Adaptive learning system based on artificial intelligence technology

Through the adaptive learning system of artificial intelligence technology, combined with personalized analysis and emotion recognition modules, the learning path is adjusted in real time, and the problem of failure to deeply understand user learning styles and emotional changes in the existing technology is solved, realizing personalized and dynamic Recommended learning resources improve learning effect and user satisfaction.

CN119067810APending Publication Date: 2024-12-03ANHUI BADU QUANTUM TECH CO LTD
View PDF 0 Cites 4 Cited by

Patent Information

Application Number
CN202411086165.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-08
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The existing technology fails to conduct in-depth research on individual learning styles, cognitive preferences and emotional state changes in educational resource recommendations, resulting in insufficient targetedness and effectiveness of the recommended content.

Method used

Adaptive learning system based on artificial intelligence, including personalized analysis module and emotion recognition module, is adopted, and the learning path and resource recommendation are adjusted in real time based on the user's learning portrait and emotional state.

Benefits of technology

It significantly improves learning efficiency and satisfaction, provides a personalized and dynamic learning experience, meets users' deep needs, and stimulates learning motivation and enthusiasm.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119067810A_ABST
    Figure CN119067810A_ABST
Patent Text Reader

Abstract

The invention discloses a self-adaptive learning system based on an artificial intelligence technology, relates to the technical field of self-adaptive learning of artificial intelligence, and solves the problem that in the prior art, in the aspect of educational resource recommendation, deep differences of individuals in learning styles, cognitive preferences and the like are not deeply researched. And the technical problem that individual demands cannot be accurately responded due to neglect of emotional state changes of individuals in different time and situations is solved. According to the method, user learning features are extracted through user learning data, and a user learning portrait is constructed; analyzing the emotional state of the user according to the state information of the user; based on the user learning portrait, the current learning demand and the current emotional state, recommending a personalized learning path; the system based on user learning data, sentiment analysis and personalized learning path recommendation has remarkable advantages in the aspects of improving learning efficiency, enhancing learning experience, realizing education personalization and the like.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of adaptive learning of artificial intelligence, and in particular to an adaptive learning system based on artificial intelligence technology. Background Art

[0002] In recent years, with the rapid development of information technology, especially the widespread application of cloud computing, big data analysis and artificial intelligence technology, the education industry has ushered in a wave of digital transformation; online learning platforms and applications have emerged in large numbers, and existing technologies provide educational data mining for improving teaching personalization and interactivity. It is used to analyze students' learning behavior patterns to identify learning difficulties and efficient learning strategies. At the same time, learning management systems widely integrate learning resource management, progress tracking, test evaluation and other functions.

[0003] Although existing technologies have achieved personalized recommendation of learning resources to a certain extent, their limitations are also very obvious. First, they mainly rely on users' past behavioral data, ignoring the changes in learners' emotional states at different times and situations, such as learning motivation, stress level and mood swings, which have an important impact on learning outcomes. Second, the recommendation system is relatively static, and does not fully consider the dynamic changes and growth of learners, and lacks the ability to adjust the learning path in real time according to learning outcomes. Finally, its degree of personalization is still limited, and it fails to delve into deep-level individual differences such as learning styles and cognitive preferences, resulting in the need to improve the pertinence and effectiveness of recommended content.

[0004] Therefore, the present invention provides an adaptive learning system based on artificial intelligence technology. Summary of the invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes an adaptive learning system based on artificial intelligence technology, which is used to solve the technical problems that the prior art does not conduct in-depth research on the deep differences in individuals' learning styles, cognitive preferences, etc. when recommending educational resources, and ignores the changes in the emotional state of individuals at different times and situations, and cannot accurately respond to individual needs.

[0006] To achieve the above-mentioned object, the first aspect of the present invention provides an adaptive learning system based on artificial intelligence technology, including an intelligent recommendation module, and a personalized analysis module and an emotion recognition module connected thereto;

[0007] Personalized analysis module: used to collect user learning data, extract user learning features, and build user learning portraits;

[0008] Emotion recognition module: used to collect user status information and analyze the user's emotional state; the status information includes voice audio and user image;

[0009] Intelligent recommendation module: used to recommend personalized learning paths based on the user's learning profile, current learning needs and current emotional state.

[0010] Preferably, the collecting of user learning data, extracting user learning features, and constructing user learning portraits include:

[0011] The user's basic information, learning style data and ability level data are extracted from the user's learning data to construct a user learning profile; among them, the user's basic information, learning style data and ability level data are the user's behavioral characteristics.

[0012] Preferably, extracting the user's basic information, learning style data and ability level data from the user's learning data includes:

[0013] Collect basic information of users; basic information of users includes user ID and grade in school;

[0014] Collect the user's historical browsing history, count the number of views and average browsing duration ratio of each video from the browsing history, and mark the videos of each type whose number of views exceeds a preset number and whose average browsing duration exceeds a preset duration ratio as frequently viewed videos; classify the frequently viewed videos according to the learning subject and video style to obtain the user's learning style data; wherein the video style includes explanation type and interactive type;

[0015] Collect the user's historical online test score records, extract the online test questions with failed scores from the historical online score records, classify the failed online test questions according to the subject type and the examination content of the test questions, and obtain the user's ability level data.

[0016] The present invention intelligently analyzes the browsing behavior of users, identifies frequently browsed videos, and further classifies them according to learning subjects and video styles, thereby gaining insight into the learning interests and style preferences of users. At the same time, by analyzing historical online test scores, it accurately identifies which subjects or knowledge points the users are deficient in, providing a basis for customized teaching plans. The personalized recommendation mechanism can stimulate students' learning interest, promote independent learning, and improve learning outcomes.

[0017] Preferably, the average browsing time ratio is the ratio of the average browsing time of the video to the total video time; wherein, the average browsing time of the video is the ratio of the sum of the browsing time of each video to the number of browsing times.

[0018] The present invention reflects the user's real participation and interest level in the video content by calculating the average browsing time ratio. A higher ratio means that the user tends to watch the video in its entirety, indicating that the content is highly relevant to their interests or very attractive. Conversely, a lower ratio may mean that the content does not match the user's needs or is not attractive enough. In addition, the user's interests and preferences may change over time. By continuously monitoring the average browsing time ratio, the system can capture these changes in a timely manner, adjust the recommendation strategy in a timely manner, and keep pace with user needs.

[0019] Preferably, collecting the user's status information and analyzing the user's emotional state includes:

[0020] During the user learning period, the user's voice audio and user image are collected from time to time; the user image includes the face image and the upper body image;

[0021] The voice audio and user image are input into the emotion state assessment model, and the user's current emotion state is output; wherein the user's emotion state includes positive state, fatigue state or lazy state; the emotion state assessment model is constructed by a recurrent neural network model.

[0022] The present invention utilizes a recurrent neural network model to process voice and image data, and can realize real-time monitoring of the user's emotional state during the learning process; the emotional state directly affects the learning efficiency and effect, and by timely identifying and processing negative emotions (such as fatigue or laziness), timely adjusting the learning difficulty or providing psychological support, preventing the decline of learning efficiency, and promoting students to maintain a positive learning attitude, thereby improving the effect of knowledge absorption and long-term memory.

[0023] Preferably, the emotional state assessment model is constructed by a recurrent neural network model, including:

[0024] Extract characteristic information of users in various states from historical data, recognize voice audio through the first level of the recurrent neural network model, and obtain the characteristics of voice audio; extract characteristic behaviors of user images through the second level of the recurrent neural network model; audio characteristics include pitch and speech speed;

[0025] The features of the speech audio obtained by the first layer of the recurrent neural network model and the characteristic behaviors of the user image obtained by the second layer of the recurrent neural network model are integrated to output the corresponding user emotional state;

[0026] By continuously training the recurrent neural network model, an emotional state assessment model is obtained.

[0027] The recurrent neural network model of the present invention can process sequential data, such as the time series of speech audio and continuous frames of user images, capture the temporal dependencies in the data, and realize the fusion of multimodal information through multi-level feature extraction, which helps to more comprehensively understand the emotional state of users in different situations. Because emotional expression is often multi-sensory, combining data from multiple modalities can significantly improve the accuracy of emotion recognition.

[0028] Preferably, the step of extracting characteristic behaviors of user images through the second level of the recurrent neural network model includes:

[0029] The second level of the RNN model extracts the user’s facial expression from the face image and the user’s sitting posture from the upper body image.

[0030] Preferably, the user learning portrait is updated according to the current user's needs, and a personalized learning path that suits the current situation is recommended; and during the learning process, the user's emotional state is paid attention to in real time to adjust the learning state.

[0031] By analyzing the user's learning profile, the system can accurately identify the user's interests, ability level and learning style, thereby tailoring a personalized learning path for the user, and paying attention to the user's emotional state in real time during the learning process, which helps to promptly discover and respond to the confusion, frustration or boredom that the user may encounter. By adjusting the learning state; by constantly updating the user's learning profile and recommending a personalized learning path that meets current needs, the user's continued interest and motivation in learning can be stimulated.

[0032] Preferably, each video in the learning platform is classified according to the teaching content, and the quality index of each video is evaluated. The video with the highest quality index in each teaching content is the high-quality video of the corresponding content; based on the content currently being learned by the user, the intelligent recommendation module will make recommendations based on the user's learning style.

[0033] The present invention screens out high-quality videos of different contents by evaluating the quality index of videos with different contents, thereby ensuring that users are exposed to high-quality teaching contents; through clear content organization, high-quality video screening, and accurate personalized recommendations, it provides users with a more efficient, convenient, and personalized learning experience, thereby improving learning effects.

[0034] Preferably, the evaluating the quality index of each video includes:

[0035] The number of likes, negative comments, views and reposts of the video is extracted, and the quality index of the video is calculated using the formula YSP=α×HPL+β×ZFL; among them, HPL is the positive rate, that is, the ratio of the number of likes to the number of negative comments; ZFL is the sharing rate, that is, the ratio of the repost rate to the number of views.

[0036] The present invention comprehensively considers various aspects of user feedback on videos, including likes, downvotes, views and reposts, so as to more comprehensively evaluate the quality of videos; and these data are easy to obtain and calculate, and can intuitively reflect the quality of videos; based on actual data, the number of likes, downvotes, views and reposts are calculated, avoiding the interference of subjective judgment, making the evaluation results more objective and fair.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] 1. The present invention constructs a user learning profile based on the user's basic information, learning style and ability level, monitors the emotional changes in learning in real time, dynamically generates or adjusts personalized learning plans, optimizes learning resource recommendations, and provides an instant feedback mechanism, which significantly improves learning efficiency and satisfaction, and ensures the continuity and depth of learning outcomes;

[0039] 2. The first level of the recurrent neural network model of the present invention is used to identify the characteristics of speech audio, and the second level of the recurrent neural network model is used to identify the user's expression and sitting posture characteristics; the user's current emotion is identified by fusing multiple information of audio, expression and sitting posture, and then the current learning state is reflected, and the user's learning state is adjusted in time, so that the user can obtain the maximum learning benefit within a limited time;

[0040] 3. The present invention extracts user learning features and constructs user learning portraits through user learning data; analyzes the user's emotional state based on the user's status information; recommends personalized learning paths based on the user's learning portraits, current learning needs and current emotional states; by constructing user learning portraits, analyzing emotional states and recommending personalized learning paths, a tailored learning path is provided for users, which meets the user's personalized needs and makes it easier for users to generate learning motivation and enthusiasm. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0042] Figure 1 It is a schematic diagram of the system structure of the present invention;

[0043] Figure 2 It is a schematic diagram of the process of the present invention;

[0044] Figure 3 A schematic diagram of the user learning portrait construction process of the present invention. DETAILED DESCRIPTION

[0045] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] See also Figure 1-Figure 2 , the first aspect of the present invention provides an adaptive learning system based on artificial intelligence technology, including an intelligent recommendation module, and a personalized analysis module and an emotion recognition module connected thereto;

[0047] The personalized analysis module collects user learning data, extracts user learning features, and constructs user learning portraits; user learning features include user basic information, learning style data, and ability level data;

[0048] See also Figure 3 Specifically, collect basic information of users; basic information of users refers to basic information registered by users on the learning platform, including user ID, name, age and grade in school, etc.;

[0049] Collect the user's historical browsing history; count the number of views and average browsing time ratio of each video from the browsing history, and mark the videos of each type whose number of views exceeds a preset number and whose average browsing time exceeds a preset time ratio as frequently viewed videos; classify the frequently viewed videos according to the learning subject and video style to obtain the user's learning style data; wherein the video style includes explanation type and interactive type;

[0050] It should be noted that the average viewing time ratio is the ratio of the average viewing time of each video to the total video time; among which, the average viewing time of each video is the ratio of the sum of the viewing time of each video to the number of views.

[0051] For example: collect browsing records of all users, including video ID, number of views and browsing time; calculate the average browsing time of each video; mark the video type (learning subject) and style (explanatory or interactive) according to the video content; and extract frequently viewed videos, where the video type and style corresponding to the frequently viewed videos are the learning styles preferred by users when studying various subjects.

[0052] Collect the user's historical online test score records; extract the online test questions with failed scores from the historical online score records, classify the failed online test questions according to the subject type and the examination content of the test questions, and obtain the user's ability level data.

[0053] It should be noted that the subjects tested by the online test questions with unsatisfactory scores and the corresponding test contents are knowledge points that the users have not mastered proficiently. These knowledge points need to be marked as important points and more auxiliary materials about these knowledge points need to be provided to the users.

[0054] Integrate the user's basic information, learning style data and ability level data to build a user learning profile.

[0055] The emotion recognition module collects the user's state information and analyzes the user's emotional state; wherein the state information includes the user's image and voice audio;

[0056] During the user learning period, the user's voice audio and user image are collected from time to time; the user image includes the face image and the upper body image;

[0057] For example: during the user's learning period, in order to monitor the user's learning status, the learning platform will periodically turn on the voice acquisition mode and image acquisition mode; the voice acquisition mode is to regularly trigger the recording function on the learning software or platform, requiring the user to read aloud the specified text or answer questions; the image acquisition mode is to capture the user's learning expression and posture through the camera.

[0058] The voice audio and user image are input into the emotion state assessment model, and the user's current emotion state is output; wherein the user's emotion state includes positive state, fatigue state or sad state; the emotion state assessment model is constructed by a recurrent neural network model.

[0059] The emotional state assessment model training process is as follows:

[0060] Extract characteristic information of users in different emotional states from historical data, recognize voice audio through the first level of the recurrent neural network model, and obtain the characteristics of voice audio; extract characteristic behaviors of user images through the second level of the recurrent neural network model; audio characteristics include pitch and speech speed;

[0061] The features of the speech audio obtained by the first layer of the recurrent neural network model and the characteristic behaviors of the user image obtained by the second layer of the recurrent neural network model are integrated to output the corresponding user emotional state;

[0062] By continuously training the recurrent neural network model, the output layer outputs the emotional state evaluation model.

[0063] It should be noted that the second level of the recurrent neural network model extracts the user's facial expression from the face image and the user's sitting posture from the upper body image.

[0064] For example: collect voice audio and image data of users in different emotional states; voice audio includes recordings of conversations between users in different emotional states; image data includes video clips or continuous images of users' facial expressions and body postures;

[0065] Use the first layer of the recurrent neural network model (such as LSTM or GRU units) to process the audio sequence and extract the time series features of the audio, such as pitch, speaking speed, etc.

[0066] The second layer of the recurrent neural network model is used to process the image sequence and extract the time series features of the image, such as changes in facial expressions, head movements, etc.

[0067] The features extracted from audio and images are fused in the time dimension to form a multimodal feature vector. The fused multimodal features are used to train a recurrent neural network model for emotional state prediction. The output layer of the model outputs the user's emotional category (such as happiness, sadness, anger, etc.).

[0068] The intelligent recommendation module recommends personalized learning paths based on the user's learning profile, current learning needs and current emotional state.

[0069] Specifically, the user learning portrait is updated according to the current user's needs, and a personalized learning path that suits the current situation is recommended; and during the learning process, the user's emotional state is paid attention to in real time, thereby adjusting the learning state.

[0070] For example: the current user inputs his or her own learning needs in the learning platform or the current user has been promoted to a higher level, and the user's learning profile adjusts the learning content based on this, and recommends corresponding learning videos for the user's preferred learning style; in addition, the user's learning status is adjusted in real time according to the user's emotional state. For example, when the user's emotional state is in a positive state, the user's current learning difficulty is increased; when the user is in a state of fatigue, the study is temporarily suspended, and the user is advised to take a break for a while or to recommend the user to engage in relaxing activities, such as eye exercises and stretching exercises; when the user is in a sad state, the learning difficulty is reduced, and the user is encouraged to participate in discussions in the learning community and exchange experiences with others.

[0071] Furthermore, the intelligent recommendation module also classifies each video in the learning platform according to the teaching content, extracts the number of likes, negative comments, views and reposts of the video, and calculates the quality index of the video through the formula YSP=α×HPL+β×ZFL; where HPL is the positive rate, that is, the ratio of the number of likes to the number of negative comments; ZFL is the sharing rate, that is, the ratio of the number of reposts to the number of views;

[0072] The quality index in the present invention is used to evaluate the teaching quality of the video, which is closely related to the number of likes, negative comments, views and reposts. If the number of positive comments and reposts is relatively high, it means that the teaching video is of good quality, and many users have learned from the teaching video and given it a high evaluation, so the quality index of the teaching video is high.

[0073] Based on the quality index, the video with the highest quality index in each teaching content is the high-quality video of the corresponding content; according to the content currently being learned by the user, the intelligent recommendation module will make recommendations based on the user's learning style.

[0074] For example: A user is studying the graphics and geometry unit of third-grade mathematics, and his preferred video learning mode is interactive; if the high-quality video corresponding to the content is of the explanatory type, the intelligent recommendation module will not recommend this high-quality video to the user; if the high-quality video corresponding to the content is of the interactive type, the intelligent recommendation module will recommend this high-quality video to the user.

[0075] Part of the data in the above formula is calculated by removing the dimension and taking its numerical value. The formula is a formula closest to the actual situation obtained by software simulation of a large amount of collected data; the preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained through simulation of a large amount of data.

[0076] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. An adaptive learning system based on artificial intelligence technology, characterized in that: It includes an intelligent recommendation module, as well as a personalized analysis module and an emotion recognition module connected thereto; Personalized analysis module: used to collect user learning data, extract user learning features, and build user learning portraits; Emotion recognition module: used to collect user status information and analyze the user's emotional state; the status information includes voice audio and user image; Intelligent recommendation module: used to recommend personalized learning paths based on the user's learning profile, current learning needs and current emotional state.

2. The adaptive learning system based on artificial intelligence technology according to claim 1 is characterized in that: The collecting of user learning data, extracting user learning features, and constructing user learning portraits include: The user's basic information, learning style data and ability level data are extracted from the user's learning data to construct a user learning profile; among them, the user's basic information, learning style data and ability level data are the user's behavioral characteristics.

3. The adaptive learning system based on artificial intelligence technology according to claim 2 is characterized in that: The extracting of the user's basic information, learning style data and ability level data from the user's learning data includes: Collect basic information of users; basic information of users includes user ID and grade in school; Collect the user's historical browsing history, count the number of views and average browsing duration ratio of each video from the browsing history, and mark the videos of each type whose number of views exceeds a preset number and whose average browsing duration exceeds a preset duration ratio as frequently viewed videos; classify the frequently viewed videos according to the learning subject and video style to obtain the user's learning style data; wherein the video style includes explanation type and interactive type; Collect the user's historical online test score records, extract the online test questions with failed scores from the historical online score records, classify the failed online test questions according to the subject type and the examination content of the test questions, and obtain the user's ability level data.

4. The adaptive learning system based on artificial intelligence technology according to claim 3 is characterized in that: The average viewing duration ratio is the ratio of the average viewing duration of the video per time to the total video duration; wherein, the average viewing duration of the video per time is the ratio of the sum of the viewing durations of the video per time to the number of views.

5. The adaptive learning system based on artificial intelligence technology according to claim 1 is characterized in that: The collecting of user status information and analyzing the user's emotional status includes: During the user learning period, the user's voice audio and user image are collected from time to time; the user image includes the face image and the upper body image; The voice audio and user image are input into the emotion state assessment model, and the user's current emotion state is output; wherein the user's emotion state includes positive state, fatigue state or lazy state; the emotion state assessment model is constructed by a recurrent neural network model.

6. The adaptive learning system based on artificial intelligence technology according to claim 5, characterized in that: The emotional state assessment model is constructed by a recurrent neural network model, including: Extract characteristic information of users in various states from historical data, recognize voice audio through the first level of the recurrent neural network model, and obtain the characteristics of voice audio; extract characteristic behaviors of user images through the second level of the recurrent neural network model; audio characteristics include pitch and speech speed; The features of the speech audio obtained by the first layer of the recurrent neural network model and the characteristic behaviors of the user image obtained by the second layer of the recurrent neural network model are integrated to output the corresponding user emotional state; By continuously training the recurrent neural network model, an emotional state assessment model is obtained.

7. The adaptive learning system based on artificial intelligence technology according to claim 6, characterized in that: The extracting characteristic behaviors of the user image through the second level of the recurrent neural network model includes: The second level of the RNN model extracts the user’s facial expression from the face image and the user’s sitting posture from the upper body image.

8. The adaptive learning system based on artificial intelligence technology according to claim 1, characterized in that: User learning profile: Update the user learning profile based on the current user's needs and recommend personalized learning paths that meet the current needs. Also, pay attention to the user's emotional state in real time during the learning process, and adjust the learning state accordingly.

9. The adaptive learning system based on artificial intelligence technology according to claim 1, characterized in that: Each video in the learning platform is classified according to the teaching content, and the quality index of each video is evaluated. The video with the highest quality index in each teaching content is the high-quality video of the corresponding content; based on the content currently being learned by the user, the intelligent recommendation module will make recommendations based on the user's learning style.

10. The adaptive learning system based on artificial intelligence technology according to claim 9, characterized in that: The quality index of each video is evaluated, including: The number of likes, negative comments, views and reposts of the video is extracted, and the quality index of the video is calculated using the formula YSP=α×HPL+β×ZFL; among them, HPL is the positive rate, that is, the ratio of the number of likes to the number of negative comments; ZFL is the sharing rate, that is, the ratio of the repost rate to the number of views.

Citation Information

Cited By

  • Education intelligent learning path recommendation system

    CN119357481A

  • An educational intelligent learning path recommendation system

    CN119357481B

  • Intelligent accompanying learning method and device, storage medium and equipment

    CN120011612A

  • An intelligent learning companion method, device, storage medium and equipment

    CN120011612B