individualized service system according to claim 7, wherein, The comparator obtains the learning deficiency features of the corresponding student based on the obtained distance by retrieving the deficiency feature pool with the highest matching degree of the obtained distance, and taking the deficiency feature as the learning deficiency features.
10. The student learning trajectory recording and personalization service system based on episodic memory according to any one of claims 7-9, characterized in that, The big prophecy module is composed of at least three big prophecy models to generate learning resource recommendations, ability diagnosis reports and learning method suggestions, respectively.