Intelligent agent system based on large language model and interaction method

Abstract

The invention relates to the technical field of artificial intelligence and natural language processing, in particular to an agent system based on a large language model and an interaction method, and the system comprises a data collection module which is used for receiving text, voice or other forms of data input by a user in real time; the natural language processing module is used for performing semantic understanding and grammatical analysis on the received data to generate an executable user intention; the knowledge reasoning module is used for reasoning an optimal response result or action scheme according to the requirements of the user and historical interaction records; the method has the beneficial effects that the interaction experience of the user is greatly improved through natural language processing, sentiment analysis and personalized response technologies. The system can perform accurate response according to the emotion and demand of the user, and can provide efficient and intimate services in applications such as intelligent customer service, intelligent assistant, personalized recommendation and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence and natural language processing, and specifically to an intelligent agent system and interaction method based on a large language model. Background Art

[0002] While large language models currently used in artificial intelligence demonstrate powerful natural language processing capabilities, their deployment as intelligent agent systems still faces numerous technical challenges. Traditional rule-based intelligent agent systems require manual pre-definition of all possible task flows, a rigid architecture that struggles to adapt to the complex and ever-changing demands of real-world scenarios. Directly applying large language models also suffers from insufficient task decomposition capabilities, particularly when handling complex tasks with multi-step logical dependencies, often leading to planning gaps or miscommunication.

[0003] In terms of interactive memory, existing systems generally use a simple conversation caching mechanism. This short-term memory approach cannot support continuous learning across sessions, resulting in a lack of personalized service capabilities. Furthermore, traditional memory retrieval methods are inefficient, making it difficult to quickly locate relevant information within massive amounts of stored data, severely impacting system responsiveness. Furthermore, imperfect memory update mechanisms introduce a large amount of noisy data, further reducing system availability.

[0004] Security issues are particularly prominent in the tool invocation process. Most current systems lack strict permission control mechanisms, which can lead to the risk of unauthorized access to API calls generated by large language models. Inaccurate parameter generation often leads to tool execution failures, and the system lacks effective recovery capabilities, often disrupting the entire task process once an error occurs. These issues are particularly prominent in security-critical fields such as finance and healthcare.

[0005] The inherent "hallucination" problem of large language models also poses challenges to the reliability of intelligent agent systems. Models may generate content containing factual errors, and existing technologies lack effective real-time verification methods. This unreliability can severely impact the system's practical value, particularly in scenarios requiring specialized knowledge. While some improvements have attempted to address these issues, they often come at the expense of system performance, resulting in increased response latency and architectural complexity, making them difficult to implement on a large scale in production environments. Summary of the Invention

[0006] The purpose of the present invention is to provide an intelligent agent system and interaction method based on a large language model to solve the problems raised in the above background technology.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an intelligent agent system based on a large language model, comprising:

[0008] Data collection module, used to receive text, voice or other forms of data input by users in real time;

[0009] The natural language processing module is used to perform semantic understanding and grammatical analysis on the received data and generate executable user intent;

[0010] The knowledge reasoning module is used to infer the optimal response or action plan based on user needs and historical interaction records;

[0011] The output module is used to feed back the generated response results to the user in the form of text, voice or other forms;

[0012] Adaptation module, used to adjust the style and content of responses based on user emotions, behaviors, and preferences, thereby improving the interactive experience.

[0013] Preferably, it also includes: a data security module for encrypting user data and ensuring the privacy and security of data in all aspects of the system.

[0014] Preferably, it also includes: a multimodal interaction module that supports multiple input and output methods such as text, voice, and image, thereby enhancing the interaction between users and intelligent agents.

[0015] An interaction method for an intelligent agent system based on a large language model comprises the following steps:

[0016] Step 1: Receive data input from the user through text, voice or other forms;

[0017] Step 2: Perform natural language understanding and intent recognition on the received data and generate appropriate response content;

[0018] Step 3: Adjust the content and style of the response based on the user's historical behavioral data, emotional state, and preferences;

[0019] Step 4: Feedback the generated response content to the user through text, voice, etc.

[0020] Step 5: Continuously optimize the response strategy during the interaction process based on user feedback data.

[0021] Preferably, in step 3, adjusting the response content and style includes sentiment analysis, and adjusting the tone and expression of the content according to the user's emotional state.

[0022] Preferably, step 5 further includes: analyzing the user's behavioral data through a machine learning algorithm and optimizing the system's response strategy to provide more accurate personalized services.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] The intelligent agent system and interaction method based on a large language model proposed in this invention greatly enhance the user's interactive experience through natural language processing, sentiment analysis, and personalized response technology. The system can accurately respond to the user's emotions and needs, and can provide efficient and friendly services in applications such as intelligent customer service, intelligent assistants, and personalized recommendations. Secondly, by automating tasks, the intelligent agent system significantly improves work efficiency and reduces manual intervention, thereby saving a lot of time and costs. Whether in enterprise production scheduling, intelligent customer service, or personal life management, the intelligent agent system can efficiently complete tedious and repetitive tasks, greatly reducing the manpower burden.

[0025] It effectively reduces labor costs, allowing businesses and organizations to reduce their reliance on manual customer service and support positions and instead provide round-the-clock services through intelligent agents, ensuring that they can respond to user needs at any time. The intelligent agent's auxiliary decision-making function further improves the quality of decision-making in various industries, especially in the fields of healthcare, education, and corporate management. The system helps users and managers make more accurate decisions through data analysis, real-time feedback, and personalized suggestions. Overall, the intelligent agent system based on the large language model not only improves efficiency, but also enhances cross-domain collaboration capabilities. At the same time, it promotes the popularization of technology in more fields, making scientific and technological services more intelligent and convenient. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0027] In order to clearly and completely describe the objectives and technical solutions of the present invention and make the advantages more clearly understood, the embodiments of the present invention are further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are part of the embodiments of the present invention, not all of them, and are only used to explain the embodiments of the present invention, not to limit the embodiments of the present invention. All other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0028] In the first embodiment, the present invention provides a technical solution: an intelligent agent system based on a large language model, which is composed of the following modules:

[0029] (1) Large language model module:

[0030] The large language model module is the core of the entire system, responsible for understanding and generating natural language. It is based on deep learning technology, typically using pre-trained models such as GPT and BERT, and fine-tuning them to meet the needs of specific domains. The main functions of this module include:

[0031] Semantic understanding: Large language models can parse text or voice commands input by users and extract key information, including intent, entities, emotions, etc.

[0032] Context processing: This module supports multi-round dialogues and can track the context information of the dialogue, maintaining the consistency of the context during long-term interactions and avoiding misunderstanding of user intentions.

[0033] Text generation: Large language models also have the ability to generate natural language, and can generate answers or operation instructions that conform to grammatical norms and are semantically accurate.

[0034] Multi-language support: This module supports multiple languages ​​and can provide cross-language services for users with different language backgrounds.

[0035] Technical Details: This system uses a Transformer architecture and a self-attention mechanism to process long text data, improving its ability to understand complex sentence structures and semantic relationships. Pre-trained on a large-scale corpus, it demonstrates strong semantic adaptability in multi-domain contexts.

[0036] (2) Knowledge base module:

[0037] The knowledge base module is the "think tank" of the system, storing a large amount of domain knowledge, user history, and answers to common questions. Its main functions are as follows:

[0038] Domain knowledge storage: Contains industry knowledge, expert literature, technical manuals, and other content related to user needs, providing data support for large language models.

[0039] History management: Records the user's past interactions (such as conversation history, preference settings, etc.), helping the system provide personalized services to the same user and speeding up task processing.

[0040] Task Mapping and Query: The knowledge base supports fast search and reasoning through structured information and semantic indexing. For example, when providing product recommendations, the system can query the knowledge base to obtain a list of recommendations that suits the user's needs.

[0041] Technical Details: The knowledge base uses a graph database, relational database, or NoSQL storage structure to support fast queries and flexible data models. Natural language understanding technology is used to match information in the knowledge base with natural language input by users.

[0042] (3)Task management module:

[0043] The task management module is the scheduling center of the intelligent agent system, responsible for analyzing user needs and converting them into specific tasks that can be executed. It mainly includes the following functions:

[0044] Task parsing: Analyzes user input (such as natural language instructions, query requests, etc.) to identify specific task objectives. For example, if a user requests "check today's weather for me," the task management module will parse the task into "check the weather."

[0045] Task Scheduling: Schedules and executes tasks based on their priorities, dependencies, and available resources. The scheduling module ensures that the system can efficiently handle multiple concurrent tasks and avoid resource conflicts.

[0046] Task feedback: After the task is completed, the management module will feedback the results to the user and may make subsequent adjustments based on user feedback.

[0047] Technical Details: Utilizes task queues, concurrent processing, and asynchronous call mechanisms to ensure efficient task execution. Machine learning techniques are combined to prioritize tasks and allocate resources, optimizing system performance.

[0048] (4) Emotion recognition and adaptation module:

[0049] The emotion recognition and adaptation module dynamically adjusts the agent's interaction strategy by analyzing the user's voice or text input to better meet the user's emotional needs. Its main functions are as follows:

[0050] Sentiment Analysis: This module analyzes the user's language, tone, keywords, and other information to identify the user's emotional state (e.g., happiness, confusion, anger, etc.). This module not only analyzes text but also combines voice and tone to identify emotions.

[0051] Interaction Adaptation: Based on the results of sentiment analysis, the agent's tone, wording, and response methods are adjusted. For example, if the user appears troubled, the system might choose a gentle, encouraging tone; if the user is in a good mood, a more relaxed and humorous tone might be adopted.

[0052] Emotion Tracking: This module can track emotional changes in real time and adjust the agent's behavior through emotional feedback from multiple rounds of conversations, providing a more humane experience.

[0053] Technical Details: Use sentiment analysis models, such as deep neural networks (DNNs) or convolutional neural networks (CNNs), to extract sentiment features. Combined with historical user interaction data, adaptive algorithms are employed to continuously optimize sentiment recognition and adaptation strategies.

[0054] (5) Multimodal interaction module:

[0055] The multimodal interaction module enables the system to interact with users through multiple input methods, including text, voice, images, etc. Its main functions are as follows:

[0056] Voice input and output: Supports users to interact with the system through voice, and converts the system's responses into voice through speech synthesis technology to provide voice assistant services.

[0057] Text input and output: Supports input through text, and the system will generate corresponding answers through text.

[0058] Image recognition and processing: Users can interact with the system by uploading images, and the system will identify the image content and provide corresponding feedback. For example, a user uploads a photo, and the system identifies the object in it and provides relevant information.

[0059] Cross-modal information fusion: The system can integrate information from different modalities to improve the richness and accuracy of interactions. For example, when a user asks about the weather via voice, the system can simultaneously display relevant weather images or videos.

[0060] Technical Details: Speech recognition uses deep learning technologies (such as Wav2Vec and DeepSpeech) to convert speech into text. Speech synthesis uses text-to-speech (TTS) technologies such as WaveNet or Tacotron to generate natural and fluent speech. Image recognition uses computer vision technologies such as convolutional neural networks (CNNs) to analyze image content.

[0061] Example 2, based on Example 1, proposes an interaction method for an intelligent agent system based on a large language model, comprising the following steps:

[0062] 1) Natural language input recognition:

[0063] When a user interacts with an agent, they first issue commands or ask questions through voice or text input. The task of the natural language input recognition module is to convert these inputs into structured data that the agent can understand. This process involves multiple sub-steps:

[0064] Speech-to-text: When a user inputs a voice message, the Automatic Speech Recognition (ASR) module first converts the voice signal into text. This process takes into account factors such as accent, noise, and speaking speed to ensure the accuracy of the text.

[0065] Text preprocessing: For text input, the system first performs necessary preprocessing, including word segmentation, stop word removal, synonym normalization, spelling correction, etc., to improve the accuracy of subsequent processing.

[0066] Syntax parsing: Using natural language processing techniques (such as syntactic analysis and dependency analysis), the system analyzes the grammatical structure of the input text and identifies key information such as the subject-verb-object relationship, time, and location. For example, the query "What is the meeting time tomorrow?" will be parsed as "inquiry" + "meeting time" + "tomorrow."

[0067] Entity recognition: Through named entity recognition (NER) technology, the system can extract key information from the text, such as names of people, places, dates, product names, etc. This step is crucial for subsequent task understanding.

[0068] Intent recognition: The intent recognition module understands the semantics of user input and determines the user's actual needs, such as whether they are requesting a query, an action, or a suggestion. Intent recognition typically relies on machine learning models such as support vector machines (SVMs) and deep neural networks (DNNs).

[0069] Technical Details: Deep Neural Networks (DNNs) are used for speech-to-text conversion. Traditional NLP methods (such as CRF and HMM) are combined with modern pre-trained language models (such as BERT and GPT) for text preprocessing, parsing, and intent recognition.

[0070] 2) Task analysis and intent recognition:

[0071] Once the user input is converted into structured data, the Task Parsing and Intent Recognition module is responsible for analyzing this data and determining the user's specific needs. This is a key step in the system's efficient response to user needs. The specific process is as follows:

[0072] User Needs Analysis: Through intent classification technology, the system first identifies the user's core needs. For example, if the user enters "What are today's stock market conditions?" the system will identify the task type as "Stock Market Quotes Query."

[0073] Task-to-Goal Mapping: During this phase, the system matches the identified intent with the task model in the knowledge base, finds the appropriate task template, and extracts relevant entities. For example, if the system recognizes that the user is asking about stock market information, it will retrieve real-time data sources related to stock market information from the knowledge base.

[0074] Context-dependent processing: Task parsing relies not only on the current input but also considers historical conversational information in the context. For example, if the user has previously asked about stock market conditions, the system can automatically identify "conditions" in subsequent conversations as referring to the stock market, eliminating the need for reconfirmation.

[0075] Multi-task parallel processing: If a user raises multiple questions or requests simultaneously, the task parsing module can effectively break these requests down into multiple subtasks and process them in parallel. For example, the request "Check the weather and my schedule" would be broken down into two subtasks: "Check the weather" and "Check the schedule."

[0076] Technical Details: Deep learning intent recognition models (such as LSTM, BERT, and BART) are used to classify user intent. Information retrieval techniques, such as TF-IDF or Embedding search, are used to extract relevant information from the knowledge base.

[0077] 3) Response Generation and Multi-Round Dialogue:

[0078] After identifying the user's intent, the response generation module generates an appropriate response, ensuring that the system's response is not only grammatically correct but also contextually appropriate. For multi-turn conversations, the system must be able to maintain contextual coherence based on the current conversation state. The main process is as follows:

[0079] Response Generation: Large language models (such as GPT and T5) generate responses based on user intent and context. The generated content must not only accurately answer the question but also be concise, clear, and consistent with the user's emotional state. For example, if the user's tone is urgent, the system may respond with a brief and direct response.

[0080] Context Management: The system maintains and tracks contextual state through the dialogue management module. For multi-turn conversations, the system needs to remember previous conversations and use them to provide more appropriate responses. For example, if a user first asks about the weather and then about the temperature, the system should understand that "temperature" refers to the temperature that day, not the user's previous question.

[0081] Multi-turn dialogue management: The challenge in multi-turn dialogues is that the system must not only understand the user's intent in each turn, but also track the conversation history to ensure that each answer is provided in the appropriate context. For example, the answer to "What's the weather like today?" might be "It's sunny today." If the user then asks "Do you need an umbrella?" the system needs to rely on the weather information from the previous conversation to generate an accurate response.

[0082] Technical Details: The system uses dialogue management technologies such as RNN, LSTM, and Transformer to handle complex dialogue structures and maintain context. Predictive models based on user behavior can be used to determine whether the user wishes to continue the conversation or whether the system can switch topics.

[0083] 4) Task execution and feedback:

[0084] After determining the user's task requirements and generating an appropriate response, the Task Execution and Feedback Module actually performs the corresponding operations and provides feedback to the user. The module's workflow is as follows:

[0085] Task Execution: If a user's request requires the system to perform an action (such as querying information, scheduling tasks, or controlling devices), the Task Execution module will call the corresponding execution engine or API based on the parsed task type. For example, if a user requests weather information, the system will call the external weather data interface; if a user requests to set an alarm, the system will call the device control module to perform the operation.

[0086] Result Feedback: After a task is completed, the system returns the task results to the user. This may include query results, a notification of completion, or further suggestions. The system also provides options for subsequent interactions. For example, if a user asks about the weather, the system may then ask if they want to bring an umbrella.

[0087] Multi-dimensional feedback adjustment: Based on user feedback, the system can adjust subsequent behaviors. For example, if a user is dissatisfied with a piece of feedback, the system can further inquire about the user's expectations and adjust the subsequent dialogue strategy.

[0088] Technical Details: Automatically execute tasks through an integrated API or internal execution engine. Use feedback mechanisms (such as system evaluation and user evaluation) to adjust the accuracy and satisfaction of task execution.

[0089] Example 3, based on Example 2, proposes that in an intelligent customer service system, an agent system based on a large language model can greatly improve service efficiency and user experience. Intelligent customer service achieves the following functions through natural language processing and multi-round dialogue capabilities:

[0090] Automatic question recognition and answering: By understanding the semantics of users' natural language questions, the system can generate accurate responses in real time. For example, if a user asks about product features, pricing, after-sales service, etc., the system will generate standardized responses based on information in the knowledge base.

[0091] Emotional Adjustment and User Experience Optimization: The intelligent customer service system uses sentiment analysis to identify users' emotional states. If a user feels anxious because their question remains unanswered, the system adjusts the tone of response based on the emotional adaptation module to soothe the user and prevent negative emotions from spreading.

[0092] Automated problem handling and transfer: The system can handle common, standardized user issues. For complex issues or those requiring manual processing, the intelligent agent will automatically transfer the call to human customer service, avoiding users switching back and forth between manual and automated processes.

[0093] 24 / 7 service: The intelligent customer service system can provide real-time online service at any time, reducing user waiting time and improving the overall service experience.

[0094] Technical Details: An integrated knowledge base and real-time data interface provide customers with the latest product and service information. A multi-round dialogue and task scheduling system effectively handles complex customer needs. Based on sentiment analysis and historical user interaction data, personalized responses are provided.

[0095] Application scenarios include e-commerce platforms: When users inquire about product issues, refunds, or order status during their shopping journey, intelligent customer service can help resolve issues or provide detailed product information, enhancing the shopping experience. Alternatively, in banking and finance, intelligent customer service can provide services such as account inquiries, transfers, and bill inquiries, while also using emotional recognition to provide gentle, soothing language to address customer concerns.

[0096] Example 4, based on Example 2, proposes that in the field of education, an intelligent agent system based on a large language model can provide students with personalized learning guidance and course recommendation services. Functions include:

[0097] Personalized learning path recommendations: By analyzing students' historical learning data, interests, and abilities, the system tailors learning plans for them, recommending appropriate learning resources and progress. For example, based on a student's weaknesses in mathematics, the system can recommend relevant exercises and tutorials.

[0098] Intelligent Q&A and real-time tutoring: When students encounter problems during the learning process, they can ask questions to the intelligent agent at any time. The system provides instant answers through natural language understanding and knowledge base support.

[0099] Adaptive assessment and feedback: The intelligent system can conduct real-time assessments based on students’ homework performance, test scores, and learning progress, provide targeted feedback, and dynamically adjust the difficulty of teaching content during the students’ learning process.

[0100] Multimodal learning experience: By combining multimodal input and output methods such as text, voice, and video, a richer and more interactive learning experience is provided. For example, students can ask questions via voice, and the system will provide detailed explanations via images or videos.

[0101] Technical Details: Combining natural language generation (NLG) with deep learning algorithms, this platform adapts learning content in real time based on student progress. Using adaptive learning technology, it optimizes learning paths based on student performance and provides personalized learning resources. It integrates interactive teaching materials, multimedia content, and real-time online testing to enhance the learning experience in a variety of ways.

[0102] Application scenario examples:

[0103] Online education platforms, such as Khan Academy and Duolingo, use intelligent agents to provide students with adaptive learning content and tests to help them master key knowledge points. Homework tutoring: The system can help students complete homework, provide answers to questions, and provide study suggestions.

[0104] Example 5, based on Example 2, proposes that in the field of medical health, the intelligent agent system can provide preliminary health consultation, symptom analysis and doctor auxiliary decision support. Its functions include:

[0105] Intelligent symptom analysis: Users can enter symptoms, and the intelligent system analyzes the symptoms to provide possible disease information and infer the disease type based on other information provided by the user. For example, "headache with fever" may indicate the possibility of influenza, cold, etc.

[0106] Health management and reminders: The intelligent agent can help users record health data (such as blood pressure, blood sugar, weight, etc.), provide health management suggestions based on the data, and set health reminders.

[0107] Medical information query and answer: The system can quickly query medical literature, drug information and treatment plans to help patients answer health questions.

[0108] Assisted diagnosis and treatment recommendations: Based on medical data and knowledge bases, intelligent agents can provide doctors with auxiliary diagnosis and treatment recommendations. While intelligent agents cannot replace doctors, they can improve diagnostic efficiency and accuracy, especially in the initial diagnosis stage.

[0109] Technical Details: Combining a medical knowledge base with big data analytics, the intelligent body can quickly match symptoms with diseases and provide preliminary diagnostic recommendations. It uses machine learning and deep learning algorithms to analyze patients' health data and provide personalized health management recommendations. It collaborates with doctors to provide decision support based on medical imaging, medical records, and laboratory results.

[0110] Application examples: Initial consultation and health management: Users consult with agents about their symptoms, and the system generates preliminary recommendations based on the input symptoms and provides subsequent health management plans. Assisting physicians in decision-making: Doctors use agents to analyze cases and recommend treatment options, reducing clinical burden and improving diagnosis and treatment efficiency.

[0111] Example 6, based on Example 2, proposes that in enterprise production and management, the intelligent agent system can help improve production efficiency, optimize decision-making and automate management. Its main functions include:

[0112] Task Automation and Scheduling: Agents automatically schedule and execute daily work tasks through the task management module. For example, they can schedule meetings, handle emails, and track progress.

[0113] Data Analysis and Decision Support: Intelligent agents can quickly process large amounts of enterprise data, generate analytical reports, and provide decision support based on this data. For example, they can analyze market trends, sales data, production efficiency, etc. to provide decision-making basis for enterprise management.

[0114] Employee training and support: Intelligent agents can provide employees with training materials, tutorials, FAQs, etc., to help them quickly resolve work questions and improve work efficiency.

[0115] Customer Relationship Management (CRM): Through continuous interaction with customers, intelligent agents can help companies improve customer satisfaction and loyalty, and enhance their market competitiveness through precise customer service and personalized recommendation strategies.

[0116] Technical details: Integrate big data analysis platform, combine machine learning and predictive analysis, generate business reports and decision support.

[0117] Use natural language generation (NLG) technology to automatically generate reports, reducing manual analysis and reporting time. Use enterprise resource planning (ERP) systems combined with intelligent agents to optimize resources and schedule tasks.

[0118] Example application scenarios: Enterprise production scheduling: Agents assist with production line scheduling and task management, automating production planning, reducing human intervention, and improving production efficiency. Market and sales analysis: Using data analysis tools, agents can provide management with real-time sales data.

[0119] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent agent system based on a large language model, characterized by: include: Data collection module, used to receive text, voice or other forms of data input by users in real time; The natural language processing module is used to perform semantic understanding and grammatical analysis on the received data and generate executable user intent; The knowledge reasoning module is used to infer the optimal response or action plan based on user needs and historical interaction records; The output module is used to feed back the generated response results to the user in the form of text, voice or other forms; Adaptation module, used to adjust the style and content of responses based on user emotions, behaviors, and preferences, thereby improving the interactive experience.

2. The agent system based on a large language model according to claim 1, characterized in that: Also includes: The data security module is used to encrypt user data and ensure the privacy and security of data in all aspects of the system.

3. The agent system based on a large language model according to claim 2, characterized in that: Also includes: The multimodal interaction module supports multiple input and output methods such as text, voice, and image, enhancing the interaction between users and intelligent agents.

4. An interactive method for an intelligent agent system based on a large language model according to claim 3, characterized in that: The following steps are involved: Step 1: Receive data input from the user through text, voice or other forms; Step 2: Perform natural language understanding and intent recognition on the received data and generate appropriate response content; Step 3: Adjust the content and style of the response based on the user's historical behavioral data, emotional state, and preferences; Step 4: Feedback the generated response content to the user in the form of text or voice; Step 5: Continuously optimize the response strategy during the interaction process based on user feedback data.

5. An interactive method according to claim 4, characterized in that: In step 3, adjusting the response content and style includes sentiment analysis and adjusting the tone and content expression according to the user's emotional state.

6. An interactive method according to claim 5, characterized in that: Step 5 further includes: analyzing user behavior data through machine learning algorithms and optimizing the system's response strategy to provide more accurate personalized services.

Images