An adaptive learning evaluation method and system based on large model driving

Abstract

The present invention provides an adaptive learning evaluation method and system driven by a large model. The method includes authenticating the user. If the authentication is successful, the user's behavioral data and emotional data are obtained; behavioral analysis is performed on the behavioral data to output behavioral analysis results; emotional state analysis is performed on the emotional data to obtain emotional analysis results; training data is obtained, the training data is input into a preset model for training and the model performance is evaluated to obtain a target model, and the behavioral analysis results, emotional analysis results, and target learning plan are input into the target model to output evaluation results. The present invention can provide more refined learning experience analysis and evaluate students' learning effects. It also provides personalized learning suggestions based on the evaluation results, and adjusts teaching content and learning paths to achieve true adaptive learning.

Description

Technical Field

[0001] The present invention belongs to the technical field of learning assessment, and in particular relates to an adaptive learning assessment method and system driven by a large model. Background Art

[0002] Currently, in the field of education, especially adult higher education, the more traditional approach is to organize learning offline, with teachers conducting adaptive learning assessments through questionnaires, interviews, interactions, homework, etc. This traditional approach has inconsistent standards, relies on manpower, lacks personalization, is difficult to involve the entire learning process, and is affected by teachers' subjective factors.

[0003] Some have also attempted to use traditional IT systems on online learning platforms, which record user usage records, learning time, test assessment scores, etc. However, they lack multimodal comprehensive assessment methods. On the one hand, there is a lack of identity authenticity verification, and on the other hand, there is a lack of other important analysis bases such as vision and voice, which in turn affects the accuracy of learning effect evaluation. Summary of the Invention

[0004] In order to solve the above technical problems, the present invention provides an adaptive learning evaluation method and system based on large model driving, which are used to solve the technical problems in the prior art.

[0005] In one aspect, the present invention provides the following technical solution: a large model-driven adaptive learning evaluation method, comprising:

[0006] Authenticate the user, and if the authentication is successful, obtain the user's behavioral data and emotional data;

[0007] Performing behavior analysis on the behavior data to output behavior analysis results;

[0008] Performing emotional state analysis on the emotional data to obtain an emotional analysis result;

[0009] Acquire training data, input the training data into a preset model for training and evaluate model performance to obtain a target model, input the behavior analysis results, the emotion analysis results, and the target learning plan into the target model to output an evaluation result.

[0010] Compared with the existing technology, the beneficial effects of the present invention are: the scheme of the present invention is not limited to a single data collection method, but integrates multiple biometric technologies such as facial recognition, eye tracking, and voice recognition, so that it can more comprehensively capture students' learning behaviors and emotional states. In addition to basic learning time and rhythm analysis, the present invention also includes analysis of the speed and order of answering questions. These are all in-depth understandings of students' cognitive processes. The present invention uses facial expression recognition and voice emotion analysis to evaluate students' emotional states, which is rare in traditional online learning systems and can provide more refined learning experience analysis. The present invention not only evaluates students' learning effects, but also provides personalized learning suggestions based on the evaluation results, and adjusts teaching content and learning paths to achieve true adaptive learning.

[0011] Preferably, the step of performing identity authentication on the user and obtaining the user's behavior data and emotion data if the identity authentication is passed includes:

[0012] Obtaining identity information of the user, and performing biometric identification on the user based on the identity information of the user to obtain a first verification result;

[0013] Obtaining the user's login information, and performing identity authentication on the user based on the login information to obtain a second verification result;

[0014] Obtaining the user's historical learning records, extracting the user's learned knowledge information from the historical learning records, generating a verification test question based on the learned knowledge information, and performing learning content identification on the user based on the verification test question to obtain a third verification result;

[0015] If the first verification result, the second verification result, and the third verification result are all verified to be passed, the behavior data and emotion data of the user are obtained.

[0016] Preferably, the step of performing behavior analysis on the behavior data to output the behavior analysis result includes:

[0017] Determine learning engagement based on the behavioral data :

[0018] ;

[0019] Where, For standard study time, For the The actual study time of each student;

[0020] Determine a learning rhythm ratio based on the behavioral data :

[0021] ; ;

[0022] Where, For the The amount of learning content for each student, Indicates the The time required for a student to complete the learning content For the The learning completion speed of each student, The speed of completion of standard learning;

[0023] Determine the answering speed ratio based on the behavioral data :

[0024] ;

[0025] Where, For the Students complete the target questions The actual response time, Target topic Standard response time;

[0026] Determine an answer sequence indicator based on the behavioral data :

[0027] ;

[0028] Where, For the The difficulty of the question, is the total number of questions, Indicates the The difficulty of the first question a student chooses to answer;

[0029] The learning engagement, the learning rhythm ratio, the answering speed ratio, and the answering sequence index are comprehensively considered to obtain a behavior analysis result.

[0030] Preferably, the step of performing emotional state analysis on the emotional data to obtain an emotional analysis result includes:

[0031] Extracting a facial image from the emotion data, and extracting key facial feature points using a feature point detection algorithm to obtain facial feature data;

[0032] Inputting the facial feature data into a trained first emotion classifier to obtain a first emotion recognition result;

[0033] Extracting voice data from the emotion data, and preprocessing the voice data to obtain processed voice data;

[0034] Extracting acoustic features from the processed speech data, and inputting the acoustic features into a trained second emotion classifier to obtain a second emotion recognition result;

[0035] The first emotion recognition result and the second emotion recognition result are combined to obtain an emotion analysis result.

[0036] Preferably, the step of inputting the training data into a preset model for training and evaluating model performance to obtain a target model includes:

[0037] The training data is input into the preset model for training, and the predicted score is calculated based on the training output results. :

[0038] ;

[0039] Where, is the intercept parameter, is the error term, To expect test scores, are the first to seventh variable coefficients, Respectively Each student's actual learning time, learning pace ratio, answer speed ratio, answer order index, concentration, confusion, and learning engagement;

[0040] Calculate the mean square error based on the predicted performance :

[0041] ;

[0042] Where, is the number of students in the training data, For the The actual test scores of each student;

[0043] Calculated based on the predicted grades Fraction :

[0044] ;

[0045] Where, is the average of the test scores;

[0046] Based on the mean square error 、 The preset model is optimized and trained to obtain the target model.

[0047] Preferably, the step of inputting the behavior analysis result, the emotion analysis result, and the target learning plan into the target model to output the learning evaluation result is specifically as follows:

[0048] The behavior analysis results, the emotion analysis results, and the target learning plan are spliced ​​and fused to obtain a comprehensive feature vector, and the comprehensive feature vector is input into the target model to obtain a learning evaluation result.

[0049] In a second aspect, the present invention provides the following technical solution: an adaptive learning evaluation system driven by a large model, the system comprising:

[0050] A verification module is used to authenticate the user and, if the authentication is successful, obtain the user's behavior data and emotion data;

[0051] A behavior module, configured to perform behavior analysis on the behavior data to output a behavior analysis result;

[0052] An emotion module, configured to perform emotion state analysis on the emotion data to obtain emotion analysis results;

[0053] The evaluation module is used to obtain training data, input the training data into a preset model for training and evaluating the model performance to obtain a target model, input the behavior analysis results, the emotion analysis results, and the target learning plan into the target model to output an evaluation result.

[0054] In a third aspect, the present invention provides the following technical solution: a computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor; when the processor executes the computer program, the adaptive learning evaluation method driven by a large model as described above is implemented.

[0055] In a fourth aspect, the present invention provides the following technical solution: a storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the above-mentioned large model-driven adaptive learning evaluation method. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. 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 any creative work.

[0057] Figure 1 Flowchart of the large model-driven adaptive learning evaluation method provided in Example 1 of the present invention;

[0058] Figure 2 This is a structural diagram of the large model-driven adaptive learning evaluation system provided in Example 2 of the present invention;

[0059] Figure 3 A schematic diagram of the hardware structure of a computer provided in another embodiment of the present invention.

[0060] The embodiments of the present invention will be further described below with reference to the accompanying drawings. DETAILED DESCRIPTION

[0061] The following describes embodiments of the present invention in detail, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the embodiments of the present invention, and should not be construed as limiting the present invention.

[0062] Example 1

[0063] In the first embodiment of the present invention, Figure 1 As shown, a large model-driven adaptive learning evaluation method includes:

[0064] S1. Authenticate the user. If the authentication is successful, obtain the user's behavioral data and emotional data.

[0065] Wherein, the step S1 includes:

[0066] S11. Obtaining identity information of the user, and performing biometric identification on the user based on the identity information of the user to obtain a first verification result;

[0067] Specifically, the identity information here includes facial information and fingerprint information, and the biometric recognition process is facial recognition and fingerprint recognition, which verifies the student's identity through biometric information.

[0068] S12. Obtaining the user's login information, and performing identity authentication on the user based on the login information to obtain a second verification result;

[0069] Specifically, for the learning assessment method provided in this application, it is specifically installed in the learning assessment system. Therefore, every time a student logs in to the system, the student's login information, such as login account, login password, etc., will be verified. Therefore, identity login recognition can verify the student's identity to ensure the accuracy of the assessment results.

[0070] S13, obtaining the user's historical learning records, extracting the user's learned knowledge information from the historical learning records, generating a verification test question based on the learned knowledge information, and performing learning content identification on the user based on the verification test question to obtain a third verification result;

[0071] Specifically, based on historical learning records, the historical learning knowledge points that students have learned, that is, the learned knowledge information, can be determined, and the big model will set corresponding questions based on the historical learning knowledge points mastered by the students, that is, generate verification questions to test whether they are the real person, and then establish the student's identity verification process at the knowledge graph level.

[0072] S14. If the first verification result, the second verification result, and the third verification result are all verified to be passed, the behavior data and emotion data of the user are obtained.

[0073] S2. performing behavior analysis on the behavior data to output a behavior analysis result;

[0074] Wherein, the step S2 includes:

[0075] S21. Determine learning engagement based on the behavioral data :

[0076] ;

[0077] Where, For standard study time, For the The actual study time of each student;

[0078] Specifically, learning engagement is used to evaluate a student's learning engagement. When the learning engagement is not less than 1, the student's learning engagement is considered to be high; when the learning engagement is less than 1, the student's learning engagement is considered to be low.

[0079] S22, determining the learning rhythm ratio based on the behavior data :

[0080] ; ;

[0081] Where, For the The amount of learning content for each student, Indicates the The time required for a student to complete the learning content For the The learning completion speed of each student, The speed of completion of standard learning;

[0082] Specifically, the learning rhythm ratio is used to reflect the student's rhythm, which includes learning speed, learning progress, etc., to judge the student's learning efficiency and learning attitude. When the learning rhythm ratio is not less than 1, it is considered that the student's learning rhythm is fast. When the learning rhythm ratio is less than 1, it is considered that the student's learning rhythm is slow.

[0083] S23. Determine the answering speed ratio based on the behavior data :

[0084] ;

[0085] Where, For the Students complete the target questions The actual response time, Target topic Standard response time;

[0086] Specifically, the answering speed ratio is used to judge the student's answering speed in the process of answering questions. It is used to evaluate the student's thinking reaction speed and problem-solving ability. When the answering speed ratio is not greater than 1, it is considered that the student's thinking reaction speed is fast. When the answering speed ratio is greater than 1, it is considered that the student's thinking reaction speed is slow.

[0087] S24. Determine the answer sequence index based on the behavior data :

[0088] ;

[0089] Where, For the The difficulty of the question, is the total number of questions, Indicates the The difficulty of the first question a student chooses to answer;

[0090] Specifically, the answer order indicator is used to assess students' learning strategies and thinking styles. This can be assessed by analyzing the order in which students choose to answer questions. This analysis is relatively complex and typically requires qualitative analysis based on the specific learning content and the specific circumstances of the student's responses. A quantitative approach involves comparing the difficulty of the question a student chooses to answer first with the overall difficulty level. If the answer order indicator is greater than 1, it indicates that the student tends to tackle difficult questions first. If it is less than 1, it indicates that the student tends to proceed from easy to difficult questions.

[0091] S25. Integrate the learning engagement, the learning rhythm ratio, the answering speed ratio, and the answering sequence index to obtain a behavior analysis result.

[0092] At the same time, in the actual evaluation process, you can first build a Transformer-based knowledge tracking model (KT-Model), fuse text data, voice data, and image data into a multimodal feature vector, train the model, and then input the vector into the trained Transformer-based knowledge tracking model to output the knowledge point mastery prediction results. At the same time, introduce the course knowledge graph as the model prior, and enhance interpretability through the graph neural network (GNN).

[0093] S3. performing an emotional state analysis on the emotional data to obtain an emotional analysis result;

[0094] Wherein, the step S3 includes:

[0095] S31, extracting a facial image from the emotion data, and extracting key facial feature points using a feature point detection algorithm to obtain facial feature data;

[0096] Specifically, first, the system locates the position of the face in the image or video frame by using a Haar cascade classifier, HOG (histogram of oriented gradients) plus SVM (support vector machine) and a CNN (convolutional neural network) model based on deep learning. After detecting the face, the system extracts the key feature points of the face, such as the position and shape of the eyes, eyebrows, nose, and mouth. This step is completed by using a feature point detection algorithm, Dlib's 68-point model or a deep learning-based method.

[0097] S32, inputting the facial feature data into the trained first emotion classifier to obtain a first emotion recognition result;

[0098] Specifically, the feature data is then input into a classifier that has been trained to recognize different emotional states, such as happiness, sadness, surprise, fear, disgust, anger, and neutrality. The classifier is based on traditional machine learning methods and is trained to match appropriate results through a CNN deep learning model.

[0099] S33, extracting voice data from the emotion data, and preprocessing the voice data to obtain processed voice data;

[0100] Specifically, the preprocessing process is as follows: the voice signal collected by JS front-end technology is converted into a digital form of WAV sampling, and preprocessing steps such as noise elimination and silence detection are performed to improve data quality.

[0101] S34, extracting acoustic features from the processed speech data, and inputting the acoustic features into a trained second emotion classifier to obtain a second emotion recognition result;

[0102] Specifically, acoustic features are extracted from the preprocessed speech data, including fundamental frequency (F0), energy, time domain features, frequency domain features, Mel-frequency cepstral coefficients (MFCCs), etc. These features can capture the emotional information of the speech. The extracted acoustic features are fed into an emotion classifier, which can identify different emotional states such as happiness, sadness, anger, etc. These classifiers are usually implemented based on deep learning technology CNN.

[0103] S35, combining the first emotion recognition result and the second emotion recognition result to obtain an emotion analysis result;

[0104] Specifically, in the actual evaluation process, the ability of semantic understanding of speech transcription can be combined with the big model to collect and archive problems encountered by students in the learning process, or detect abnormal semantics such as chat topics that are not focused, so as to improve the comprehensiveness of the evaluation.

[0105] S4. Obtain training data, input the training data into a preset model for training and evaluate model performance to obtain a target model, input the behavior analysis results, the emotion analysis results, and the target learning plan into the target model to output an evaluation result.

[0106] The step of inputting the training data into a preset model for training and evaluating the model performance to obtain a target model includes:

[0107] S41: Input the training data into a preset model for training, and calculate the predicted score based on the training output results. :

[0108] ;

[0109] Where, is the intercept parameter, is the error term, To expect test scores, are the first to seventh variable coefficients, Respectively Each student's actual learning time, learning pace ratio, answer speed ratio, answer order index, concentration, confusion, and learning engagement;

[0110] Specifically, is the intercept term, which is used to represent the expected score when all independent variables are 0, and the error term is used for random variation that the model cannot explain.

[0111] S42, calculating the mean square error based on the predicted results :

[0112] ;

[0113] Where, is the number of students in the training data, For the The actual test scores of each student.

[0114] S43, calculation based on the predicted score Fraction :

[0115] ;

[0116] Where, is the average test score.

[0117] S44, based on the mean square error 、 The preset model is optimized and trained to obtain the target model.

[0118] The step of inputting the behavior analysis result, the emotion analysis result, and the target learning plan into the target model to output the learning evaluation result is specifically as follows:

[0119] The behavior analysis results, the emotion analysis results, and the target learning plan are spliced ​​and fused to obtain a comprehensive feature vector, and the comprehensive feature vector is input into the target model to obtain a learning evaluation result.

[0120] Specifically, while outputting the learning evaluation results, the system will integrate the above data and extract features related to learning effects, such as learning time, learning rhythm, emotional state, etc. Multimodal data and analysis result reports are mainly reflected in three aspects:

[0121] 1. Visual Reports: Use charts, progress bars, and color coding to display metrics such as study time, learning pace, problem-solving speed, answering order, and emotional state. For example, use a line chart to display learning progress, with color changes indicating changes in emotional state. 2. Numerical Scores: Assign numerical scores to various aspects of learning behavior and emotional state for quantitative comparison and progress tracking. 3. Written Summary: Provide a brief written summary outlining the student's learning performance and emotional trends, highlighting strengths and areas for improvement.

[0122] At the same time, this application can construct a teaching decision-making agent based on the above model, generating natural language feedback and learning plans based on the evaluation results. Based on the learning effect evaluation results (when there is a difference between the actual and predicted scores, the specific scores of each factor affecting the score and the results of each knowledge module), the agent provides real-time feedback to students and teachers, and adjusts the teaching content and learning path accordingly to achieve adaptive learning;

[0123] Student feedback: The intelligent body will provide students with personalized learning reports, including learning progress, learning effect evaluation, learning suggestions, etc., to help students understand their learning situation and improve their learning methods.

[0124] Teacher feedback: The AI ​​will provide teachers with student learning reports and evaluation results to help them better understand students' learning situation and adjust teaching content and teaching strategies based on the evaluation results.

[0125] Adjustment of teaching content and learning paths: Based on learning effect evaluation and feedback results, the system will automatically adjust the learning content and learning paths to provide students with teaching resources and learning recommendations that are more in line with their learning needs and levels.

[0126] At the same time, the present invention can achieve:

[0127] Calibration of independent variable coefficients. The relationship between learning methods and influencing factors in various courses is not fixed. Therefore, we first need to adjust the reliability of the independent variable coefficients of each course using training data to calculate the coefficients that are relatively suitable for the predicted and actual grades of the current course.

[0128] Analyze differences between individual behavior and the standard sample. Based on individual students' specific learning behaviors and performance during the test-taking process, analyze the specific reasons for the difference between their performance and the standard sample. Examples include insufficient time invested in learning, lack of focus, and confusion about the content; during the test-taking process, factors such as the order of answering questions, speed of answering questions, and time spent thinking; and personal interest in the current knowledge.

[0129] Adaptive learning content and learning content recommendations. Students who, due to personal subjective factors, have not invested time or are uninterested need teachers to communicate with them and encourage them to learn. For students who have questions or do not understand something during the learning process, teachers recommend Q&A content and teaching cases for corresponding chapters and related difficult knowledge points to help them further their learning. Students who have learned but were careless in their answers will be given guidance and offered the opportunity to retake or make up the exam.

[0130] Furthermore, in practical applications of the present invention, students must undergo multimodal authentication before logging into the learning platform. This authentication uses biometric technologies such as facial recognition and fingerprint recognition to verify their real-person identity, ensuring that the learning system is used solely by the student. Furthermore, the system uses the student's knowledge graph mastery to verify the large-scale model's question generation, allowing for the determination of knowledge substitution. During the learning process, the system collects multimodal student data, such as keyboard operation data, mouse click data, eye tracking data, facial expression data, and voice input data. By analyzing this data, the system can gain a preliminary understanding of the student's learning behavior and emotional state. Combined with the large-scale model's enhanced semantic understanding capabilities, it can uncover detailed factors in the learning process. The system conducts in-depth analysis of student learning behavior, including analysis of learning time, learning rhythm, question answering speed, and answer sequence. Through these analyses, the system can assess students' learning efficiency and mastery, helping them understand their own learning progress. Furthermore, the results can be refined through the large-scale model to generate natural language and conversational feedback. The system analyzes students' facial expression data and voice data to assess their emotional state, such as their focus, confusion, and interest. This can better understand students' learning experience and emotional state, provide students with personalized learning support, and at the same time, combine students' interests, allowing the big model to give some interesting examples based on the theoretical knowledge of the course to attract students and increase their enthusiasm; based on the above multimodal data and analysis results, after students complete the math test, the system calls the big model for multi-dimensional analysis. Specifically, the CLIP model is used to align the screenshots of wrong questions with the knowledge graph to locate weak knowledge points (such as "trigonometric function image transformation"); GPT-4 is used to generate wrong question analysis video scripts and automatically synthesize explanation videos with virtual teachers; the intelligent agent recommends the golden learning time of 18:00-20:00 every day based on students' historical data (such as the fluctuation pattern of concentration); at the same time, the intelligent agent provides real-time feedback to students and teachers based on the learning effect evaluation results, and adjusts the teaching content and learning path accordingly to achieve adaptive learning. Students can understand their own learning situation and improve their learning methods based on personalized learning reports, while teachers can understand students' learning situation based on the evaluation results and adjust teaching content and teaching strategies accordingly.

[0131] The first embodiment of the present invention provides an adaptive learning assessment method based on large model driving. The solution of the present invention is not limited to a single data collection method, but integrates multiple biometric technologies such as facial recognition, eye tracking, and voice recognition, so as to more comprehensively capture students' learning behaviors and emotional states. In addition to basic learning time and rhythm analysis, the present invention also includes analysis of the speed and order of answering questions, which are all in-depth understandings of students' cognitive processes. The present invention uses facial expression recognition and voice emotion analysis to evaluate students' emotional states, which is rare in traditional online learning systems and can provide more refined learning experience analysis. The present invention not only evaluates students' learning effects, but also provides personalized learning suggestions based on the evaluation results, and adjusts teaching content and learning paths to achieve true adaptive learning.

[0132] Example 2

[0133] like Figure 2 As shown, in the second embodiment of the present invention, a large model-driven adaptive learning evaluation system is provided, and the system includes:

[0134] Verification module 1 is used to authenticate the user and obtain the user's behavior data and emotion data if the authentication is successful;

[0135] Behavior module 2, used for performing behavior analysis on the behavior data to output behavior analysis results;

[0136] Emotion module 3, used for performing emotional state analysis on the emotional data to obtain an emotional analysis result;

[0137] Evaluation module 4 is used to obtain training data, input the training data into a preset model for training and evaluating model performance to obtain a target model, input the behavior analysis results, the emotion analysis results, and the target learning plan into the target model to output an evaluation result;

[0138] The verification module 1 includes:

[0139] a first verification submodule, configured to obtain identity information of the user, and perform biometric identification on the user based on the identity information of the user to obtain a first verification result;

[0140] A second verification submodule is configured to obtain the user's login information and perform identity verification on the user based on the login information to obtain a second verification result;

[0141] a third verification submodule, configured to obtain the user's historical learning records, extract the user's learned knowledge information from the historical learning records, generate verification test questions based on the learned knowledge information, and perform learning content identification on the user based on the verification test questions to obtain a third verification result;

[0142] The final verification submodule is configured to obtain the behavior data and emotion data of the user if the first verification result, the second verification result, and the third verification result are all verified to be passed.

[0143] The behavior module 2 includes:

[0144] The first calculation submodule is used to determine the learning engagement based on the behavior data :

[0145] ;

[0146] Where, For standard study time, For the The actual study time of each student;

[0147] The second calculation submodule is used to determine the learning rhythm ratio based on the behavior data. :

[0148] ; ;

[0149] Where, For the The amount of learning content for each student, Indicates the The time required for a student to complete the learning content For the The learning completion speed of each student, The speed of completion of standard learning;

[0150] The third calculation submodule is used to determine the answer speed ratio based on the behavior data :

[0151] ;

[0152] Where, For the Students complete the target questions The actual response time, Target topic Standard response time;

[0153] The fourth calculation submodule is used to determine the answer order index based on the behavior data :

[0154] ;

[0155] Where, For the The difficulty of the question, is the total number of questions, Indicates the The difficulty of the first question a student chooses to answer;

[0156] The final calculation submodule is used to comprehensively consider the learning engagement, the learning rhythm ratio, the answering speed ratio, and the answering sequence index to obtain a behavior analysis result.

[0157] The emotion module 3 includes:

[0158] A feature point extraction submodule is used to extract a facial image from the emotion data and extract key facial feature points using a feature point detection algorithm to obtain facial feature data;

[0159] a first emotion recognition submodule, configured to input the facial feature data into a trained first emotion classifier to obtain a first emotion recognition result;

[0160] A preprocessing submodule, configured to extract voice data from the emotion data and preprocess the voice data to obtain processed voice data;

[0161] a second emotion recognition submodule, configured to extract acoustic features from the processed speech data, and input the acoustic features into a trained second emotion classifier to obtain a second emotion recognition result;

[0162] The final emotion recognition submodule is used to combine the first emotion recognition result and the second emotion recognition result to obtain an emotion analysis result.

[0163] The evaluation module 4 includes:

[0164] The prediction submodule is used to input the training data into the preset model for training and calculate the prediction results based on the training output results :

[0165] ;

[0166] Where, is the intercept parameter, is the error term, To expect test scores, are the first to seventh variable coefficients, Respectively Each student's actual learning time, learning pace ratio, answer speed ratio, answer order index, concentration, confusion, and learning engagement;

[0167] The mean square error submodule is used to calculate the mean square error based on the predicted score :

[0168] ;

[0169] Where, is the number of students in the training data, For the The actual test scores of each student;

[0170] Score submodule, used to calculate based on the predicted score Fraction :

[0171] ;

[0172] Where, is the average of the test scores;

[0173] Optimization submodule for 、 The preset model is optimized and trained to obtain the target model.

[0174] The evaluation module 4 is specifically used for:

[0175] The behavior analysis results, the emotion analysis results, and the target learning plan are spliced ​​and fused to obtain a comprehensive feature vector, and the comprehensive feature vector is input into the target model to obtain a learning evaluation result.

[0176] In other embodiments of the present invention, the embodiments of the present invention provide the following technical solutions: a computer comprising a memory 102, a processor 101, and a computer program stored on the memory 102 and executable on the processor 101; the processor 101 implements the large model-driven adaptive learning evaluation method as described above when executing the computer program.

[0177] Specifically, the processor 101 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured as one or more integrated circuits for implementing the embodiments of the present invention.

[0178] Memory 102 may include a large-capacity memory for data or instructions. By way of example, and not limitation, memory 102 may include a hard disk drive (HDD), a floppy disk drive, a solid-state drive (SSD), flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 102 may include removable or non-removable (or fixed) media. Where appropriate, memory 102 may be internal or external to the data processing device. In certain embodiments, memory 102 is non-volatile memory. In certain embodiments, memory 102 includes read-only memory (ROM) and random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically alterable ROM (EAROM) or a flash memory (FLASH), or a combination of two or more of these. Under appropriate circumstances, the RAM can be a static random access memory (SRAM) or a dynamic random access memory (DRAM), where the DRAM can be a fast page mode dynamic random access memory (FPMDRAM), an extended data output dynamic random access memory (EDODRAM), a synchronous dynamic random access memory (SDRAM), etc.

[0179] The memory 102 may be used to store or cache various data files that need to be processed and / or used for communication, as well as possible computer program instructions executed by the processor 101 .

[0180] The processor 101 implements the above-mentioned large model-driven adaptive learning evaluation method by reading and executing computer program instructions stored in the memory 102.

[0181] In some embodiments, the computer may further include a communication interface 103 and a bus 100. Figure 3 As shown, the processor 101 , the memory 102 , and the communication interface 103 are connected via a bus 100 and communicate with each other.

[0182] The communication interface 103 is used to implement communication between the various modules, devices, units, and / or equipment in the embodiments of the present invention. The communication interface 103 can also implement data communication with other components such as external devices, image / data acquisition equipment, databases, external storage, and image / data processing workstations.

[0183] Bus 100 includes hardware, software, or both, and couples components of a computer device to each other. Bus 100 includes, but is not limited to, at least one of the following: a data bus, an address bus, a control bus, an expansion bus, and a local bus. By way of example and not limitation, bus 100 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or a combination of two or more of these. Bus 100 may include one or more buses, where appropriate. Although embodiments of the present invention describe and illustrate a particular bus, the present invention contemplates any suitable bus or interconnect.

[0184] The computer can execute the large-model-driven adaptive learning evaluation method of the present invention based on the acquisition of the large-model-driven adaptive learning evaluation system, thereby realizing the large-model-driven adaptive learning evaluation.

[0185] In some further embodiments of the present invention, in combination with the above-mentioned large-model-driven adaptive learning evaluation method, the embodiments of the present invention provide the following technical solutions: a storage medium having a computer program stored thereon, which implements the above-mentioned large-model-driven adaptive learning evaluation method when executed by a processor.

[0186] Those skilled in the art will appreciate that the logic and / or steps represented in the flowcharts or otherwise described 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 an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device), or in conjunction with such instruction execution system, apparatus, or device. For purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by an instruction execution system, apparatus, or device, or in conjunction with such instruction execution system, apparatus, or device.

[0187] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0188] It should be understood that various components of the present invention may be implemented using hardware, software, firmware, or a combination thereof. In the aforementioned embodiments, multiple steps or methods may be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one or a combination of the following technologies known in the art may be used: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0189] The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0190] The above-described embodiments merely represent several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent. It should be noted that a person of ordinary skill in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, and these variations and improvements fall within the scope of protection of the present invention. Therefore, the scope of protection of the patent for this invention shall be determined by the appended claims.

Claims

1. A large model-driven adaptive learning evaluation method, characterized in that: include: Authenticate the user, and if the authentication is successful, obtain the user's behavioral data and emotional data; Performing behavior analysis on the behavior data to output behavior analysis results; Performing emotional state analysis on the emotional data to obtain an emotional analysis result; Acquire training data, input the training data into a preset model for training and evaluate model performance to obtain a target model, input the behavior analysis results, the emotion analysis results, and the target learning plan into the target model to output an evaluation result; The step of performing behavior analysis on the behavior data to output a behavior analysis result includes: Determine learning engagement based on the behavioral data : ； Where, For standard study time, For the The actual study time of each student; Determine a learning rhythm ratio based on the behavioral data : ； ； Where, For the The amount of learning content for each student, Indicates the The time required for a student to complete the learning content For the The learning completion speed of each student, The speed of completion of standard learning; Determine the answering speed ratio based on the behavioral data : ； Where, For the Students complete the target questions The actual response time, Target topic Standard response time; Determine an answer sequence indicator based on the behavioral data : ； Where, For the The difficulty of the question, is the total number of questions, Indicates the The difficulty of the first question a student chooses to answer; Combining the learning engagement, the learning rhythm ratio, the answering speed ratio, and the answering sequence index to obtain a behavior analysis result; The step of inputting the training data into a preset model for training and evaluating the model performance to obtain a target model includes: The training data is input into the preset model for training, and the predicted score is calculated based on the training output results. : ； Where, is the intercept parameter, is the error term, To expect test scores, are the first to seventh variable coefficients, Respectively Each student's actual learning time, learning pace ratio, answer speed ratio, answer order index, concentration, confusion, and learning engagement; Calculate the mean square error based on the predicted performance : ； Where, is the number of students in the training data, For the The actual test scores of each student; Calculated based on the predicted grades Fraction : ； Where, is the average of the test scores; Based on the mean square error 、 The preset model is optimized and trained to obtain the target model.

2. The large model-driven adaptive learning evaluation method according to claim 1 is characterized in that: The step of authenticating the user and obtaining the user's behavior data and emotion data if the authentication is successful includes: Obtaining identity information of the user, and performing biometric identification on the user based on the identity information of the user to obtain a first verification result; Obtaining the user's login information, and performing identity authentication on the user based on the login information to obtain a second verification result; Obtaining the user's historical learning records, extracting the user's learned knowledge information from the historical learning records, generating a verification test question based on the learned knowledge information, and performing learning content identification on the user based on the verification test question to obtain a third verification result; If the first verification result, the second verification result, and the third verification result are all verified to be passed, the behavior data and emotion data of the user are obtained.

3. The large model-driven adaptive learning evaluation method according to claim 1 is characterized in that: The step of performing emotional state analysis on the emotional data to obtain an emotional analysis result includes: Extracting a facial image from the emotion data, and extracting key facial feature points using a feature point detection algorithm to obtain facial feature data; Inputting the facial feature data into a trained first emotion classifier to obtain a first emotion recognition result; Extracting voice data from the emotion data, and preprocessing the voice data to obtain processed voice data; Extracting acoustic features from the processed speech data, and inputting the acoustic features into a trained second emotion classifier to obtain a second emotion recognition result; The first emotion recognition result and the second emotion recognition result are combined to obtain an emotion analysis result.

4. The large model-driven adaptive learning evaluation method according to claim 1, characterized in that: The step of inputting the behavior analysis result, the emotion analysis result, and the target learning plan into the target model to output the learning evaluation result is specifically as follows: The behavior analysis results, the emotion analysis results, and the target learning plan are spliced ​​and fused to obtain a comprehensive feature vector, and the comprehensive feature vector is input into the target model to obtain a learning evaluation result.

5. A large-model driven adaptive learning evaluation system, wherein the system adopts the large-model driven adaptive learning evaluation method according to claim 1, characterized in that: The system comprises: A verification module is used to authenticate the user and, if the authentication is successful, obtain the user's behavior data and emotion data; A behavior module, configured to perform behavior analysis on the behavior data to output a behavior analysis result; An emotion module, configured to perform emotion state analysis on the emotion data to obtain emotion analysis results; The evaluation module is used to obtain training data, input the training data into a preset model for training and evaluating the model performance to obtain a target model, input the behavior analysis results, the emotion analysis results, and the target learning plan into the target model to output an evaluation result.

6. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the adaptive learning evaluation method based on large model driving is implemented as described in any one of claims 1 to 4.

7. A storage medium, characterized in that: The storage medium stores a computer program, which, when executed by a processor, implements the large model-driven adaptive learning evaluation method according to any one of claims 1 to 4.

Images