Generative artificial intelligence teaching assisting system and device

Abstract

The invention focuses on artificial intelligence to be applied to the field of education, aims to solve the problems of a traditional education method and an existing artificial intelligence teaching assisting device, and provides a generative artificial intelligence teaching assisting system and a miniaturized device. According to the system, by means of a high-performance cloud server, a teaching assisting system client is built, a cloud resource library with classified storage, encryption management and rich resources is built, and a teaching feedback subsystem supporting multi-form feedback and intelligent analysis is developed. An advanced generative artificial intelligence program and a large model are accessed, and operations such as data preprocessing, training fine tuning, evaluation optimization, integration fusion and security privacy protection are carried out. The client application process comprises the links of receiving a registration request, selecting a teaching stage and a school category, establishing a file, uploading data, sending a resource demand instruction, calling and distributing resources and the like. The system has multiple functions of personalized teaching analysis, intelligent question and answer and the like. The beneficial effects are obvious, the miniaturization design is convenient for carrying and deployment, and the hardware cost is reduced; the teaching efficiency and quality are improved through multiple functions; the cloud architecture realizes resource sharing optimization and promotes artificial intelligence program learning; the teaching feedback subsystem enhances communication; the unique algorithm trains the accurate feedback capability, and provides powerful support for the technological innovation and development of the education industry.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence application technology, and in particular to an artificial intelligence teaching assistant system and device. Background Art

[0002] In today's era of rapid technological development, generative artificial intelligence large models have made substantial breakthroughs and quickly moved from the theoretical exploration stage to the stage of widespread large-scale application. This breakthrough has brought opportunities for change in many fields, and the education industry is no exception. Generative artificial intelligence is injecting strong impetus into the innovation of education, helping it to improve quality and efficiency, and becoming an important enabling factor in promoting the development of the education industry.

[0003] At present, although traditional education methods can meet basic teaching needs in long-term practice, with the development of the times and the advancement of educational concepts, their shortcomings are gradually becoming more prominent. As the core force of traditional teaching, teachers have limited energy and time, and it is difficult to formulate exclusive teaching plans for each student's unique learning progress, interests, hobbies, knowledge mastery and other personalized needs, resulting in some students not being able to receive sufficient attention and guidance. At the same time, the manual teaching model also faces challenges in teaching efficiency. Teachers need to spend a lot of time on repetitive tasks, such as correcting homework and preparing teaching materials, which limits the efficient use and sharing of teaching resources.

[0004] Although some AI teaching assistant devices have appeared on the market, which have alleviated some problems of traditional teaching methods to a certain extent, these devices generally have limitations. Their teaching scene design is relatively simple and it is difficult to adapt to the complex and changeable actual teaching needs; the AI ​​technology installed in the device is not smart and flexible enough, and its performance is relatively rigid, and it cannot make accurate and targeted responses according to the diverse teaching situations; and the device is generally large in size, which not only takes up a lot of space, but is also not convenient in deployment and use, which brings inconvenience to the equipment management and teaching implementation of educational institutions.

[0005] In response to these existing problems, the present invention proposes a teaching assistant system that integrates generative artificial intelligence with innovative thinking and technical means, and integrates it into a miniaturized device. The system and device innovatively integrate multiple functional modules, which can provide comprehensive and efficient auxiliary support for teachers in multiple key teaching links, and provide a new idea and effective solution to solve existing educational technology problems. Summary of the invention

[0006] The present invention aims to solve the many problems existing in the above-mentioned traditional teaching methods and existing artificial intelligence teaching assistant devices, and to provide a highly integrated generative artificial intelligence teaching assistant system and a miniaturized device to achieve a more efficient, personalized and intelligent teaching process, thereby improving the overall teaching quality and the efficiency of educational resource utilization.

[0007] The present invention is achieved through the following technical solutions:

[0008] A generative artificial intelligence teaching assistant system and device, wherein the generative artificial intelligence teaching assistant system is constructed by the following method:

[0009] Build the teaching assistant system client:

[0010] We selected cloud servers with high performance and high reliability, such as servers from mainstream cloud service providers such as Alibaba Cloud and Tencent Cloud, to build the teaching assistant client. Taking into account the concurrent access requirements in different educational scenarios, we reasonably configured the server's computing resources (such as the number of CPU cores, memory size) and storage capacity based on the estimation of user scale and access frequency. At the same time, we adopted load balancing technology to ensure that the client can still run stably and smoothly under high concurrency, providing a solid foundation for subsequent data processing and function realization.

[0011] Establish cloud resource library and teaching feedback subsystem:

[0012] Cloud resource library: Create a large and comprehensive cloud resource library on the cloud server. First, clarify the data architecture of the resource library, and classify and store it according to the types of teaching resources (such as documents, videos, audio, etc.), subjects (covering major subjects such as Chinese, mathematics, English, etc.), and applicable grades (from kindergarten to university stages) to facilitate efficient retrieval and call. Secondly, formulate a resource collection and review mechanism to collect teaching resources through various channels, including teacher uploads and screening of public resources on the Internet. All resources entering the resource library must be reviewed by professional educators to ensure the quality, accuracy and consistency of the resources with the education syllabus. In addition, use data encryption technology to encrypt and store sensitive data in the resource library to prevent data leakage.

[0013] Teaching feedback subsystem: Develop a dedicated teaching feedback subsystem, which will serve as an important bridge for communication between teachers, parents and students. It is developed using a microservice architecture to improve the scalability and maintainability of the system. In terms of functional design, different feedback portals are provided for teachers, parents and students. Teachers can provide feedback on problems encountered in the teaching process, suggestions for improving the resource library, etc.; parents can provide feedback on their children's learning situation at home, opinions on teaching methods, etc.; students can provide feedback on their confusion and learning experience in the learning process. The subsystem supports multiple feedback forms, such as text, pictures, voice, etc., to meet the usage habits of different users. At the same time, natural language processing technology is used to intelligently classify and analyze the feedback content to provide data support for subsequent problem solving and system optimization.

[0014] Access to generative AI programs and large models:

[0015] Introduce the current advanced generative artificial intelligence programs and large models adapted to them, such as the GPT series, BERT, etc. During the access process, a series of adaptation and optimization work needs to be done on the model.

[0016] Data preprocessing: Collect a large amount of education-related data, including textbook content, teaching cases, student homework, test papers, etc. Clean these data to remove duplicate, erroneous and irrelevant data. Then annotate them. For example, for student homework and test paper data, annotate the knowledge points, difficulty level, problem-solving ideas and other information. Divide the annotated data into training sets, validation sets and test sets according to a certain ratio to provide high-quality data support for model training.

[0017] Model training and fine-tuning: Based on the pre-trained large model, supervised training is performed using labeled education data. During the training process, a suitable loss function (such as the cross entropy loss function) is selected to measure the difference between the model's predicted results and the actual results, and the model's parameters are continuously adjusted through optimization algorithms (such as stochastic gradient descent and its variants Adagrad, Adadelta, Adam, etc.) to minimize the loss function and improve the model's performance in the field of education. In addition, the model is fine-tuned for specific teaching tasks and needs. For example, for the intelligent question-and-answer function, by adding relevant question-and-answer pairs to the training data, the model can better understand and answer questions in the field of education.

[0018] Model evaluation and optimization: Use the validation set to evaluate the trained model, and use indicators such as accuracy, recall, and F1 value to measure the performance of the model on different tasks (such as personalized teaching analysis, intelligent question and answer, etc.). According to the evaluation results, analyze the problems of the model, such as overfitting or underfitting. If overfitting occurs, you can use regularization techniques (such as L1, L2 regularization), increase the amount of training data, adjust the model structure, and other methods to optimize; if it is underfitting, you can consider increasing the complexity of the model, adjusting the training parameters, etc. In the process of model optimization, training, evaluation, and adjustment are repeated until the model performance reaches the expected goal.

[0019] Model integration and fusion: In order to further improve the stability and accuracy of the model, the model integration method can be used to combine multiple models with different structures or parameters. For example, the prediction results of multiple models can be combined through a voting mechanism or weighted average to reduce the error and uncertainty of a single model. In addition, the generative artificial intelligence model can also be integrated with other traditional machine learning models (such as decision trees, support vector machines, etc.) to give full play to the advantages of different models and improve the system's processing capabilities in complex educational scenarios.

[0020] Model security and privacy protection:

[0021] Security and privacy protection are crucial when building a generative AI teaching assistant system.

[0022] Data security: Strict access control is implemented on user data (such as students' personal information, learning records, etc.) and teaching resource data involved in the system. Identity authentication and authorization mechanisms are used to ensure that only authorized users and system components can access the corresponding data. At the same time, data is backed up regularly to prevent data loss due to hardware failures, natural disasters, etc.

[0023] Model security: Prevent the model from adversarial attacks, such as adversarial sample attacks, model stealing attacks, etc. Use adversarial training technology to introduce adversarial samples during the training process, so that the model has a certain degree of robustness against common adversarial attacks. In addition, monitor and filter the output of the model to prevent the generation of harmful, erroneous or inappropriate content.

[0024] Privacy protection: Comply with relevant privacy protection laws and regulations (such as GDPR, the Personal Information Protection Law of the People's Republic of China, etc.), and fully protect the privacy of users during data collection, storage, use and sharing. Differential privacy technology is used to perturb user data to a certain extent without affecting the model training effect, ensuring that even if the data is leaked, the user's personal information cannot be accurately identified from it.

[0025] System monitoring and maintenance:

[0026] Build a system monitoring platform to monitor the performance indicators of the generative AI teaching assistant system in real time, such as the server's CPU usage, memory usage, network bandwidth, etc., as well as the model's operating status (such as response time, accuracy fluctuations, etc.). When the indicator exceeds the preset threshold, an alarm is issued in time to notify the system administrator. At the same time, establish a regular system maintenance mechanism, including software updates, vulnerability repairs, data cleaning, etc., to ensure the long-term stable operation and performance optimization of the system.

[0027] Here, we can use a simple mathematical function to describe the input-output relationship of the model. Assuming that the input is the user's demand information X (which can be in vector form, containing multi-dimensional information such as teaching stage, subject, specific problem, etc.), and the output is the generated teaching resources or solutions Y, then the function Y = f (X) can be used to abstractly represent this process, where f represents the complex calculation process and algorithm logic inside the artificial intelligence model.

[0028] Teaching assistant system client application process:

[0029] Receiving registration requests: The teaching assistant client has a friendly user interface. Users can submit registration requests through an intuitive operation portal, and the system will respond and process the requests in real time.

[0030] Select teaching stage and school category: During the registration process, users need to accurately select the current teaching stage (such as specific class hours, complete courses, semesters or school years, etc.) and school category (covering kindergartens, primary schools, middle schools, vocational schools, universities, special schools and other educational institutions) from a variety of options based on actual teaching conditions. This step can help the system provide users with more targeted services and resource recommendations.

[0031] Create a dedicated profile file: The system will generate a unique profile file for each user based on the user's registration information. The profile file will serve as the user's personal data center in the system to store and manage the user's teaching-related information.

[0032] Upload learning behavior data: Users can conveniently upload various behavior data closely related to teaching and learning, including common teaching resources, student homework, test papers, classroom videos, excellent notes, learning videos and attendance records, as well as diversified data such as students' classroom performance evaluation and learning interest survey results. These data will provide rich materials for the system to conduct intelligent analysis and personalized teaching.

[0033] Send resource demand instructions: During the teaching process, if the user needs specific teaching resources, he can send clear resource demand instructions to the system through the client. After receiving the instructions, the system will quickly start the resource matching and scheduling mechanism.

[0034] Resource calling and allocation: Generative AI programs will quickly screen, call and allocate high-quality teaching resources that meet the current teaching stage and user needs from a huge cloud resource library based on user instructions, ensuring that users can obtain the required information in a timely manner.

[0035] In order to more accurately describe the resource allocation process, a weight function can be introduced. Assume that there are n dimensions in the teaching stage (such as grade, subject difficulty, etc.), and the weight of each dimension is w i (i=1,2,···,n), and For a resource R, its matching degree with user needs can be expressed by the function To calculate, where s i (R) represents the matching score between resource R and user demand in the i-th dimension. The system will select and allocate the most appropriate resource based on the size of M(R).

[0036] Client menu interface application:

[0037] Personalized teaching analysis application: When the user selects this application, the system will make full use of the high-definition camera installed on the device, and through advanced image recognition and behavior analysis technology, continuously and accurately record each student's academic progress, including classroom participation, homework completion, knowledge mastery and other multi-dimensional data. Then, combined with the rich teaching cases and educational theories in the artificial intelligence cloud resource library, according to each student's academic progress in the semester, a personalized teaching plan is tailored for each student to ensure that the teaching content and methods can meet the students' personalized learning needs to the greatest extent.

[0038] When analyzing the degree of students' mastery of knowledge points, we can use probability functions to quantify them. Let the probability of students mastering a certain knowledge point k be P k , which can be calculated by the student's performance in related assignments and tests. For example, if a student gets n correct answers in m tests involving this knowledge point, then To preliminarily estimate the probability of mastering the content, the system can then formulate a corresponding teaching plan based on this probability.

[0039] Smart Question and Answer Application: After entering the application, users can input text by touching the screen, or directly use the voice interaction function to communicate with the system and issue various operation instructions. After receiving the instructions, the system will quickly mobilize the massive knowledge in the artificial intelligence cloud resource library, use natural language processing technology to deeply understand and analyze the questions, and then generate comprehensive and accurate answers in various forms including text, pictures, detailed courseware, etc., to provide users with strong teaching support.

[0040] For the natural language processing part, a word vector model can be used to represent the problem. Assume that a question consists of N words, each word can be represented as a vector Then the whole problem can be expressed as a vector set The system processes this vector set through a neural network model (which can be represented by function g) to obtain the semantic representation of the problem. The answer is then retrieved and generated from the resource repository based on this semantic representation.

[0041] Interactive teaching application:

[0042] Online interactive teaching: After the user selects the online interactive teaching function in the teaching assistant client, the system will start the high-definition camera to record the real-time video, and record the teaching process completely and clearly. At the same time, with the help of network communication technology, real-time interactive communication with remote students can be achieved, such as online Q&A, group discussions, etc., breaking geographical restrictions and expanding teaching space.

[0043] Offline interactive teaching: When choosing offline interactive teaching, users can use the convenient operation tool of flying mouse to seamlessly connect with the projection teaching whiteboard to realize flexible display and operation of teaching content. The system will intelligently generate learning auxiliary materials related to the teaching content based on the real-time captured teaching video and the user's teaching status, and conduct efficient on-site interactive teaching with students, enhancing the fun and participation of teaching.

[0044] In the interactive teaching process, in order to measure the student's participation, a participation function can be introduced. Assume that students have p kinds of participation behaviors in the interactive teaching process (such as speaking, asking questions, submitting homework, etc.), and the frequency of each behavior is f i (i=1,···,p), the importance weight of each behavior is Then the student's participation E can be expressed as The system can adjust teaching strategies based on this engagement value.

[0045] Learning effectiveness evaluation application: After the user opens the application, the camera will capture the student's learning process and homework status in real time, and comprehensively record the student's performance in various aspects such as classroom learning, homework completion, and test answering. The system will carefully compare the acquired data with the standard data and excellent cases in the artificial intelligence cloud resource library, use scientific evaluation algorithms to correct students' homework in real time, and comprehensively summarize and deeply evaluate students' learning effects from multiple dimensions such as knowledge mastery, learning ability, and thinking development. Finally, the detailed learning effect evaluation report will be promptly fed back to the teaching feedback subsystem, so that teachers, parents, and students can view it at any time to adjust teaching and learning strategies.

[0046] When evaluating learning outcomes, a comprehensive evaluation function can be used. Suppose that the evaluation is conducted from q dimensions, such as knowledge mastery, learning ability, and thinking development, and the scores of each dimension are s1, s2, s q , the weight of each dimension is Then the comprehensive score S of the student's learning effectiveness can be expressed as This score can intuitively reflect the students' overall learning effect.

[0047] Course attendance record application: After the user enters the application, the camera will automatically turn on the real-time monitoring function to accurately record students' attendance and early class leave. The system will promptly upload these attendance data to the artificial intelligence cloud resource library for storage and management, providing objective and accurate data support for the teaching management department. At the same time, the system will perform intelligent analysis of the attendance data, generate attendance statistics and trend analysis charts, and help teachers and administrators intuitively understand students' attendance patterns and changing trends.

[0048] To analyze attendance trends, we can use time series analysis functions. Assume that the attendance of students is recorded in days, with attendance as 1 and absence as 0. In a period of time t1, t2, ···, t T The attendance sequence in is {x1,x2,···,x T}. You can use the moving average function (where n is the time window size of the moving average) to smooth the data and observe attendance trends.

[0049] Real-time creation application: When the user selects the real-time creation application, the user takes pictures of the student's work or other related pictures through the camera and uploads them to the system. The artificial intelligence program in the system will use a variety of technologies such as image recognition, content understanding and creative generation to quickly generate creative results in various forms such as pictures, exquisite picture books, and vivid videos based on the content of the pictures, adding rich creative elements and teaching materials to teaching activities.

[0050] In the image recognition and generation part, the related functions of convolutional neural network (CNN) can be used. For the input image I, after a series of convolutional layers, pooling layers and other operations (which can be represented by function h), the feature representation is obtained. Then, based on this feature representation, a new creative result is generated through a generative network (which can be represented by function k)

[0051] The hardware components of the artificial intelligence teaching assistant system are as follows:

[0052] The artificial intelligence teaching assistant system device is composed of a body (1), a front cover (4), a middle seat (3), and a rear cover (2).

[0053] Front cover: A high-resolution, sensitive touch screen (6) is installed on the front cover (4), and the user can conveniently interact with the system through touch operation to call and set various functions.

[0054] Middle seat: The middle seat (3) is equipped with a variety of interfaces and devices, including a universal USB port (7) for users to connect external storage devices, input devices, etc.; a stable and reliable power socket (12) to provide continuous power support for the device; a high-speed network port (11) to ensure that the device can stably access the network, realize rapid data transmission and real-time acquisition of resources; high-quality speakers (13) to ensure clear sound playback during the teaching process; and a motherboard that carries the core computing and control functions of the device.

[0055] Both sides of the machine body: A high-definition, wide-angle camera (5) is installed on the right side of the machine body (1), which can clearly capture various details in the teaching scene; a convenient flying mouse (8) is installed on the left side, which is convenient for users to flexibly operate the projected content during offline interactive teaching; a high-performance network antenna (9) enhances the network signal receiving capability; and an easy-to-operate power button (10).

[0056] Furthermore, the teaching assistant client and cloud resource library, when constructing the teaching assistant client and cloud resource library, make full use of the distributed storage and computing technology provided by the cloud server to ensure the security of data storage and the efficiency of system operation. Use data encryption and backup mechanisms to prevent data loss and leakage. At the same time, regularly clean and optimize the cloud resource library to ensure the accuracy and timeliness of resources.

[0057] Furthermore, the artificial intelligence program and its algorithm deeply integrate the supervised learning algorithm and annealing algorithm with the cloud resource library, continuously update and expand the training data, so that the artificial intelligence program can continue to learn and evolve. The reinforcement learning mechanism is introduced to dynamically adjust the algorithm parameters according to user feedback and usage effects, further improving the feedback accuracy and response speed of the artificial intelligence program for various functional applications such as personalized teaching analysis and intelligent question and answer.

[0058] In supervised learning algorithms, loss functions can be used to measure the difference between the model's predicted results and the actual results. For example, the mean square error loss function where θ is the parameter of the model, y i is the true value, is the predicted value of the model, and N is the number of samples. By minimizing this loss function, the model parameters are optimized and the prediction accuracy of the model is improved.

[0059] In the annealing algorithm, the change of temperature T with time t can be expressed by the function T(t) = T0·a t To express it, T0 is the initial temperature, a (0<a<1) is the cooling coefficient, and the global search and local search capabilities of the algorithm are balanced by controlling the temperature drop process to find a better solution.

[0060] Furthermore, the artificial intelligence teaching assistant system is an organic collection, which is composed of a teaching assistant client, a cloud resource library, a teaching feedback subsystem and an artificial intelligence program working together. In order to improve the user experience, the system is designed to focus on seamless connection and data sharing between modules. After the user logs in to the client through the touch screen (6), not only can the user operate by touch, but also the user can make full use of advanced speech recognition and synthesis technology to achieve more natural and convenient voice operation and improve operation efficiency.

[0061] Furthermore, the teaching feedback subsystem, in addition to being installed on the PC and Android terminals, will also develop a version adapted to the iOS system to cover a wider user group. At the same time, the form of feedback content will be enriched. In addition to text feedback, multimedia feedback methods such as pictures and videos will be supported to make feedback information more intuitive and comprehensive. A feedback processing mechanism will be established to timely classify, process and respond to user feedback to improve user satisfaction.

[0062] Furthermore, the personalized teaching analysis, intelligent question and answer, interactive teaching, learning effectiveness evaluation, course attendance record and real-time creation are functional applications in the artificial intelligence teaching assistant system, and the following functions are continuously upgraded and expanded. For example, big data analysis and machine learning technology are introduced in personalized teaching analysis to more accurately predict students' learning trends; multi-language support functions are added to intelligent question and answer to meet the needs of different regions and students; virtual reality (VR) and augmented reality (AR) technologies are integrated into interactive teaching to create a more immersive teaching experience.

[0063] Furthermore, the artificial intelligence teaching assistant system and device are miniaturized devices that can flexibly adapt to various teaching environments.

[0064] The beneficial effects brought by the present invention are:

[0065] First, the miniaturized design of the present invention makes the AI ​​teaching assistant system and device easier to carry and deploy, and can flexibly adapt to various teaching environments. Compared with the large-volume and single-function teaching assistants on the market, the present invention has obvious advantages in space occupation, greatly reducing the investment cost of educational institutions in hardware equipment.

[0066] Secondly, the present invention integrates multiple functional applications, such as personalized teaching analysis, intelligent question and answer, interactive teaching, learning effectiveness evaluation, course attendance records and real-time creation, etc. The organic integration of these functional applications effectively improves the teaching efficiency and quality. Through the intelligent processing of the artificial intelligence teaching assistant system, teachers can understand the students' learning status more accurately and provide personalized teaching plans for each student, thereby improving the teaching quality and promoting the all-round development of students.

[0067] Thirdly, the present invention uses an efficient and stable cloud server to build a teaching assistant client, and establishes a cloud resource library and a teaching feedback subsystem to achieve the sharing and optimization of teaching resources. The establishment of a cloud resource library enables the effective integration of teaching resources, and teachers can quickly call and allocate resources according to teaching needs, thereby improving teaching efficiency. At the same time, the continuous enrichment and improvement of the cloud resource library provides a basis for the self-learning and optimization of the artificial intelligence program, so that the system can better adapt to the educational characteristics of different schools.

[0068] At the same time, the development of the teaching feedback subsystem provides a convenient communication platform for teachers, parents and students, which helps to improve education participation and satisfaction.

[0069] Finally, the artificial intelligence program and its algorithm of the present invention are repeatedly trained through supervised learning algorithms and annealing algorithms combined with an ever-increasing cloud resource library, and gradually train themselves to be able to respond quickly and provide accurate feedback on functional applications such as personalized teaching analysis.

[0070] In summary, the artificial intelligence teaching assistant system and device of the present invention have shown significant beneficial effects in improving teaching efficiency, optimizing teaching resources, realizing personalized teaching, and promoting educational equity, providing strong support for technological innovation and development of my country's education industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 Functional application diagram of the artificial intelligence teaching assistant system

[0072] Figure 2 Device diagram for artificial intelligence teaching assistant DETAILED DESCRIPTION

[0073] The present invention will be described in detail below in conjunction with the accompanying drawings.

[0074] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0075] like Figure 1 As shown, the artificial intelligence teaching assistant system of this embodiment can be applied to various teaching environments. In this embodiment, the artificial intelligence teaching assistant system is constructed by the following steps: Use an efficient and stable cloud server, such as Alibaba Cloud Server, to build a teaching assistant client. Establish a cloud resource library on the cloud server to store data such as teaching resources, student homework, test papers, classroom videos, and attendance records. Establish a teaching feedback subsystem to feedback student status information to users. The user logs in to the system through the teaching assistant client and selects the current teaching stage and school category. The artificial intelligence program calls and allocates resources that meet the teaching stage from the cloud resource library according to user needs. Users can upload various types of learning behavior data to the teaching assistant client. The system will synchronize these data to the cloud resource library in real time, and through intelligent analysis, learning and cloud resources, realize the intelligent allocation of teaching resources for the characteristics of the user's school. The artificial intelligence teaching assistant system summarizes the student status information in the cloud resource library every day to form a report. Feedback student status information to users through the teaching feedback subsystem.

[0076] In order to facilitate understanding of the above technical solutions of the present invention, the working principle or operation mode of the present invention in the actual process is described in detail below. Figure 2 As shown, after the device of the artificial intelligence teaching assistant system is turned on, the user logs in to the client and can use the touch screen to enter the menu to select the artificial intelligence application he needs, and combine different hardware tools to complete different functions to assist users in better teaching. For example, to use the learning effect evaluation function application, the user logs in to the client and turns on the camera to take real-time photos of the student's learning process or homework, test papers. After obtaining the data from the artificial intelligence cloud resource library, the artificial intelligence teaching assistant system compares the student's homework, then makes real-time corrections, summarizes and evaluates the learning effect, and finally feeds back the learning effect evaluation to the user in a timely manner through the teaching feedback subsystem.

[0077] In summary, with the help of the above technical solutions of the present invention, the present invention realizes the optimal configuration of teaching resources and the intelligent management of the teaching process by integrating artificial intelligence, artificial intelligence cloud resource library, personalized teaching application, intelligent question and answer application, interactive teaching application, learning effect evaluation application and attendance record summary application, real-time creation application and other artificial intelligence functional applications. Compared with traditional education methods, the present invention can significantly improve teaching efficiency, reduce teachers' workload, and enable teachers to focus more on improving teaching quality and teaching effect.

[0078] The principles and implementation methods of the present invention are described in this article using specific embodiments. The description of the above embodiments is only used to help understand the method and core ideas of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made to the present invention without departing from the principles of the present invention, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.

Claims

1. An artificial intelligence teaching assistant system and device, characterized in that: include: The present invention aims to solve the many problems existing in the above-mentioned traditional teaching methods and existing artificial intelligence teaching assistant devices, and to provide a highly integrated generative artificial intelligence teaching assistant system and a miniaturized device to achieve a more efficient, personalized and intelligent teaching process, thereby improving the overall teaching quality and the efficiency of educational resource utilization. The present invention is achieved through the following technical solutions: A generative artificial intelligence teaching assistant system and device, wherein the generative artificial intelligence teaching assistant system is constructed by the following method: Build the teaching assistant system client: We selected cloud servers with high performance and high reliability, such as servers from mainstream cloud service providers such as Alibaba Cloud and Tencent Cloud, to build the teaching assistant client. Taking into account the concurrent access requirements in different educational scenarios, we reasonably configured the server's computing resources (such as the number of CPU cores, memory size) and storage capacity based on the estimation of user scale and access frequency. At the same time, we adopted load balancing technology to ensure that the client can still run stably and smoothly under high concurrency, providing a solid foundation for subsequent data processing and function realization. Establish cloud resource library and teaching feedback subsystem: Cloud resource library: Create a large and comprehensive cloud resource library on the cloud server. First, clarify the data architecture of the resource library, and classify and store it according to the types of teaching resources (such as documents, videos, audio, etc.), subjects (covering major subjects such as Chinese, mathematics, English, etc.), and applicable grades (from kindergarten to university stages) to facilitate efficient retrieval and call. Secondly, formulate a resource collection and review mechanism to collect teaching resources through various channels, including teacher uploads and screening of public resources on the Internet. All resources entering the resource library must be reviewed by professional educators to ensure the quality, accuracy and consistency of the resources with the education syllabus. In addition, use data encryption technology to encrypt and store sensitive data in the resource library to prevent data leakage. Teaching feedback subsystem: Develop a dedicated teaching feedback subsystem, which will serve as an important bridge for communication between teachers, parents and students. It is developed using a microservice architecture to improve the scalability and maintainability of the system. In terms of functional design, different feedback portals are provided for teachers, parents and students. Teachers can provide feedback on problems encountered in the teaching process, suggestions for improving the resource library, etc.; parents can provide feedback on their children's learning situation at home, opinions on teaching methods, etc.; students can provide feedback on their confusion and learning experience in the learning process. The subsystem supports multiple feedback forms, such as text, pictures, voice, etc., to meet the usage habits of different users. At the same time, natural language processing technology is used to intelligently classify and analyze the feedback content to provide data support for subsequent problem solving and system optimization. Access to generative AI programs and large models: Introduce the current advanced generative artificial intelligence programs and large models adapted to them, such as the GPT series, BERT, etc. During the access process, a series of adaptation and optimization work needs to be done on the model. Data preprocessing: Collect a large amount of education-related data, including textbook content, teaching cases, student homework, test papers, etc. Clean these data to remove duplicate, erroneous and irrelevant data. Then annotate them. For example, for student homework and test paper data, annotate the knowledge points, difficulty level, problem-solving ideas and other information. Divide the annotated data into training sets, validation sets and test sets according to a certain ratio to provide high-quality data support for model training. Model training and fine-tuning: Based on the pre-trained large model, supervised training is performed using labeled education data. During the training process, a suitable loss function (such as the cross entropy loss function) is selected to measure the difference between the model's predicted results and the actual results, and the model's parameters are continuously adjusted through optimization algorithms (such as stochastic gradient descent and its variants Adagrad, Adadelta, Adam, etc.) to minimize the loss function and improve the model's performance in the field of education. In addition, the model is fine-tuned for specific teaching tasks and needs. For example, for the intelligent question-and-answer function, by adding relevant question-and-answer pairs to the training data, the model can better understand and answer questions in the field of education. Model evaluation and optimization: Use the validation set to evaluate the trained model, and use indicators such as accuracy, recall, and F1 value to measure the performance of the model on different tasks (such as personalized teaching analysis, intelligent question and answer, etc.). According to the evaluation results, analyze the problems of the model, such as overfitting or underfitting. If overfitting occurs, you can use regularization techniques (such as L1, L2 regularization), increase the amount of training data, adjust the model structure, and other methods to optimize; if it is underfitting, you can consider increasing the complexity of the model, adjusting the training parameters, etc. In the process of model optimization, training, evaluation, and adjustment are repeated until the model performance reaches the expected goal. Model integration and fusion: In order to further improve the stability and accuracy of the model, the model integration method can be used to combine multiple models with different structures or parameters. For example, the prediction results of multiple models can be combined through a voting mechanism or weighted average to reduce the error and uncertainty of a single model. In addition, the generative artificial intelligence model can also be integrated with other traditional machine learning models (such as decision trees, support vector machines, etc.) to give full play to the advantages of different models and improve the system's processing capabilities in complex educational scenarios. Model security and privacy protection: Security and privacy protection are crucial when building a generative AI teaching assistant system. Data security: Strict access control is implemented on user data (such as students' personal information, learning records, etc.) and teaching resource data involved in the system. Identity authentication and authorization mechanisms are used to ensure that only authorized users and system components can access the corresponding data. At the same time, data is backed up regularly to prevent data loss due to hardware failures, natural disasters, etc. Model security: Prevent the model from adversarial attacks, such as adversarial sample attacks, model stealing attacks, etc. Use adversarial training technology to introduce adversarial samples during the training process, so that the model has a certain degree of robustness against common adversarial attacks. In addition, monitor and filter the output of the model to prevent the generation of harmful, erroneous or inappropriate content. Privacy protection: Comply with relevant privacy protection laws and regulations (such as GDPR, the Personal Information Protection Law of the People's Republic of China, etc.), and fully protect the privacy of users during data collection, storage, use and sharing. Differential privacy technology is used to perturb user data to a certain extent without affecting the model training effect, ensuring that even if the data is leaked, the user's personal information cannot be accurately identified from it. System monitoring and maintenance: Build a system monitoring platform to monitor the performance indicators of the generative AI teaching assistant system in real time, such as the server's CPU usage, memory usage, network bandwidth, etc., as well as the model's operating status (such as response time, accuracy fluctuations, etc.). When the indicator exceeds the preset threshold, an alarm is issued in time to notify the system administrator. At the same time, establish a regular system maintenance mechanism, including software updates, vulnerability repairs, data cleaning, etc., to ensure the long-term stable operation and performance optimization of the system. Here, we can use a simple mathematical function to describe the input-output relationship of the model. Assuming that the input is the user's demand information X (which can be in vector form, containing multi-dimensional information such as teaching stage, subject, specific problem, etc.), and the output is the generated teaching resources or solutions Y, then the function Y = f (X) can be used to abstractly represent this process, where f represents the complex calculation process and algorithm logic inside the artificial intelligence model. Teaching assistant system client application process: Receiving registration requests: The teaching assistant client has a friendly user interface. Users can submit registration requests through an intuitive operation portal, and the system will respond and process the requests in real time. Select teaching stage and school category: During the registration process, users need to accurately select the current teaching stage (such as specific class hours, complete courses, semesters or school years, etc.) and school category (covering kindergartens, primary schools, middle schools, vocational schools, universities, special schools and other educational institutions) from a variety of options based on actual teaching conditions. This step can help the system provide users with more targeted services and resource recommendations. Create a dedicated profile file: The system will generate a unique profile file for each user based on the user's registration information. The profile file will serve as the user's personal data center in the system to store and manage the user's teaching-related information. Upload learning behavior data: Users can conveniently upload various behavior data closely related to teaching and learning, including common teaching resources, student homework, test papers, classroom videos, excellent notes, learning videos and attendance records, as well as diversified data such as students' classroom performance evaluation and learning interest survey results. These data will provide rich materials for the system to conduct intelligent analysis and personalized teaching. Send resource demand instructions: During the teaching process, if the user needs specific teaching resources, he can send clear resource demand instructions to the system through the client. After receiving the instructions, the system will quickly start the resource matching and scheduling mechanism. Resource calling and allocation: Generative AI programs will quickly screen, call and allocate high-quality teaching resources that meet the current teaching stage and user needs from a huge cloud resource library based on user instructions, ensuring that users can obtain the required information in a timely manner. In order to more accurately describe the resource allocation process, a weight function can be introduced. Assume that there are n dimensions in the teaching stage (such as grade, subject difficulty, etc.), and the weight of each dimension is w i (i=1,2,···,n), and For a resource R, its matching degree with user needs can be expressed by the function To calculate, where s i (R) represents the matching score between resource R and user demand in the i-th dimension. The system will select and allocate the most appropriate resource based on the size of M(R). Client menu interface application: Personalized teaching analysis application: When the user selects this application, the system will make full use of the high-definition camera installed on the device, and through advanced image recognition and behavior analysis technology, continuously and accurately record each student's academic progress, including classroom participation, homework completion, knowledge mastery and other multi-dimensional data. Then, combined with the rich teaching cases and educational theories in the artificial intelligence cloud resource library, according to each student's academic progress in the semester, a personalized teaching plan is tailored for each student to ensure that the teaching content and methods can meet the students' personalized learning needs to the greatest extent. When analyzing the degree of students' mastery of knowledge points, we can use probability functions to quantify them. Let the probability of students mastering a certain knowledge point k be P k , which can be calculated by the student's performance in related assignments and tests. For example, if a student gets n correct answers in m tests involving this knowledge point, then To preliminarily estimate the probability of mastering the content, the system can then formulate a corresponding teaching plan based on this probability. Smart Question and Answer Application: After entering the application, users can input text by touching the screen, or directly use the voice interaction function to communicate with the system and issue various operation instructions. After receiving the instructions, the system will quickly mobilize the massive knowledge in the artificial intelligence cloud resource library, use natural language processing technology to deeply understand and analyze the questions, and then generate comprehensive and accurate answers in various forms including text, pictures, detailed courseware, etc., to provide users with strong teaching support. For the natural language processing part, a word vector model can be used to represent the problem. Assume that a question consists of N words, each word can be represented as a vector Then the whole problem can be expressed as a vector set The system processes this vector set through a neural network model (which can be represented by function g) to obtain the semantic representation of the problem. The answer is then retrieved and generated from the resource repository based on this semantic representation. Interactive teaching application: Online interactive teaching: After the user selects the online interactive teaching function in the teaching assistant client, the system will start the high-definition camera to record the real-time video, and record the teaching process completely and clearly. At the same time, with the help of network communication technology, real-time interactive communication with remote students can be achieved, such as online Q&A, group discussions, etc., breaking geographical restrictions and expanding teaching space. Offline interactive teaching: When choosing offline interactive teaching, users can use the convenient operation tool of flying mouse to seamlessly connect with the projection teaching whiteboard to realize flexible display and operation of teaching content. The system will intelligently generate learning auxiliary materials related to the teaching content based on the real-time captured teaching video and the user's teaching status, and conduct efficient on-site interactive teaching with students, enhancing the fun and participation of teaching. In the interactive teaching process, in order to measure the student's participation, a participation function can be introduced. Assume that students have p kinds of participation behaviors in the interactive teaching process (such as speaking, asking questions, submitting homework, etc.), and the frequency of each behavior is f i (i=1,···,p), the importance weight of each behavior is Then the student's participation E can be expressed as The system can adjust teaching strategies based on this engagement value. Learning effectiveness evaluation application: After the user opens the application, the camera will capture the student's learning process and homework status in real time, and comprehensively record the student's performance in various aspects such as classroom learning, homework completion, and test answering. The system will carefully compare the acquired data with the standard data and excellent cases in the artificial intelligence cloud resource library, use scientific evaluation algorithms to correct students' homework in real time, and comprehensively summarize and deeply evaluate students' learning effects from multiple dimensions such as knowledge mastery, learning ability, and thinking development. Finally, the detailed learning effect evaluation report will be promptly fed back to the teaching feedback subsystem, so that teachers, parents, and students can view it at any time to adjust teaching and learning strategies. When evaluating learning outcomes, a comprehensive evaluation function can be used. Suppose that the evaluation is conducted from q dimensions, such as knowledge mastery, learning ability, and thinking development, and the scores of each dimension are s1, s2, s q , the weight of each dimension is Then the comprehensive score S of the student's learning effectiveness can be expressed as This score can intuitively reflect the students' overall learning effect. Course attendance record application: After the user enters the application, the camera will automatically turn on the real-time monitoring function to accurately record students' attendance and early class leave. The system will promptly upload these attendance data to the artificial intelligence cloud resource library for storage and management, providing objective and accurate data support for the teaching management department. At the same time, the system will perform intelligent analysis of the attendance data, generate attendance statistics and trend analysis charts, and help teachers and administrators intuitively understand students' attendance patterns and changing trends. To analyze attendance trends, we can use time series analysis functions. Assume that the attendance of students is recorded in days, with attendance as 1 and absence as 0. In a period of time t1, t2, ···, t T The attendance sequence in is {x1,x2,···,x T }. You can use the moving average function (where n is the time window size of the moving average) to smooth the data and observe attendance trends. Real-time creation application: When the user selects the real-time creation application, the user takes pictures of the student's work or other related pictures through the camera and uploads them to the system. The artificial intelligence program in the system will use a variety of technologies such as image recognition, content understanding and creative generation to quickly generate creative results in various forms such as pictures, exquisite picture books, and vivid videos based on the content of the pictures, adding rich creative elements and teaching materials to teaching activities. In the image recognition and generation part, the related functions of convolutional neural network (CNN) can be used. For the input image I, after a series of convolutional layers, pooling layers and other operations (which can be represented by function h), the feature representation is obtained. Then, based on this feature representation, a new creative result is generated through a generative network (which can be represented by function k) The hardware components of the artificial intelligence teaching assistant system are as follows: The artificial intelligence teaching assistant system device is composed of a body (1), a front cover (4), a middle seat (3), and a rear cover (2). Front cover: A high-resolution, sensitive touch screen (6) is installed on the front cover (4), and the user can conveniently interact with the system through touch operation to call and set various functions. Middle seat: The middle seat (3) is equipped with a variety of interfaces and devices, including a universal USB socket (7) to facilitate users to connect external storage devices, input devices, etc.; a stable and reliable power socket (12) to provide continuous power support for the device; a high-speed network socket (11) to ensure that the device can stably access the network, realize fast data transmission and real-time acquisition of resources; high-quality audio (13) to ensure clear sound playback during the teaching process; and a motherboard that carries the device's core computing and control functions. Both sides of the machine body: A high-definition, wide-angle camera (5) is installed on the right side of the machine body (1), which can clearly capture various details in the teaching scene; a convenient flying mouse (8) is installed on the left side, which is convenient for users to flexibly operate the projected content during offline interactive teaching; a high-performance network antenna (9) enhances the network signal receiving capability; and an easy-to-operate power button (10).

2. An artificial intelligence teaching assistant system and device according to claim 1, characterized in that: The teaching assistant client and cloud resource library, when constructing the teaching assistant client and cloud resource library, make full use of the distributed storage and computing technology provided by the cloud server to ensure the security of data storage and the efficiency of system operation. Data encryption and backup mechanisms are used to prevent data loss and leakage. At the same time, the cloud resource library is regularly cleaned and optimized to ensure the accuracy and timeliness of resources.

3. An artificial intelligence teaching assistant system and device according to claim 1, characterized in that: The artificial intelligence program and its algorithm deeply integrate the supervised learning algorithm and annealing algorithm with the cloud resource library, continuously update and expand the training data, so that the artificial intelligence program can continue to learn and evolve. The reinforcement learning mechanism is introduced to dynamically adjust the algorithm parameters according to user feedback and usage effects, further improving the feedback accuracy and response speed of the artificial intelligence program for various functional applications such as personalized teaching analysis and intelligent question and answer. In supervised learning algorithms, loss functions can be used to measure the difference between the model's predicted results and the actual results. For example, the mean square error loss function where θ is the parameter of the model, y i is the true value, is the predicted value of the model, and N is the number of samples. By minimizing this loss function, the model parameters are optimized and the prediction accuracy of the model is improved. In the annealing algorithm, the change of temperature T with time t can be expressed by the function T(t) = T0·a t To express it, T0 is the initial temperature, a (0<a<1) is the cooling coefficient, and the global search and local search capabilities of the algorithm are balanced by controlling the temperature drop process to find a better solution.

4. An artificial intelligence teaching assistant system and device according to claim 1, characterized in that: The artificial intelligence teaching assistant system is an organic collection, which is composed of a teaching assistant client, a cloud resource library, a teaching feedback subsystem and an artificial intelligence program working together. In order to improve the user experience, the system is designed to focus on seamless connection and data sharing between modules. After the user logs in to the client through the touch screen (6), not only can the user operate by touch, but also can make full use of advanced speech recognition and synthesis technology to achieve more natural and convenient voice operation and improve operation efficiency.

5. An artificial intelligence teaching assistant system and device according to claim 1, characterized in that: The teaching feedback subsystem, in addition to being installed on the PC and Android terminals, will also develop a version adapted to the iOS system to cover a wider user group. At the same time, the form of feedback content will be enriched. In addition to text feedback, multimedia feedback methods such as pictures and videos will be supported to make feedback information more intuitive and comprehensive. A feedback processing mechanism will be established to timely classify, process and respond to user feedback to improve user satisfaction.

6. An artificial intelligence teaching assistant system and device according to claim 1, characterized in that: The personalized teaching analysis, intelligent question and answer, interactive teaching, learning effectiveness evaluation, course attendance record and real-time creation are the functional applications in the artificial intelligence teaching assistant system, and the following functions are continuously upgraded and expanded. For example, big data analysis and machine learning technology are introduced in personalized teaching analysis to more accurately predict students' learning trends; multi-language support functions are added to intelligent question and answer to meet the needs of different regions and students; virtual reality (VR) and augmented reality (AR) technologies are integrated into interactive teaching to create a more immersive teaching experience.

7. An artificial intelligence teaching assistant system and device according to claim 1, characterized in that: The artificial intelligence teaching assistant system and device are miniaturized devices that can flexibly adapt to various teaching environments.

Images