Course and AI partner training personalized learning method and system

Abstract

The embodiment of the invention provides a course and AI partner training personalized learning method and system. The method comprises the following steps executed by a computer program: S1, providing a theoretical course of at least one field for a user; s2, after the user completes theoretical course learning, generating a personalized practice scene based on the content and learning condition of the theoretical course learned by the user; s3, in the personalized practice scene, enabling an AI large model to simulate a virtual role through role playing, and carrying out interactive dialogue practice with a user based on the virtual role; s4, enabling the AI large model to provide feedback or suggestions according to the performance of the user in the interactive dialogue practice. The embodiment of the invention aims to create a personalized learning scheme integrating course learning and AI partner training, so that a user can perform high-simulation dialogue training through AI role playing on the basis of learning related theoretical courses.

Description

Technical Field

[0001] The present invention relates to the field of personalized learning technology, and in particular to a personalized learning method and system combining course learning with AI training. Background Art

[0002] In the existing technology, users often learn communication theories through videos, books or online courses. However, these methods can only provide static knowledge input, and users lack a practical practice and feedback environment. Although some existing AI dialogue systems can simulate dialogues, most of them are limited to a single scene or fixed templates, and cannot truly support users' personalized needs in complex situations such as parent-child communication and intimate relationships. Existing solutions lack a seamless connection between theoretical learning and practical operations, making it difficult for users to internalize and apply the knowledge they have learned in actual applications.

[0003] In the process of implementing the present invention, the inventors found that there are at least the following problems in the prior art:

[0004] Course learning is disconnected from actual operations. Users are unable to conduct situational simulation exercises immediately after learning the theory, resulting in insufficient in-depth application of knowledge.

[0005] Existing AI dialogue systems are mostly static and templated, cover few scenarios, cannot customize scenarios, and cannot adjust dialogue situations according to actual user needs.

[0006] It is impossible to provide users with a sufficiently personalized experience, and it is difficult to provide customized solutions for diverse areas such as parent-child communication and intimate relationships. Summary of the invention

[0007] In view of this, an embodiment of the present invention aims to provide a personalized learning method and system that combines course learning with AI training to solve at least one of the above technical problems.

[0008] To achieve the above-mentioned purpose, in a first aspect, an embodiment of the present invention provides a personalized learning method of course plus AI training, comprising the following steps executed by a computer program:

[0009] S1: Provide users with theoretical courses in at least one field;

[0010] S2: After the user completes the theoretical course, a personalized practice scenario is generated based on the content and learning status of the theoretical course studied by the user;

[0011] S3: In the personalized practice scenario, the AI ​​big model simulates a virtual character through role-playing, and performs interactive dialogue practice with the user based on the virtual character;

[0012] S4: Enable the AI ​​big model to provide feedback or suggestions based on the user’s performance in interactive dialogue exercises.

[0013] In some possible implementations, step S1 specifically includes:

[0014] S1.1: Obtain the user's learning preferences and learning style;

[0015] S1.2: Based on the user's learning preferences and learning style, select appropriate theoretical course content and presentation methods from a preset course library;

[0016] S1.3: During the theoretical course learning process, real-time monitoring of the user's learning behavior and emotional state;

[0017] S1.4: Dynamically adjust the content difficulty, presentation order and learning rhythm of the theoretical course based on the real-time monitoring results.

[0018] In some possible implementations, step S2 specifically includes:

[0019] S2.1: After confirming that the user has completed the theoretical course learning, the user's performance data during the theoretical course learning process is analyzed based on the machine learning algorithm to obtain quantitative indicators of the user's learning performance;

[0020] S2.2: Based on the quantitative indicators, determine the user's knowledge mastery and learning characteristics through a preset evaluation model;

[0021] S2.3: Using a content matching algorithm, based on the content of the theoretical course studied by the user, a basic scenario related to the content of the theoretical course studied by the user is selected from a preset scenario library;

[0022] S2.4: Adjust the difficulty of the selected basic scenario according to the user's knowledge level to obtain a practice scenario with appropriate difficulty;

[0023] S2.5: Based on the learning characteristics of the user, the difficulty-adapted practice scene is personalized to obtain a preliminary personalized practice scene;

[0024] S2.6: Match the preliminary personalized practice scenario with the content of the theoretical course learned by the user to obtain a practice scenario with matching knowledge points;

[0025] S2.7: Generate the final personalized practice scenario based on the practice scenario matched with the knowledge points.

[0026] In some possible implementations, step S3 specifically includes:

[0027] S3.1: Based on the personalized practice scenario, configure the corresponding role information, background knowledge and behavior pattern for the AI ​​big model;

[0028] S3.2: Control the AI ​​big model to generate virtual characters with different personalities and / or different emotional responses through preset prompt words or user-defined prompt words;

[0029] S3.3: Utilize a multimodal algorithm to parse and obtain the user's input information;

[0030] S3.4: generating a reply content of the virtual character based on a preset dialogue strategy, the user's input information, the character information, the background knowledge and the behavior pattern;

[0031] S3.5: identifying the emotional characteristics in the user's input information through a sentiment analysis algorithm, and reflecting the corresponding emotional characteristics in the reply content of the virtual character;

[0032] S3.6: When it is detected that the user has difficulty with a knowledge point in the theoretical course, the AI ​​big model is triggered to switch to explanation mode to provide explanations for the knowledge point.

[0033] In some possible implementations, step S3.3 includes:

[0034] Acquire user data from a variety of input devices, including a keyboard, a microphone, a camera, and a touch screen;

[0035] Perform modality-specific preprocessing and feature extraction on the collected data, which specifically includes: segmenting text data, removing stop words and part-of-speech tagging, and then using a pre-trained model to extract semantic features to obtain text semantic features; denoising and framing speech data, and then using a deep learning end-to-end ASR model for speech recognition to obtain speech-to-text results and acoustic features; performing image enhancement and size normalization on image data, and then using a convolutional neural network or visual converter to recognize facial expressions and body movements to obtain visual features and expression / action labels; performing signal filtering on tactile data, and then using a machine learning algorithm to analyze tactile patterns to obtain tactile features and tactile pattern labels;

[0036] Use a multimodal fusion model to integrate the features and labels of each modality, including text semantic features, speech-to-text results, acoustic features, visual features, expression / action labels, tactile features, and tactile pattern labels, to obtain a fused multimodal feature representation;

[0037] Based on the fused multimodal feature representation, user intent recognition and key entity extraction are performed, including: using conditional random fields or BiLSTM-CRF models to sequence the fused features to identify key entities; using pre-trained multi-label classification models to identify multiple user intents;

[0038] The processing results including the modal information input by the user, the identified multiple intents and the extracted key entities are converted into a standardized format. In some possible implementations, step S4 specifically includes:

[0039] S4.1: Real-time monitoring of users’ performance in interactive dialogue exercises, including answer accuracy, response speed, language expression ability, and emotional expression;

[0040] S4.2: Analyze the conversation content of the user in the interactive dialogue exercise using a natural language processing algorithm to obtain the user's ability assessment result;

[0041] S4.3: Based on the preset evaluation criteria and the ability evaluation results, the user's performance in the interactive dialogue exercise is quantitatively scored to obtain quantitative indicators of multiple dimensions of the user in the interactive dialogue exercise;

[0042] S4.4: Compare and analyze the quantitative indicators of multiple dimensions of the user in the interactive dialogue exercise with the quantitative indicators of the user's learning performance to obtain comparative analysis results;

[0043] S4.5: According to the comparative analysis results, select appropriate feedback content from a preset feedback template library;

[0044] S4.6: Use natural language generation algorithms to personalize the selected feedback content and generate personalized feedback or suggestions;

[0045] S4.7: Adjust the tone and expression of the personalized feedback or suggestions through sentiment analysis algorithms.

[0046] In some possible implementations, step S4.1 specifically includes:

[0047] Use speech recognition tools to convert users’ spoken responses into text in real time;

[0048] Use natural language understanding models to evaluate the accuracy and relevance of users’ spoken responses based on the converted text;

[0049] Record the time interval from when users receive questions to when they start to answer them to assess response speed;

[0050] Apply linguistic analysis tools to evaluate the user's vocabulary richness, syntactic complexity and expression fluency on the converted text;

[0051] The user's tone, volume, and speaking speed changes are analyzed through speech emotion recognition methods to evaluate emotional expression.

[0052] In some possible implementations, step S4.2 specifically includes:

[0053] Preprocessing the conversation content of the user in the interactive conversation exercise to obtain preprocessed conversation content, wherein the preprocessing includes word segmentation, stop word removal and standardization;

[0054] Using a natural language processing algorithm to extract features of the preprocessed conversation content, the features include word frequency statistics, grammatical structure analysis, and semantic vector representation;

[0055] The extracted features are analyzed using a pre-trained machine learning model to evaluate the user's performance in language fluency, vocabulary richness, grammatical accuracy, and semantic coherence, and to obtain the user's ability assessment results in various language ability dimensions.

[0056] In some possible implementations, step S4.3 specifically includes:

[0057] Establish a multi-dimensional scoring matrix and set weights for each dimension;

[0058] Map the results of step S4.1 and step S4.2 to the scoring criteria of each dimension;

[0059] The overall score was calculated using a weighted average algorithm;

[0060] Generate radar charts or other visualizations to intuitively display the user's performance in various dimensions.

[0061] In some possible implementations, step S4.4 specifically includes:

[0062] Establish a user ability assessment indicator system, including: interactive dialogue practice indicators, which include language fluency, vocabulary richness, grammatical accuracy and semantic coherence; learning performance indicators, which include test scores, homework completion, learning time and participation;

[0063] Each indicator is standardized to convert indicators of different dimensions into comparable standard scores;

[0064] The correlation coefficient matrix between the interactive dialogue practice indicators and the learning performance indicators was calculated to quantify the degree of association between the two sets of indicators;

[0065] Cluster analysis method is used to classify users into different ability groups based on interactive dialogue practice indicators and learning performance indicators;

[0066] Generate a comparative analysis report, including: descriptive statistical results of each indicator, including mean, standard deviation, maximum value, minimum value, etc.; correlation analysis results between interactive dialogue practice indicators and learning performance indicators, including correlation coefficient matrix; cluster analysis results, which show the characteristics and distribution of different ability groups; comparative charts of each ability group in interactive dialogue practice indicators and learning performance indicators.

[0067] In the second aspect, a personalized learning system with courses and AI training is provided, which includes:

[0068] Theoretical course module, used to provide users with theoretical courses in at least one field;

[0069] The practice scenario generation module is used to generate personalized practice scenarios based on the content and learning status of the theoretical courses learned by the users after the users complete the theoretical courses.

[0070] An AI training module is used to enable the AI ​​big model to simulate a virtual character through role-playing in the personalized training scenario, and to conduct interactive dialogue training with the user based on the virtual character;

[0071] The feedback generation module is used to enable the AI ​​model to provide feedback or suggestions based on the user's performance in the interactive dialogue exercise.

[0072] In some possible implementations, the theoretical course modules include:

[0073] A learning preference acquisition unit, used to acquire the user's learning preference and learning style;

[0074] A course selection unit, for selecting suitable theoretical course content and presentation method from a preset course library based on the user's learning preference and learning style;

[0075] The learning monitoring unit is used to monitor the user's learning behavior and emotional state in real time during the theoretical course learning process;

[0076] The course adjustment unit is used to dynamically adjust the content difficulty, presentation order and learning rhythm of the theoretical course according to the real-time monitoring results.

[0077] In some possible implementations, the practice scenario generation module includes:

[0078] A learning performance analysis unit is used to analyze the user's performance data during the theoretical course learning process based on a machine learning algorithm after confirming that the user has completed the theoretical course learning, and obtain a quantitative indicator of the user's learning performance;

[0079] A knowledge mastery evaluation unit, used to determine the user's knowledge mastery and learning characteristics through a preset evaluation model based on the quantitative indicators;

[0080] A scenario selection unit, for selecting a basic scenario related to the content of the theoretical course studied by the user from a preset scenario library by using a content matching algorithm based on the content of the theoretical course studied by the user;

[0081] A difficulty adjustment unit is used to adjust the difficulty of the selected basic scenario according to the user's knowledge level to obtain a practice scenario with suitable difficulty;

[0082] A personalized customization unit, used to customize the difficulty-adapted practice scene based on the user's learning characteristics to obtain a preliminary personalized practice scene;

[0083] A knowledge point matching unit is used to match the preliminary personalized practice scenario with the content of the theoretical course learned by the user to obtain a practice scenario with matching knowledge points;

[0084] The final scenario generation unit is used to generate a final personalized practice scenario based on the practice scenario matched with the knowledge points.

[0085] In some possible implementations, the AI ​​training module includes:

[0086] A role configuration unit, configured to configure corresponding role information, background knowledge and behavior patterns for the AI ​​big model based on the personalized practice scenario;

[0087] A virtual character generation unit, used to control the AI ​​big model to generate virtual characters with different personalities and / or different emotional reactions through preset prompt words or user-defined prompt words;

[0088] A user input parsing unit, used to parse and obtain user input information using a multimodal algorithm;

[0089] A reply generating unit, configured to generate reply content of the virtual character based on a preset dialogue strategy, user input information, the character information, the background knowledge and the behavior pattern;

[0090] The sentiment analysis unit is used to identify the sentiment characteristics in the user's input information through a sentiment analysis algorithm, and reflect the corresponding sentiment characteristics in the reply content of the virtual character;

[0091] The explanation mode switching unit is used to trigger the AI ​​big model to switch to the explanation mode and provide explanations for the knowledge points when it detects that the user has difficulties with the knowledge points of the theoretical course.

[0092] In some possible implementations, the feedback generation module includes:

[0093] Performance monitoring unit, used to monitor the user's performance in interactive dialogue exercises in real time, including answer accuracy, response speed, language expression ability and emotional expression;

[0094] A capability assessment unit, used to analyze the conversation content of the user in the interactive dialogue exercise using a natural language processing algorithm to obtain the user's capability assessment result;

[0095] A quantitative scoring unit, used to quantitatively score the user's performance in the interactive dialogue exercise based on a preset evaluation standard and the ability evaluation result, and obtain quantitative indicators of multiple dimensions of the user in the interactive dialogue exercise;

[0096] A comparative analysis unit, used to compare and analyze the quantitative indicators of multiple dimensions of the user in the interactive dialogue exercise with the quantitative indicators of the user's learning performance to obtain a comparative analysis result;

[0097] A feedback selection unit, configured to select appropriate feedback content from a preset feedback template library according to the comparison and analysis result;

[0098] A personalized feedback generation unit, used to personalize the selected feedback content and generate personalized feedback or suggestions by using a natural language generation algorithm;

[0099] The tone adjustment unit is used to adjust the tone and expression of the personalized feedback or suggestion through a sentiment analysis algorithm.

[0100] The above technical solution has the following beneficial effects:

[0101] By combining theoretical courses with AI training, this method can provide each user with a tailored learning experience. This personalized approach can better meet the needs of different learners and improve learning efficiency and effectiveness.

[0102] This method closely combines theoretical course learning with practical exercises, allowing users to consolidate and apply what they have learned through interactive dialogue exercises on the basis of mastering theoretical knowledge, thereby promoting in-depth understanding of knowledge and improvement of practical application ability.

[0103] Generate personalized practice scenarios based on the user's learning content and status to ensure that the practice content matches the user's learning progress and needs, avoiding the inadaptability problems that may be caused by traditional fixed practice content.

[0104] Interactive dialogue exercises using virtual characters simulated by large AI models provide users with a highly interactive and immersive learning environment, helping to increase learning interest and engagement.

[0105] The AI ​​big model can provide timely feedback or suggestions based on the user's performance in interactive dialogue exercises. This instant feedback mechanism helps users quickly identify and correct errors and continuously optimize learning strategies.

[0106] Through AI tutoring, this approach can reduce dependence on human teachers to a certain extent, potentially reducing education costs while providing readily available learning support.

[0107] The data generated during the entire learning process can be used to continuously optimize learning content, practice scenarios, and feedback mechanisms, thereby continuously improving the overall learning effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0108] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0109] Figure 1 It is a flow chart of a personalized learning method of course plus AI training in an embodiment of the present invention;

[0110] Figure 2 is a specific flow chart of step S1 of an embodiment of the present invention;

[0111] Figure 3 is a specific flow chart of step S2 of an embodiment of the present invention;

[0112] Figure 4 is a specific flow chart of step S3 of an embodiment of the present invention;

[0113] Figure 5 is a specific flow chart of step S4 of an embodiment of the present invention;

[0114] Figure 6 This is an overall functional block diagram of a personalized learning system with courses and AI training in an embodiment of the present invention;

[0115] Figure 7 It is a specific functional block diagram of a personalized learning system of a course plus AI training according to an embodiment of the present invention;

[0116] Figure 8 is a functional block diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0117] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0118] The technical problem solved by the embodiments of the present invention is how to provide users with an efficient platform that combines learning theory with practical training, especially in the fields of parent-child communication, intimate relationships, sleep management, workplace communication, etc. After learning the theoretical knowledge of relevant courses, users often lack an environment for real-time and personalized sparring. The embodiments of the present invention combine course learning with role-playing training of large AI models to help users better apply what they have learned in daily life and improve their communication skills and coping abilities.

[0119] The embodiment of the present invention aims to create a personalized learning program that integrates course learning and AI training, so that users can conduct highly simulated dialogue training through AI role-playing on the basis of learning relevant theoretical courses. Through the two-way interaction between courses and AI, this program helps users quickly apply what they have learned to practical situations and improve practical communication skills such as parent-child communication and workplace dialogue.

[0120] Embodiment 1

[0121] like Figure 1 As shown, an embodiment of the present invention provides a personalized learning method of course plus AI training, including the following steps executed by a computer program:

[0122] S1: Provide users with theoretical courses in at least one field.

[0123] In this step, the system provides users with theoretical courses in the fields of intimate relationships, workplace communication, and sleep management. For example, intimate relationship courses can include emotional communication, conflict resolution, trust building, etc.; workplace communication courses can include teamwork, presentation skills, negotiation strategies, etc.; sleep management courses can include topics such as sleep cycles, relaxation techniques, and sleep environment optimization; these theoretical courses are presented in a variety of forms such as text, audio, and video to suit the learning preferences of different users.

[0124] S2: After the user completes the theoretical course, a personalized practice scenario is generated based on the content and learning status of the theoretical course studied by the user.

[0125] After the user completes the theoretical course, the system analyzes the user's learning content and status. For example, if the system finds that the user's performance in the "Handling Difficult Conversations" section of the workplace communication course is poor, a personalized practice scenario simulating a difficult workplace conversation will be generated. This scenario can be a virtual team meeting that contains challenging communication situations that the user needs to handle.

[0126] S3: In personalized practice scenarios, the AI ​​big model simulates a virtual character through role-playing, and conducts interactive dialogue exercises with users based on the virtual character.

[0127] In the generated personalized practice scenarios, the AI ​​big model plays one or more virtual roles. For example, in an intimate relationship scenario, the AI ​​plays the role of the user's partner and simulates a conversation about a disagreement about family finances. The user needs to use the communication skills they have learned to deal with this sensitive topic and practice how to express their own ideas while also listening to and understanding their partner's views.

[0128] For example, in a workplace communication scenario, AI plays the role of a difficult colleague or client, making sharp criticisms in a team meeting. Users need to play the role of a project manager and use the skills learned in the course to respond to criticisms calmly and resolve conflicts.

[0129] For example, in the sleep management scenario, AI plays the role of a sleep expert and conducts a one-on-one sleep consultation session with the user. The user needs to describe his or her sleep problems, answer questions raised by AI, and try to develop a personalized sleep improvement plan based on AI's suggestions.

[0130] The above AI big model may include any one or more combinations of the following:

[0131] GPT (Generative Pre-trained Transformer) is a series of autoregressive language models developed by OpenAI that excels at natural language generation and understanding. This series of models is widely used in natural language processing tasks such as text generation, question answering, and summarization. Representative models include GPT-3 and GPT-4, which can generate coherent and natural language text and are suitable for a variety of language tasks.

[0132] BERT (Bidirectional Encoder Representations from Transformers) is a bidirectional transformer encoder model developed by Google. BERT enhances language representation through bidirectional contextual understanding, making it perform well in natural language understanding tasks such as question answering and sentiment analysis.

[0133] T5 (Text-to-Text Transfer Transformer) is a unified framework model developed by Google that treats all natural language processing tasks as text-to-text conversions. This highly general architecture enables T5 to adapt to a variety of tasks, such as translation, summarization, and question-answering, and performs well in a wide range of application scenarios.

[0134] LaMDA (Language Model for Dialogue Applications) is a language model optimized by Google for open-ended conversations. LaMDA is specially trained for natural conversations and can provide a more coherent, interactive and natural conversation experience, and can be applied in a variety of conversation scenarios.

[0135] PaLM (Pathways Language Model) is a large-scale language model developed by Google based on its Pathways system. PaLM can efficiently handle various language processing tasks.

[0136] LLaMA (Large Language Model Meta AI) is a large language model developed by Meta for the research community.

[0137] Claude is a large-scale language model developed by Anthropic, which emphasizes the safety and ethics of the model. Claude is able to perform complex language tasks while ensuring the compliance and safety of the generated content.

[0138] ERNIE (Enhanced Representation through kNowledgeIntEgration) is a knowledge-enhanced language understanding model developed by Baidu. By combining knowledge graphs, ERNIE has a stronger ability to understand entities and relationships and is suitable for a variety of natural language understanding tasks.

[0139] ChatGLM (Chat Generative Language Model) is an open source bilingual conversational language model developed by Tsinghua University. The model performs well in Chinese and English bilingual environments and can perform conversation and text generation tasks.

[0140] S4: Enable the AI ​​big model to provide feedback or suggestions based on the user’s performance in interactive dialogue exercises.

[0141] During and after the interactive dialogue exercise, the AI ​​big model analyzes the user's performance and provides personalized feedback and suggestions. For example: In an intimate relationship scenario, the AI ​​big model will point out the user's tone problem when expressing needs, and recommends using more "I think" statements instead of accusatory language. In a workplace communication scenario, the AI ​​big model will praise the user for successfully using active listening skills, but recommends that they can be more confident and firm when responding to criticism. In a sleep management scenario, the AI ​​big model will recommend that users record their sleep patterns in more detail and provide some specific relaxation techniques to improve the problem of difficulty falling asleep.

[0142] In this way, users can not only learn theoretical knowledge, but also practice and apply what they have learned in close-to-real scenarios, while receiving timely and targeted feedback. This method of combining theoretical learning, practical exercises and intelligent feedback can provide users with an efficient and personalized learning experience, helping users make substantial progress in important life areas such as intimate relationships, workplace communication and sleep management.

[0143] like Figure 2 As shown, in some embodiments, step S1 (providing a user with a theoretical course in at least one field) can be further subdivided into the following steps:

[0144] In step S1.1, the user's learning preferences and learning style are obtained. The system learns about the user's learning habits and needs in a variety of ways, which may include asking the user to complete a detailed learning style questionnaire to assess whether they are visual, auditory, or hands-on learners; analyzing the user's historical learning data, such as the completion of past courses, video viewing habits, and the frequency of participating in interactive exercises; and allowing users to directly select their preferred learning methods, such as whether they prefer video explanations, text materials, or interactive exercises. For example, the system finds that a user is a visual learner who prefers to learn through diagrams and videos and tends to study at night.

[0145] Step S1.2 selects appropriate theoretical course content and presentation methods from the preset course library based on the user's learning preferences and learning style. Based on the preference information obtained in S1.1, the system will prioritize the content and presentation methods that are most suitable for the user from the course library. Continuing with the above example, for visual learners, the system will prioritize course content that contains a large number of charts, infographics, and short videos, and even convert text-intensive content into more graphical presentations, such as flowcharts or mind maps. At the same time, if the user tends to study at night, the system can also adjust the interface to dark mode to reduce the user's visual fatigue.

[0146] In step S1.3, the system monitors the user's learning behavior and emotional state in real time during the learning process of the theoretical course. This includes tracking the user's learning time, pause frequency, number of times the content is repeatedly watched, and other behavioral indicators. It can even analyze the user's facial expressions through the camera to determine whether the user is confused, tired, or showing interest. At the same time, the system will also record the user's interactive behavior, such as clicking, sliding, or taking notes. For example, the system can find that when the user is learning the content of "efficient meeting management", he frequently pauses and repeatedly watches certain clips, and his facial expression shows signs of confusion.

[0147] Step S1.4 dynamically adjusts the content difficulty, presentation order and learning rhythm of the theoretical course based on the real-time monitoring results. Based on the monitoring data in S1.3, the system can make adaptive adjustments during the learning process. For example, if the system finds that the user is confused in a certain content module, it can insert basic explanations or examples to assist understanding; it can also adjust the presentation order of subsequent content based on the user's response, and bring some relatively simple concepts forward to help users gradually build understanding and confidence. If it is detected that the user's attention is beginning to disperse, the system can prompt the user to take a break, or switch to interactive exercises to refocus.

[0148] In this way, the system not only customizes the learning content according to the user's learning preferences before the course starts, but also makes real-time adjustments during the learning process to dynamically adapt to the user's immediate needs and status. This highly personalized and flexible adjustment mechanism can significantly improve the user's learning efficiency and experience. For example, for users who encounter difficulties in the "Efficient Meeting Management" content module, the system can break down complex techniques and convert them into small, easy-to-understand portions of knowledge by inserting a short graphical meeting overview; it can also provide interactive scenario simulations when users become fatigued to practice and consolidate what they have learned. Through these real-time adjustments, the system ensures that each user can get the learning experience that best suits them, thereby maximizing their learning effects and improving overall learning efficiency.

[0149] like Figure 3 As shown, in some embodiments, step S2 (generating a personalized practice scenario based on the content and learning status of the theoretical course learned by the user) can be further subdivided into the following steps:

[0150] S2.1: After confirming that the user has completed the theoretical course, the system analyzes the user's performance data during the theoretical course based on a machine learning algorithm to obtain quantitative indicators of the user's learning performance. These data may include the user's study time, test scores, interaction frequency, etc. For example, for users who have completed the "Parent-Child Communication" course, the system may analyze their participation in the course (such as video viewing completion rate), performance in situational tests, completion of interactive exercises, etc. Through these data, the system can derive quantitative indicators such as "85% mastery of listening skills", "70% emotional management ability", and "75% positive expression ability".

[0151] Specifically, machine learning algorithms include: supervised learning algorithms, which predict or classify through existing data, such as linear regression for predicting learning performance, decision trees and random forests for classifying user learning levels, and support vector machines (SVM) for binary classification tasks to complete predictions. Unsupervised learning algorithms, which are used for pattern recognition and dimensionality reduction, such as K-means clustering to group users to identify different learning patterns, and principal component analysis (PCA) to extract key learning features after dimensionality reduction. Deep learning algorithms, such as recurrent neural networks (RNN) for analyzing learning behaviors of time series, and convolutional neural networks (CNN) for analyzing user interaction data and engagement. Reinforcement learning algorithms, such as Q-learning, optimize personalized learning paths and recommend optimal learning actions through continuous learning. Ensemble learning methods, such as gradient boosting decision trees XGBoost, comprehensively analyze multiple factors, accurately predict learning effects and rank the importance of influencing factors. These algorithms can be used independently or in combination to more comprehensively and accurately analyze user learning performance and provide personalized feedback.

[0152] S2.2: Based on the quantitative indicators obtained in S2.1, the system determines the user's knowledge mastery and learning characteristics through a preset evaluation model. The evaluation model can be an algorithm that takes into account multiple factors. It not only evaluates the user's overall performance, but also identifies the user's strengths and weaknesses. For example, for a user studying the "Intimate Relationships" course, the system can determine that the user performs well in understanding the needs of his partner, but needs more practice in dealing with conflicts and expressing his own needs. At the same time, the system can find that the user performs well in daily warm interaction scenarios, but performs poorly in communication under stressful situations.

[0153] The preset assessment model uses a multi-dimensional framework that covers multiple key aspects of the course content. These dimensions are set according to the nature of the course and the learning objectives, and usually include knowledge understanding, skill application, situational adaptation and other aspects. Each dimension uses a unified scoring standard to ensure that the assessment of different learners is consistent and comparable. The preset assessment model assigns specific weights to each assessment dimension to reflect its relative importance in the overall learning objectives. The weights are assigned based on the focus of the course design to ensure that the assessment results can accurately reflect the most critical learning outcomes. The system uses a weighted average method to calculate the user's overall score. This method combines the scores of each dimension with the corresponding weights to derive a comprehensive learning performance indicator to help better evaluate the user's overall learning situation. Based on the comprehensive score, the preset assessment model divides the user's performance into multiple levels. This grading can quickly assess the learner's overall level, that is, the degree of knowledge mastery. The system compares the user's scores in each dimension through the strength and weakness identification algorithm to identify the user's strengths and weaknesses. This algorithm not only focuses on the user's absolute score, but also analyzes his relative performance between different dimensions to provide a basis for personalized learning recommendations. The strength and weakness identification algorithm helps determine the user's knowledge mastery and learning characteristics in different aspects. The preset evaluation model further analyzes the differences in user performance in different question types and learning scenarios. This analysis can identify the characteristics of users in terms of theoretical and practical abilities, performance in different situations, etc., which helps to adjust learning strategies and optimize learning experience.

[0154] S2.3: The system uses a content matching algorithm to select basic scenarios related to the content of the theoretical courses that users are learning from the preset scenario library based on the content of the theoretical courses that users are learning. For example, for users who have learned the "Workplace Communication" course, the system can select basic scenarios related to "discussing salary increases with superiors", "dealing with conflicts between colleagues", "hosting team meetings", "giving feedback to subordinates", etc. from the scenario library. These scenarios directly correspond to the workplace communication skills and principles learned in the course.

[0155] Specifically, the content matching algorithm is a technology used to match the theoretical course content that users are learning with the basic scenarios in the preset scenario library. First, the content matching algorithm extracts keywords and topics from the course content that users are learning, and then performs semantic analysis on these keywords to understand their importance in the context. Then, the content matching algorithm compares these extracted keywords with the labels of each scenario in the scenario library, and calculates the relevance score of each scenario based on the matching degree and semantic similarity. Based on the score, the content matching algorithm sorts the scenarios and selects the ones that are most relevant to the course content. In this process, the content matching algorithm also takes into account the user's learning progress and difficulty level to ensure that the selected scenarios are both suitable for the current learning stage and can effectively help users consolidate their knowledge. In addition, the content matching algorithm also ensures that different aspects of the course content are covered by ensuring the diversity of scenarios, ultimately providing users with a comprehensive and personalized learning experience.

[0156] S2.4: Based on the user's knowledge level, the system adjusts the difficulty of the basic scenario selected in S2.3 to obtain a practice scenario with appropriate difficulty. Considering that learners of the "Sleep Management" course need more practice in establishing a sleep routine, the system will design a series of scenarios with increasing difficulty, from the simple "making a sleep schedule" to the more complex "maintaining good sleep habits while traveling", gradually improving the user's ability to cope with various sleep challenges.

[0157] S2.5: Based on the user's learning characteristics, the system customizes the difficulty-adapted practice scenarios in S2.4 to obtain preliminary personalized practice scenarios. For example, for learners of the "Parent-Child Communication" course, if the system finds that the user has particular challenges in dealing with the emotional problems of their teenage children, a scenario can be designed to simulate a situation where a teenager is depressed due to academic pressure, allowing the user to practice how to listen to, understand, and appropriately respond to their children's emotional needs.

[0158] S2.6: The system matches the preliminary personalized practice scenarios in S2.5 with the content of the theoretical course that the user is learning, and obtains practice scenarios with matching knowledge points. In this step, the system ensures that each practice scenario clearly corresponds to a specific knowledge point in the course. For example, in the "Handling Disagreements" scenario of the "Intimate Relationships" course, the system can specifically emphasize the application of active listening skills and how to find common goals in disagreements, which are key knowledge points in the course.

[0159] S2.7: Based on the practice scenarios matched with the knowledge points in S2.6, the system generates the final personalized practice scenario. In this final step, the system adds some details and background information to make the scenario more realistic and vivid. For example, in the "Host a Team Meeting" scenario of the "Workplace Communication" course, the system creates a detailed background profile of a virtual team, including the personality traits, work styles, potential team conflicts, etc. of the team members, so that users can apply the learned meeting hosting skills in a more realistic environment. For the "Sleep Management" course, the system can design a sleep adjustment scenario involving shift work based on the user's work and rest habits and the nature of their work, allowing users to apply the circadian rhythm adjustment skills learned in the course in practice. The final personalized practice scenario generated in this way is not only of moderate difficulty and targeted at the user's learning needs, but also ensures that users can fully apply the knowledge points learned in the course in practice, whether in parent-child relationships, intimate relationships, workplace communication or personal health management.

[0160] like Figure 4 As shown, in some embodiments, step S3 (generating a virtual character based on a personalized practice scenario using an AI big model and interacting with the user in a dialogue) can be further subdivided into the following steps:

[0161] S3.1: Based on personalized practice scenarios, the system configures the corresponding role information, background knowledge, and behavior patterns for the AI ​​large model. For example, in the parent-child communication scenario, the system can configure the role information of a 14-year-old teenager for the AI, including characteristics such as the character is in a rebellious period, feels anxious about learning pressure, and likes to play games. In the workplace communication scenario, AI can be configured as an experienced department manager who has rich project management experience but is not good at handling team conflicts. For the sleep management scenario, AI can be set as an office worker who suffers from insomnia due to work pressure, has the habit of staying up late to work, and is dependent on coffee.

[0162] S3.2: Through preset prompt words or user-defined prompt words, the system controls the AI ​​large model to generate virtual characters with different personalities and / or different emotional responses. In intimate relationship scenarios, the system can use "You are an introverted partner who is not good at expressing emotions" as a prompt word to let AI generate a more passive role in communication. For workplace communication, you can use "You are an ambitious but impatient new employee" to create a challenging communication partner. In parent-child communication, the system can let AI play the role of children with different personalities such as "angry and stubborn teenagers" or "sensitive and melancholy elementary school students" according to the user's choice.

[0163] S3.3: The system uses a multimodal algorithm to parse the user's input information. Using a multimodal algorithm to parse the user's data in various modes as input, the user's input information can be more comprehensively parsed, including data in various forms such as text, voice, and images. For example, in a scenario about sleep management, the user can input by voice: "I can't sleep every night, and I can't help but look at my phone until late at night." The system will analyze the semantic structure of this sentence and identify key information such as "can't sleep", "look at the phone", "very late", etc., to prepare for subsequent responses.

[0164] Specifically, the system obtains user data from multiple input devices, including keyboards to obtain text data, microphones to obtain voice data, cameras to obtain image data, and touch screens to obtain tactile data. This multimodal data collection method can fully capture user interaction information and provide a rich input source for subsequent processing.

[0165] Next, the system performs specific preprocessing and feature extraction on the data of each modality. For text data, it performs word segmentation, stop word removal, part-of-speech tagging, and semantic feature extraction; for speech data, it performs noise reduction, framing, speech recognition, and acoustic feature extraction; for image data, it performs image enhancement, size normalization, facial expression and body movement recognition, and visual feature extraction; for tactile data, it performs signal filtering, tactile pattern analysis, and tactile feature extraction. These processing steps ensure the quality and consistency of data from each modality, laying the foundation for subsequent fusion and analysis.

[0166] After obtaining the features and labels of each modality, the system uses a multimodal fusion model for integration. This embodiment uses a Transformer model based on an attention mechanism to encode the features and labels of each modality into vectors, uses a multi-head attention mechanism to calculate the association between different modalities, and performs feature fusion through a feedforward neural network and residual connections to finally obtain a fused multimodal feature representation. This fusion method can effectively capture the relationship between different modalities and improve the system's ability to understand.

[0167] Based on the fused multimodal feature representation, the system performs user intent recognition and key entity extraction. Key entity extraction uses the BiLSTM-CRF model to sequence the fused features and identify key entities in the text; multiple intent recognition uses a pre-trained multi-label classification model, such as TextCNN or BERT, to identify the user's multiple intents and output possible multiple intents and their confidence levels. This step enables the system to deeply understand the user's needs and intentions.

[0168] Finally, the system converts the processing results into a standardized JSON format, including the modal information input by the user, the multiple intents identified and their confidence, the key entities extracted and their types. Through this comprehensive multimodal processing flow, the system can accurately understand and respond to the user's complex interaction needs. S3.4: Based on the preset dialogue strategy, the user's input information, role information, background knowledge and behavior patterns, the system generates the virtual character's reply content. In the parent-child communication scenario, if the user (playing the role of a parent) says: "Your grades have dropped recently. Are you playing too many games?" The system will let the virtual character (playing the role of a teenager) reply: "You are always like this. You only care about my grades and never ask me if I have any pressure." This reply reflects the rebellious characteristics of the adolescent character and the anxiety about learning pressure.

[0169] S3.5: The system uses sentiment analysis algorithms to identify the emotional characteristics of the user's input information and reflects the corresponding emotional characteristics in the virtual character's reply content. For example, in an intimate relationship scenario, if the system detects disappointment and frustration in the user's tone, the partner role played by AI may respond: "I feel like you are a little disappointed. I may not be very good at expressing myself, but I really care about your feelings. Can you tell me more?" Such a response not only reflects the character's introverted characteristics, but also shows sensitivity to the partner's emotions.

[0170] Specifically, this step specifically includes the following steps: obtaining the user's input information; using a pre-trained sentiment analysis algorithm to process the user's input information to identify the sentiment characteristics therein; matching the identified sentiment characteristics with predefined sentiment categories to determine the user's current sentiment state; selecting a corresponding reply framework from a preset reply template library according to the determined user's current sentiment state; personalizing the selected reply framework based on the predefined personality characteristics of the virtual character to obtain a personalized reply framework; combining the personalized reply framework with the user's current sentiment characteristics to generate the final reply content; and providing the generated final reply content to the user. The sentiment analysis algorithm includes but is not limited to: a sentiment analysis algorithm based on a dictionary, a sentiment analysis algorithm based on a machine learning foundation, and a sentiment analysis algorithm based on a deep learning model. The predefined sentiment categories include at least: happiness, sadness, anger, fear, surprise, and disgust. The predefined personality characteristics of the virtual character include but are not limited to: introversion, extroversion, sensitivity, rationality, humor, and seriousness.

[0171] S3.6: When the system detects that the user has difficulty with the knowledge points of the theoretical course, it will trigger the AI ​​model to switch to explanation mode and provide explanations for the knowledge points. For example, in a workplace communication scenario, if the user shows hesitation or inappropriate handling when dealing with team conflicts, the system will pause the role-playing and switch to explanation mode: "It seems that you have encountered some challenges in dealing with disagreements between team members. Let's review several key points about conflict management in the course: first, remain neutral and objective; second, encourage all parties to express their views; finally, guide everyone to focus on common goals rather than personal differences." In the sleep management scenario, if the user shows confusion about establishing a sleep routine, the system can explain: "Establishing healthy sleep habits requires a step-by-step process. We can start by setting a fixed bedtime, and then gradually establish a relaxing bedtime ritual, such as meditation or light stretching. Remember, avoid using electronic devices before bedtime, because blue light affects the secretion of melatonin."

[0172] like Figure 5 As shown, in some embodiments, step S4 (generating personalized feedback or suggestions based on the user's performance in the interactive dialogue exercise) can be further subdivided into the following steps:

[0173] S4.1: Real-time monitoring of the user's performance in interactive dialogue exercises, including accuracy of answers, reaction speed, language expression ability, and emotional expression. For example, in a parent-child communication scenario, the system will record the speed at which the user (playing the role of a parent) responds to the child's questions to evaluate whether he or she can quickly understand and appropriately respond to the child's emotional needs. The system will also analyze the user's language choices, such as whether empathetic statements such as "I understand how you feel" are used. In workplace communication scenarios, the system will pay attention to the user's calmness when dealing with difficult problems, and whether they can express their views clearly and professionally. For sleep management scenarios, the system will evaluate the user's specificity and logic when describing their sleep problems and formulating improvement plans.

[0174] S4.2: Use natural language processing algorithms to analyze the content of users' conversations in interactive dialogue exercises and obtain the user's ability assessment results. In intimate relationship scenarios, the algorithm will identify whether the user uses active listening skills, such as repeating the partner's words to confirm understanding. The system can also analyze whether the user can express his or her needs appropriately while also taking into account the feelings of his or her partner. For example, if the user says, "I understand that you are busy at work, but I also need us to have more time together. Can we find a time to discuss how to balance work and relationships?" The system will identify this expression as a positive communication skill.

[0175] S4.3: Based on the preset evaluation criteria and ability evaluation results, the system will quantify the user's performance in the interactive dialogue exercise and obtain quantitative indicators of multiple dimensions of the user in the interactive dialogue exercise. For example, in the workplace communication scenario, the system will evaluate the following dimensions: clear expression (1-10 points), active listening (1-10 points), conflict resolution (1-10 points), emotional management (1-10 points) and professionalism (1-10 points). If the user performs well in a simulated team conflict resolution exercise, the system may give the following scores: 8 points for clear expression, 9 points for active listening, 7 points for conflict resolution, 8 points for emotional management, and 9 points for professionalism.

[0176] S4.4: Compare and analyze the quantitative indicators of multiple dimensions of the user in the interactive dialogue exercise with the quantitative indicators of the user's learning performance to obtain comparative analysis results. For example, in the parent-child communication scenario, if the user scored 85% in "Listening Skills" in the theoretical course, but only scored 7 points (equivalent to 70%) in "Active Listening" in the actual dialogue exercise, the system will identify this gap. Similarly, if the user scored high in the theoretical knowledge test of "Emotional Management" but scored low in "Emotional Management" in practice, the system will also notice this.

[0177] S4.5: Based on the comparative analysis results, select appropriate feedback content from the preset feedback template library. For example, for the sleep management scenario, if the system finds that the user performs well in understanding the theory of sleep cycles, but encounters difficulties in actually making a sleep plan, it may select a feedback template like this: "You have a very good understanding of the sleep cycle, which lays a theoretical foundation for improving sleep quality. However, in actual application, we found that you still have room for improvement in XX specific issues. Here are some suggestions to help you better transform theoretical knowledge into practice."

[0178] S4.6: Use natural language generation algorithms to personalize the selected feedback content and generate personalized feedback or suggestions. In intimate relationship scenarios, if the system finds that the user is too direct in expressing his or her needs, it may generate personalized feedback such as: "I noticed that you are very direct in expressing your needs, and this kind of frankness is good. However, considering that your partner has a more sensitive personality, you can try to use a softer way of expression.

[0179] S4.7: Adjust the tone and expression of personalized feedback or suggestions through sentiment analysis algorithms. For example, in a workplace communication scenario, if the system detects that the user is showing frustration during the exercise, the tone of the feedback can be adjusted to make it more encouraging and supportive. The feedback can start like this: "First, I want to affirm your efforts in this exercise. Dealing with workplace conflicts is indeed challenging, and you have shown a good willingness to communicate", and then gradually introduce specific suggestions for improvement. For parent-child communication scenarios, if the system finds that the user is impatient when talking to a simulated teenager, the feedback can be reminded in a gentle but firm tone: "I understand that communicating with teenagers can be frustrating. Remember that being patient and calm is essential to building a good parent-child relationship. Let's explore some strategies that can help you stay calm when you are frustrated."

[0180] The embodiment of the present invention forms a complete learning and practice platform by combining online course learning with AI big model training. During the course learning stage, users purchase and learn theoretical courses in the fields of parent-child communication, intimate relationships, workplace communication, etc. through the platform. The course content covers communication skills, emotional expression, problem solving and other knowledge. In the AI ​​training stage, users enter the AI ​​training module after completing the course learning. The AI ​​training module will provide personalized practice scenarios. The AI ​​big model simulates virtual characters that meet user needs (such as children, partners, colleagues, etc.) through role-playing, and provides highly simulated scenarios for users to practice.

[0181] The AI ​​big model generates characters with different personalities and emotional responses through system-existing or user-defined prompt words. For example, in a parent-child communication scenario, the AI ​​big model can play the role of a rebellious child, and users can interact with the AI ​​character through text or voice input to practice dealing with tricky communication situations. The AI ​​sparring module not only helps users practice through simulated conversations, but also provides feedback and suggestions based on the user's performance, guides users to improve their communication skills, and makes learning and practice complement each other, promoting the internalization of knowledge and improving capabilities. This solution is not only suitable for parent-child communication, but can also be extended to intimate relationships, workplace communication, sleep management and other fields. The AI ​​big model can automatically adjust the role-playing scenarios and dialogue modes according to the course content in different fields.

[0182] In this embodiment, the course learning and AI training are seamlessly connected, and users can immediately perform high-simulation exercises through the AI ​​big model after learning the course knowledge. The personalized role-playing technology of the AI ​​big model can simulate a variety of personalities, a variety of emotional states, and a variety of communication scenarios, enhancing the realism and sense of substitution of the user's practice. The system can dynamically generate exercise content according to the user's progress and performance in the course learning to ensure that theory and practice are closely integrated. The personalized role-playing ability of the AI ​​character is realized through the design of prompt words, especially the generation of different scenarios and dialogue reactions according to the course content and user needs. This embodiment has a two-way interactive mechanism, and AI can provide feedback and suggestions to users, so that they can continuously improve their communication skills during the practice process. The arrangement of real-person training is complicated and it is difficult to ensure the frequency, and the existing AI system lacks sufficient flexibility and personalization. By combining course learning and AI training, the embodiment of the present invention not only ensures the immediacy of practice, but also provides a highly personalized interactive experience.

[0183] By combining theoretical learning with practical operations, the embodiments of the present invention solve the problem of "learning without practice" in the prior art and give users the ability to practice immediately. Compared with a single course learning or a templated AI dialogue system, the present invention has the following advantages: after completing the course learning, the user can practice through AI immediately, which shortens the distance between learning and practical application, and helps users quickly master and apply new skills. The AI ​​big model can generate customized characters based on the scenes and roles owned by the system or set by the user, so that each practice of the user is targeted and realistic. AI training can provide feedback based on the user's performance, helping the user to continuously improve, thereby improving the learning effect. The embodiments of the present invention are not limited to the field of parent-child communication, but can also be extended to various situations such as intimate relationships and workplace communication, and have a wide range of applications.

[0184] Embodiment 2

[0185] like Figure 6 As shown, this embodiment provides a personalized learning system 200 for courses plus AI training, which includes:

[0186] Theoretical course module 210, used to provide theoretical courses in at least one field to users;

[0187] The practice scenario generation module 220 is used to generate a personalized practice scenario based on the content and learning status of the theoretical course studied by the user after the user completes the theoretical course study;

[0188] The AI ​​training module 230 is used to enable the AI ​​big model to simulate a virtual character through role-playing in a personalized training scenario, and to conduct interactive dialogue training with the user based on the virtual character;

[0189] The feedback generation module 240 is used to enable the AI ​​big model to provide feedback or suggestions based on the user's performance in the interactive dialogue exercise.

[0190] like Figure 7 As shown, in some embodiments, the theory course module 210 includes:

[0191] A learning preference acquisition unit, used to acquire the user's learning preference and learning style;

[0192] A course selection unit is used to select appropriate theoretical course content and presentation methods from a preset course library based on the user's learning preferences and learning style;

[0193] The learning monitoring unit is used to monitor the user's learning behavior and emotional state in real time during the theoretical course learning process;

[0194] The course adjustment unit is used to dynamically adjust the content difficulty, presentation order and learning rhythm of theoretical courses based on real-time monitoring results.

[0195] In some embodiments, the practice scenario generation module 220 includes:

[0196] A learning performance analysis unit is used to analyze the user's performance data during the theoretical course learning process based on a machine learning algorithm after confirming that the user has completed the theoretical course learning, and obtain a quantitative indicator of the user's learning performance;

[0197] A knowledge mastery evaluation unit is used to determine the user's knowledge mastery and learning characteristics based on quantitative indicators and through a preset evaluation model;

[0198] A scenario selection unit, for selecting a basic scenario related to the content of the theoretical course studied by the user from a preset scenario library by using a content matching algorithm based on the content of the theoretical course studied by the user;

[0199] A difficulty adjustment unit is used to adjust the difficulty of the selected basic scenario according to the user's knowledge level to obtain a practice scenario with suitable difficulty;

[0200] A personalized customization unit is used to customize the difficulty-adapted practice scene based on the user's learning characteristics to obtain a preliminary personalized practice scene;

[0201] A knowledge point matching unit is used to match the preliminary personalized practice scenario with the content of the theoretical course learned by the user to obtain a practice scenario with matching knowledge points;

[0202] The final scenario generation unit is used to generate a final personalized practice scenario based on the practice scenario matched with the knowledge points.

[0203] In some embodiments, the AI ​​training module 230 includes:

[0204] The role configuration unit is used to configure the corresponding role information, background knowledge and behavior patterns for the AI ​​big model based on personalized practice scenarios;

[0205] A virtual character generation unit, used to control the AI ​​big model to generate virtual characters with different personalities and / or different emotional responses through preset prompt words or user-defined prompt words;

[0206] A user input parsing unit, used to parse and obtain user input information using a multimodal algorithm;

[0207] A reply generation unit, used to generate reply content of the virtual character based on a preset dialogue strategy, user input information, character information, background knowledge and behavior pattern;

[0208] The sentiment analysis unit is used to identify the sentiment characteristics in the user's input information through a sentiment analysis algorithm, and reflect the corresponding sentiment characteristics in the reply content of the virtual character;

[0209] The explanation mode switching unit is used to trigger the AI ​​big model to switch to the explanation mode and provide explanations for the knowledge points when it detects that the user has difficulties with the knowledge points of the theoretical course.

[0210] In some embodiments, the feedback generation module 240 includes:

[0211] Performance monitoring unit, used to monitor the user's performance in interactive dialogue exercises in real time, including answer accuracy, response speed, language expression ability and emotional expression;

[0212] A capability assessment unit, used to analyze the conversation content of the user in the interactive dialogue exercise using a natural language processing algorithm to obtain the user's capability assessment result;

[0213] A quantitative scoring unit is used to quantify the performance of the user in the interactive dialogue exercise based on the preset evaluation criteria and ability evaluation results, and obtain quantitative indicators of multiple dimensions of the user in the interactive dialogue exercise;

[0214] A comparative analysis unit, used to compare and analyze the quantitative indicators of multiple dimensions of the user in the interactive dialogue exercise with the quantitative indicators of the user's learning performance to obtain a comparative analysis result;

[0215] A feedback selection unit, used to select appropriate feedback content from a preset feedback template library according to the comparative analysis results;

[0216] A personalized feedback generation unit, used to personalize the selected feedback content and generate personalized feedback or suggestions by using a natural language generation algorithm;

[0217] The tone adjustment unit is used to adjust the tone and expression of personalized feedback or suggestions through sentiment analysis algorithms.

[0218] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0219] Embodiment 3

[0220] An embodiment of the present invention further provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, each step of the above method is implemented.

[0221] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. Of course, there are other ways of readable storage media, such as quantum memory, graphene memory, etc. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media does not include electrical carrier signals and telecommunication signals.

[0222] Embodiment 4

[0223] The embodiment of the present invention further provides an electronic device, such as Figure 3 As shown, it includes one or more processors 301, a communication interface 302, a memory 303 and a communication bus 304, wherein the processor 301, the communication interface 302 and the memory 303 communicate with each other through the communication bus 304. The memory 303 is used to store computer programs; the processor 301 is used to execute the programs stored in the memory 303 to implement the steps of the above method.

[0224] The processor 301 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0225] The memory 303 may include a large capacity memory for data or instructions. For example, but not limitation, the memory 303 may include a hard disk drive (HDD), a floppy disk drive, a 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. In appropriate cases, the memory 303 may include a removable or non-removable (or fixed) medium. In a specific embodiment, the memory 303 is a non-volatile solid-state memory. In a specific embodiment, the memory 303 includes a read-only memory (ROM). In appropriate cases, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM) or a flash memory or a combination of two or more of these.

[0226] Communication bus 304 includes hardware, software or both, for coupling the above-mentioned components to each other.For example, bus may include accelerated graphics port (AGP) or other graphics bus, enhanced industrial standard architecture (EISA) bus, front side bus (FSB), hypertransport (HT) interconnection, industrial standard architecture (ISA) bus, infinite bandwidth interconnection, low pin count (LPC) bus, memory bus, micro channel architecture (MCA) bus, peripheral component interconnection (PCI) bus, PCI-Express (PCI-X) bus, serial advanced technology attachment (SATA) bus, video electronics standard association local (VLB) bus or other suitable bus or two or more of these combinations.In suitable cases, bus may include one or more buses.Although the embodiment of the present invention describes and shows a specific bus, the present invention considers any suitable bus or interconnection.

[0227] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0228] Although the present application provides method operation steps such as embodiments or flow charts, more or fewer operation steps may be included based on conventional or non-creative labor. The order of steps listed in the embodiments is only one way of executing the order of many steps and does not represent the only execution order. When the actual device or client product is executed, it can be executed in the order of the method shown in the embodiments or the drawings or in parallel (for example, in a parallel processor or multi-threaded processing environment).

[0229] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0230] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0231] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0232] The present invention uses specific embodiments to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A personalized learning method of course plus AI training, characterized in that: The method comprises the following steps executed by a computer program: S1: Provide users with theoretical courses in at least one field; S2: After the user completes the theoretical course, a personalized practice scenario is generated based on the content and learning status of the theoretical course studied by the user; S3: In the personalized practice scenario, the AI ​​big model simulates a virtual character through role-playing, and performs interactive dialogue practice with the user based on the virtual character; S4: Enable the AI ​​big model to provide feedback or suggestions based on the user’s performance in interactive dialogue exercises.

2. The method according to claim 1, characterized in that Step S1 specifically includes: S1.1: Obtain the user's learning preferences and learning style; S1.2: Based on the user's learning preferences and learning style, select appropriate theoretical course content and presentation methods from a preset course library; S1.3: During the theoretical course learning process, real-time monitoring of the user's learning behavior and emotional state; S1.4: Dynamically adjust the content difficulty, presentation order and learning rhythm of the theoretical course based on the real-time monitoring results.

3. The method according to claim 1, characterized in that Step S2 specifically includes: S2.1: After confirming that the user has completed the theoretical course learning, the user's performance data during the theoretical course learning process is analyzed based on the machine learning algorithm to obtain quantitative indicators of the user's learning performance; S2.2: Based on the quantitative indicators, determine the user's knowledge mastery and learning characteristics through a preset evaluation model; S2.3: Using a content matching algorithm, based on the content of the theoretical course studied by the user, a basic scenario related to the content of the theoretical course studied by the user is selected from a preset scenario library; S2.4: Adjust the difficulty of the selected basic scenario according to the user's knowledge level to obtain a practice scenario with appropriate difficulty; S2.5: Based on the learning characteristics of the user, the difficulty-adapted practice scene is personalized to obtain a preliminary personalized practice scene; S2.6: Match the preliminary personalized practice scenario with the content of the theoretical course learned by the user to obtain a practice scenario with matching knowledge points; S2.7: Generate the final personalized practice scenario based on the practice scenario matched with the knowledge points.

4. The method according to claim 1, characterized in that: Step S3 specifically includes: S3.1: Based on the personalized practice scenario, configure the corresponding role information, background knowledge and behavior pattern for the AI ​​big model; S3.2: Control the AI ​​big model to generate virtual characters with different personalities and / or different emotional responses through preset prompt words or user-defined prompt words; S3.3: Utilize a multimodal algorithm to parse and obtain the user's input information; S3.4: generating a reply content of the virtual character based on a preset dialogue strategy, the user's input information, the character information, the background knowledge and the behavior pattern; S3.5: identifying the emotional characteristics in the user's input information through a sentiment analysis algorithm, and reflecting the corresponding emotional characteristics in the reply content of the virtual character; S3.6: When it is detected that the user has difficulty with a knowledge point in the theoretical course, the AI ​​big model is triggered to switch to explanation mode to provide explanations for the knowledge point.

5. The method according to claim 3, characterized in that: Step S4 specifically includes: S4.1: Real-time monitoring of users’ performance in interactive dialogue exercises, including answer accuracy, response speed, language expression ability, and emotional expression; S4.2: Analyze the conversation content of the user in the interactive dialogue exercise using a natural language processing algorithm to obtain the user's ability assessment result; S4.3: Based on the preset evaluation criteria and the ability evaluation results, the user's performance in the interactive dialogue exercise is quantitatively scored to obtain quantitative indicators of multiple dimensions of the user in the interactive dialogue exercise; S4.4: Compare and analyze the quantitative indicators of multiple dimensions of the user in the interactive dialogue exercise with the quantitative indicators of the user's learning performance to obtain comparative analysis results; S4.5: According to the comparative analysis results, select appropriate feedback content from a preset feedback template library; S4.6: Use natural language generation algorithms to personalize the selected feedback content and generate personalized feedback or suggestions; S4.7: Adjust the tone and expression of the personalized feedback or suggestions through sentiment analysis algorithms.

6. A personalized learning system with courses and AI training, characterized in that: include: Theoretical course module, used to provide users with theoretical courses in at least one field; The practice scenario generation module is used to generate personalized practice scenarios based on the content and learning status of the theoretical courses learned by the users after the users complete the theoretical courses. An AI training module is used to enable the AI ​​big model to simulate a virtual character through role-playing in the personalized training scenario, and to conduct interactive dialogue training with the user based on the virtual character; The feedback generation module is used to enable the AI ​​model to provide feedback or suggestions based on the user's performance in the interactive dialogue exercise.

7. The system according to claim 6, characterized in that The theoretical course modules include: A learning preference acquisition unit, used to acquire the user's learning preference and learning style; A course selection unit, for selecting suitable theoretical course content and presentation method from a preset course library based on the user's learning preference and learning style; The learning monitoring unit is used to monitor the user's learning behavior and emotional state in real time during the theoretical course learning process; The course adjustment unit is used to dynamically adjust the content difficulty, presentation order and learning rhythm of the theoretical course according to the real-time monitoring results.

8. The system according to claim 6, characterized in that The practice scenario generation module comprises: A learning performance analysis unit is used to analyze the user's performance data during the theoretical course learning process based on a machine learning algorithm after confirming that the user has completed the theoretical course learning, and obtain a quantitative indicator of the user's learning performance; A knowledge mastery evaluation unit, used to determine the user's knowledge mastery and learning characteristics through a preset evaluation model based on the quantitative indicators; A scenario selection unit, for selecting a basic scenario related to the content of the theoretical course studied by the user from a preset scenario library by using a content matching algorithm based on the content of the theoretical course studied by the user; A difficulty adjustment unit is used to adjust the difficulty of the selected basic scenario according to the user's knowledge level to obtain a practice scenario with suitable difficulty; A personalized customization unit, used to customize the difficulty-adapted practice scene based on the user's learning characteristics to obtain a preliminary personalized practice scene; A knowledge point matching unit is used to match the preliminary personalized practice scenario with the content of the theoretical course learned by the user to obtain a practice scenario with matching knowledge points; The final scenario generation unit is used to generate a final personalized practice scenario based on the practice scenario matched with the knowledge points.

9. The system according to claim 6, characterized in that The AI ​​training module includes: A role configuration unit, configured to configure corresponding role information, background knowledge and behavior patterns for the AI ​​big model based on the personalized practice scenario; A virtual character generation unit, used to control the AI ​​big model to generate virtual characters with different personalities and / or different emotional reactions through preset prompt words or user-defined prompt words; A user input parsing unit, used to parse and obtain user input information using a multimodal algorithm; A reply generating unit, configured to generate reply content of the virtual character based on a preset dialogue strategy, user input information, the character information, the background knowledge and the behavior pattern; The sentiment analysis unit is used to identify the sentiment characteristics in the user's input information through a sentiment analysis algorithm, and reflect the corresponding sentiment characteristics in the reply content of the virtual character; The explanation mode switching unit is used to trigger the AI ​​big model to switch to the explanation mode and provide explanations for the knowledge points when it detects that the user has difficulties with the knowledge points of the theoretical course.

10. The system according to claim 8, characterized in that The feedback generation module comprises: Performance monitoring unit, used to monitor the user's performance in interactive dialogue exercises in real time, including answer accuracy, response speed, language expression ability and emotional expression; A capability assessment unit, used to analyze the conversation content of the user in the interactive dialogue exercise using a natural language processing algorithm to obtain the user's capability assessment result; A quantitative scoring unit, used to quantitatively score the user's performance in the interactive dialogue exercise based on a preset evaluation standard and the ability evaluation result, and obtain quantitative indicators of multiple dimensions of the user in the interactive dialogue exercise; A comparative analysis unit, used to compare and analyze the quantitative indicators of multiple dimensions of the user in the interactive dialogue exercise with the quantitative indicators of the user's learning performance to obtain a comparative analysis result; A feedback selection unit, configured to select appropriate feedback content from a preset feedback template library according to the comparison and analysis result; A personalized feedback generation unit, used to personalize the selected feedback content and generate personalized feedback or suggestions by using a natural language generation algorithm; The tone adjustment unit is used to adjust the tone and expression of the personalized feedback or suggestion through a sentiment analysis algorithm.

Images