Multi-agent question answering system based on large language model

Abstract

The invention relates to the technical field of question answering systems. The question answering system comprises a language understanding generation module, a multi-agent coordination module, a knowledge graph retrieval module, an emotion analysis adaptation module, a virtual reality interaction module, a time travel simulation module, a cross-language processing module and a future prediction reasoning module. The language understanding generation module is used for semantic analysis and response generation; the multi-agent coordination module is used for agent distribution and agent interaction; according to the multi-agent question answering system based on the large language model, through deep semantic analysis of the large language model, the intention and key information of a user question can be accurately understood, and according to the type and field of the question, the most suitable agent can be dynamically allocated for answering; necessary communication and cooperation can be carried out among a plurality of agents, information circulation and answer consistency are ensured, a cross-domain knowledge graph is constructed, and rich knowledge support is provided for answer of complex questions.

Description

Technical Field

[0001] The present invention relates to the technical field of question answering systems, and in particular to a multi-agent question answering system based on a large language model. Background Art

[0002] A large language model refers to a natural language processing model with a large number of parameters and complex structures, such as BERT and GPT. These models are trained on a large amount of text data through deep learning technology and can capture the complex characteristics and rules of language. Large language models have powerful language generation and comprehension capabilities, can generate natural and fluent text, understand and answer complex questions, and even perform creative writing and translation tasks.

[0003] An intelligent agent refers to a computer program or system that has the ability to make autonomous decisions and actions. It can perceive the environment, understand tasks, make plans and execute actions to achieve goals. In question-answering systems, intelligent agents are usually used as entities that answer user questions. They can dynamically call different knowledge and algorithms to generate answers based on the type and domain of the question.

[0004] Therefore, a question-answering system is a computer system that can automatically answer user questions. It extracts relevant information from a large amount of text data and generates answers that are easy for users to understand through natural language processing, information retrieval, and reasoning. Question-answering systems are widely used in intelligent customer service, intelligent education, intelligent medical care and other fields, and can improve work efficiency and user satisfaction.

[0005] However, in general, question-answering systems usually match predefined keyword and answer pairs. When the question entered by the user matches a keyword, the system returns the corresponding answer. It is unable to handle complex questions and semantic understanding, and may not be able to give correct answers to similar questions with different expressions. The rule-based question-answering method generates answers by defining a series of rules and templates. The rules are usually based on manually written logic and conditions. The rules are complex to write and difficult to maintain. They may not be able to handle situations that are not covered, and lack flexibility and scalability.

[0006] In summary, it is necessary to propose a multi-agent question-answering system based on a large language model to solve the above problems. Summary of the invention

[0007] The purpose of the present invention is to provide a multi-agent question-answering system based on a large language model to solve the problems raised in the above background technology.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A multi-agent question-answering system based on a large language model, including a language understanding generation module, a multi-agent coordination module, a knowledge graph retrieval module, a sentiment analysis adaptation module, a virtual reality interaction module, a time travel simulation module, a cross-language processing module, and a future prediction reasoning module;

[0010] The language understanding generation module is used for semantic analysis and response generation;

[0011] The multi-agent coordination module is used for agent allocation and agent interaction;

[0012] The knowledge graph retrieval module is used for graph construction and graph query;

[0013] The sentiment analysis adaptation module is used for sentiment detection and sentiment adaptation;

[0014] The virtual reality interaction module is used for scene construction and interactive control;

[0015] The time travel simulation module is used for historical data integration and time simulation;

[0016] The cross-language processing module is used for language identification and translation adaptation;

[0017] The future prediction reasoning module is used for trend analysis and hypothesis reasoning.

[0018] Preferably, the language understanding generation module further includes a semantic parsing unit and a response generation unit;

[0019] The semantic parsing unit uses a pre-trained language model to perform deep semantic parsing on the input question to extract key information and intent;

[0020] The response generation unit is used to generate a natural and fluent answer based on a large language model.

[0021] Preferably, the multi-agent coordination module further includes an agent allocation unit and an agent interaction unit;

[0022] The agent allocation unit is used to dynamically allocate the most suitable agent to answer the question according to the question type and field;

[0023] The agent interaction unit is used to manage the communication and collaboration between multiple agents to ensure information flow and consistency.

[0024] Preferably, the knowledge graph retrieval module further includes a graph construction unit and a graph query unit;

[0025] The graph construction unit is used to construct and maintain a cross-domain knowledge graph to support answers to complex questions;

[0026] The graph query unit is used to retrieve relevant information in the knowledge graph and present it in a structured manner in response to user questions.

[0027] Preferably, the emotion analysis adaptation module further comprises an emotion detection unit and an emotion adaptation unit;

[0028] The emotion detection unit is used to identify the emotional tendency in the user's question to adjust the tone and style of the answer;

[0029] The emotion adaptation unit is used to dynamically adjust the answer strategy according to the emotion detection result to improve user satisfaction.

[0030] Preferably, the virtual reality interaction module further includes a scene construction unit and an interaction control unit;

[0031] The scene construction unit is used to explain complex concepts or processes and construct virtual reality scenes to assist understanding;

[0032] The interaction control unit is used to allow the user to interact with objects in the virtual reality environment to enhance the experience.

[0033] Preferably, the time travel simulation module further comprises a historical data integration unit and a time simulation unit;

[0034] The historical data integration unit is used to collect and integrate historical data to provide a basis for time travel simulation;

[0035] The time simulation unit is based on historical data and is used to simulate the social and cultural status at a specific time point and answer hypothetical questions.

[0036] Preferably, the cross-language processing module further includes a language identification unit and a translation adaptation unit;

[0037] The language recognition unit is used to automatically recognize the language type of the user input and supports multi-language input;

[0038] The translation adaptation unit is used to translate non-native language questions into a language supported by the system and adapt the language habits.

[0039] Preferably, the future prediction reasoning module further includes a trend analysis unit and a hypothesis reasoning unit;

[0040] The trend analysis unit is used to analyze existing data, predict future trends, and provide answers to forward-looking questions;

[0041] The hypothesis reasoning unit is based on logical reasoning and hypothesis analysis and is used to reason about situations that have not occurred or will not occur.

[0042] Based on the above system, the present invention also proposes a multi-agent question-answering method based on a large language model, comprising the following steps:

[0043] S1. User input: receiving questions raised by users in natural language, in text, voice or other forms;

[0044] S2. Language understanding: Use the semantic parsing unit in the language understanding generation module to perform deep semantic analysis on the question and extract key information and intent;

[0045] S3. Agent allocation and interaction: According to the problem type and domain, the most suitable agent is dynamically allocated through the agent allocation unit in the multi-agent coordination module, and multiple agents communicate and collaborate as necessary through the agent interaction unit;

[0046] S4. Knowledge retrieval and reasoning: In the knowledge graph retrieval module, the graph query unit is used to retrieve relevant information in the knowledge graph. For problems that require prediction or reasoning, the future prediction and reasoning module is used for analysis.

[0047] S5. Sentiment analysis and adaptation: The sentiment analysis and adaptation module detects the sentiment of user questions and adjusts the tone and style of the answer based on the results;

[0048] S6.Response generation: Generate natural and fluent responses in the response generation unit based on the large language model;

[0049] S7. Cross-language processing: For non-native language input, the cross-language processing module performs language recognition and translation adaptation to ensure that the system can understand and answer;

[0050] S8. Virtual reality interaction: For problems that require virtual reality to assist understanding, use the virtual reality interaction module to build scenes and provide interactive experience;

[0051] S9. Time Travel Simulation: For answers to historical or hypothetical questions, use the Time Travel Simulation module to simulate and reason;

[0052] S10. Output feedback: Return the generated answer to the user in the form of text, voice or other forms, and collect user feedback for subsequent optimization.

[0053] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention can accurately understand the intention and key information of user questions through deep semantic analysis of a large language model, and can give accurate answers even for complex questions and questions with diverse expressions. According to the type and field of the question, the most suitable intelligent agent can be dynamically assigned to answer, and multiple intelligent agents can also perform necessary communication and collaboration to ensure the flow of information and consistency of answers, build a cross-domain knowledge graph, and provide rich knowledge support for answers to complex questions. Through graph query and reasoning technology, the present invention can accurately retrieve and present relevant information, perform sentiment detection on user questions, and dynamically adjust the tone and style of answers according to sentiment tendencies, thereby improving user satisfaction and experience, supporting multi-language input and output, and realizing language recognition and translation adaptation through a cross-language processing module to ensure that questions in different languages ​​can be understood and answered. For questions that require virtual reality assistance to understand or involve historical hypothetical questions, virtual reality interaction and time travel simulation functions can be provided to enhance user understanding and experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 A topological diagram of a multi-agent question-answering system based on a large language model according to the present invention;

[0055] Figure 2 This is a flow chart of the multi-agent question-answering method of the present invention. DETAILED DESCRIPTION

[0056] 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.

[0057] Example 1

[0058] See also Figure 1 ,The present invention proposes a multi-agent question-answering system based on a large language model, including a language understanding generation module, a multi-agent coordination module, a knowledge graph retrieval module, a sentiment analysis adaptation module, a virtual reality interaction module, a time travel simulation module, a cross-language processing module, and a future prediction reasoning module;

[0059] Among them, the present invention performs semantic parsing and response generation through a language understanding generation module, performs agent allocation and agent interaction through a multi-agent coordination module, performs graph construction and graph query through a knowledge graph retrieval module, performs emotion detection and emotion adaptation through a sentiment analysis adaptation module, performs scene construction and interactive control through a virtual reality interaction module, performs historical data integration and time simulation through a time travel simulation module, performs language recognition and translation adaptation through a cross-language processing module, and uses a future prediction reasoning module for trend analysis and hypothesis reasoning.

[0060] In this embodiment, it should also be noted that the language understanding generation module of the system also includes a semantic parsing unit and a response generation unit. The semantic parsing unit uses a pre-trained language model to perform deep semantic parsing on the input question to extract key information and intent, and the response generation unit is used to generate natural and fluent answers based on the large language model.

[0061] In this embodiment, it should also be noted that the multi-agent coordination module of the system also includes an agent allocation unit and an agent interaction unit. The agent allocation unit is used to dynamically allocate the most suitable agent to answer the question according to the type and field of the question, and the agent interaction unit is used to manage the communication and collaboration between multiple agents to ensure information flow and consistency.

[0062] In this embodiment, it should also be noted that the knowledge graph retrieval module of the system also includes a graph construction unit and a graph query unit. The graph construction unit is used to construct and maintain cross-domain knowledge graphs to support answers to complex questions. The graph query unit is used to retrieve relevant information in the knowledge graph and present it in a structured manner based on user questions.

[0063] In this embodiment, it should also be noted that the sentiment analysis adaptation module of the system also includes a sentiment detection unit and a sentiment adaptation unit. The sentiment detection unit is used to identify the sentiment tendency in the user's questions to adjust the tone and style of the answer. The sentiment adaptation unit is used to dynamically adjust the answer strategy according to the sentiment detection results to improve user satisfaction.

[0064] In this embodiment, it should also be noted that the virtual reality interaction module of the system also includes a scene construction unit and an interaction control unit. The scene construction unit is used to explain complex concepts or processes and construct virtual reality scenes to assist understanding. The interaction control unit is used to allow users to interact with objects in a virtual reality environment to enhance the experience.

[0065] In this embodiment, it should also be noted that the time travel simulation module of the system also includes a historical data integration unit and a time simulation unit. The historical data integration unit is used to collect and integrate historical data to provide a basis for time travel simulation. The time simulation unit is based on historical data and is used to simulate the social and cultural status at a specific time point and answer hypothetical questions.

[0066] In this embodiment, it should also be noted that the cross-language processing module of the system also includes a language recognition unit and a translation adaptation unit. The language recognition unit is used to automatically identify the language type of the user input and supports multi-language input. The translation adaptation unit is used to translate non-native language questions into a language supported by the system and adapt to language habits.

[0067] In this embodiment, it should also be noted that the future prediction and reasoning module of the system also includes a trend analysis unit and a hypothesis reasoning unit. The trend analysis unit is used to analyze existing data, predict future trends, and provide answers to forward-looking questions. The hypothesis reasoning unit is based on logical reasoning and hypothesis analysis, and is used to reason about situations that have not occurred or are not likely to occur.

[0068] Example 2

[0069] In the actual application process, the multi-agent question-answering method based on the above system specifically includes the following steps:

[0070] S1. User input

[0071] S1.1. Receiving input: The system receives user input through the user interface (such as text box, voice recognition interface, etc.);

[0072] S1.2. Input form recognition: The system automatically determines the input form (text, voice, image, etc.) and performs corresponding preprocessing;

[0073] For example, voice input needs to be converted into text through speech recognition;

[0074] S1.3. Input storage and preprocessing: The questions input by the user are stored in the system and basic preprocessing is performed, such as removing irrelevant characters and word segmentation, in preparation for subsequent processing;

[0075] S2. Language understanding

[0076] S2.1. Semantic parsing initialization: Call the semantic parsing unit in the language understanding generation module to initialize the model and algorithm required for parsing;

[0077] S2.2. Deep semantic analysis: Use pre-trained language models (such as BERT, GPT, etc.) to perform deep semantic analysis on questions;

[0078] The algorithm extracts key information and intent from the question through techniques such as context understanding, entity recognition, and relationship extraction;

[0079] For example, the BERT model captures language features through a multi-layer Transformer structure and outputs a vector representation of the question;

[0080] S2.3. Parsing result storage: The key information and intent obtained from the parsing are stored in the system for use in subsequent steps;

[0081] This information includes the subject of the question, the specific content of the inquiry, the user's intention, etc.

[0082] S3. Agent Allocation and Interaction

[0083] S3.1. Agent allocation strategy selection: According to the type and domain of the problem, the most appropriate agent allocation strategy is selected through methods such as rule-based matching and machine learning model prediction;

[0084] S3.2. Dynamic allocation of agents: The agent allocation unit in the multi-agent coordination module dynamically selects the most suitable agent to answer according to the allocation strategy, including the evaluation of the agent's capabilities, expertise and current load;

[0085] S3.3. Interaction and collaboration between agents: If a question requires multiple agents to answer together, the agent interaction unit manages the communication and collaboration between agents, including determining the interaction protocol, transmitting necessary information, and coordinating the consistency of answers;

[0086] S4. Knowledge retrieval and reasoning

[0087] S4.1. Knowledge graph retrieval initialization: call the graph query unit in the knowledge graph retrieval module to initialize the algorithm and database connection required for retrieval;

[0088] S4.2. Knowledge graph retrieval: Retrieve information related to the question in the knowledge graph, including traversal of the graph, matching of nodes and edges, and relevance sorting;

[0089] For example, using the SPARQL query language to retrieve data in an RDF graph;

[0090] S4.3. Reasoning and analysis: For problems that require prediction or reasoning, use the future prediction and reasoning module for analysis, including algorithms such as time series analysis, causal reasoning, and hypothesis testing;

[0091] For example, using LSTM models for trend forecasting, or rule-based logical reasoning systems for hypothesis analysis;

[0092] S5. Sentiment Analysis and Adaptation

[0093] S5.1. Emotion detection model loading: Load the model and algorithm required by the emotion detection unit in the emotion analysis adaptation module;

[0094] S5.2. Sentiment tendency identification: Detect the sentiment of user questions and identify the sentiment tendency (such as positive, negative, neutral, etc.), including sentiment classification algorithms for texts, such as dictionary-based methods, machine learning models, etc.

[0095] S5.3. Answer strategy adjustment: According to the emotion detection results, the answer strategy is dynamically adjusted through the emotion adaptation unit;

[0096] For example, for users with negative emotions, the answers are more comforting and encouraging; for users with positive emotions, the answers are more relaxed and humorous;

[0097] S6. Response Generation

[0098] S6.1. Large language model loading: Load the large language model required by the response generation unit (such as GPT, BERT generation mode, etc.);

[0099] S6.2. Answer generation: Generate natural and fluent answers based on the large language model, including understanding of the question, grasp of the context, and language generation;

[0100] For example, the GPT model gradually generates each word of the answer through autoregression;

[0101] S6.3. Answer optimization and verification: Optimize and verify the generated answers to ensure that they are grammatically correct, semantically clear, and meet user expectations, including post-processing, error correction, and quality control of the answers;

[0102] S7. Cross-language processing

[0103] S7.1. Language identification model loading: Load the model and algorithm required by the language identification unit in the cross-language processing module;

[0104] S7.2. Language type identification: Automatically identify the language type of user input, including language classification algorithms for text, such as methods based on character sets, lexical features, or machine learning models;

[0105] S7.3. Translation and Adaptation: For non-native language input, translation and adaptation are performed through the translation adaptation unit, including using machine translation algorithms to translate non-native language questions into a language supported by the system, and adjusting and optimizing according to the habits of the target language;

[0106] S8. Virtual Reality Interaction

[0107] S8.1. Virtual reality scene construction: Call the scene construction unit in the virtual reality interaction module to build a virtual reality scene according to the needs of the problem, including 3D modeling, scene rendering and interaction logic design;

[0108] S8.2. Interaction control logic setting: Setting the user's interaction control logic in the virtual reality environment, including defining the user's operation mode, the response of the interactive object, and the scene change rules;

[0109] S8.3. Virtual reality experience provision: Provide the constructed virtual reality scene and interactive control logic to users, allowing users to explore and interact in the scene, which is achieved through VR devices or software platforms;

[0110] S9. Time travel simulation

[0111] S9.1. Historical data integration and loading: Load the data and algorithms required by the historical data integration unit in the time travel simulation module;

[0112] Integrate and process data to support temporal simulation;

[0113] S9.2. Time simulation algorithm selection: Select appropriate time simulation algorithms based on the needs of the problem, including analysis of historical data, trend prediction, and construction of hypothetical scenarios;

[0114] S9.3. Simulation and Reasoning: Perform simulation and reasoning based on historical data and selected time simulation algorithms, including simulating social and cultural states at specific points in time and answering hypothetical questions;

[0115] For example, using Monte Carlo simulation to simulate the outcomes of historical events;

[0116] S10. Output feedback

[0117] S10.1. Answer output: Return the generated answer to the user in text, voice or other forms, including the formatting of the answer, the selection of presentation method and the control of the output device;

[0118] S10.2. User feedback collection: Collect user feedback on answers through the user interface or other channels, including user comments, suggestions, questions, etc.;

[0119] Feedback is provided in the form of text, voice, rating, etc.

[0120] S10.3. Subsequent optimization and improvement: The system will be optimized and improved based on the collected user feedback, including analyzing feedback data, identifying problems, adjusting algorithms and models, and updating system functions and interfaces. Through continuous optimization and improvement, the system performance and user experience will be improved.

[0121] In summary, the present invention can accurately understand the intention and key information of user questions through deep semantic analysis of a large language model, and can give accurate answers even to complex questions and questions with diverse expressions. It can dynamically assign the most appropriate intelligent agent to answer according to the type and field of the question, and multiple intelligent agents can also perform necessary communication and collaboration to ensure the flow of information and consistency of answers, build a cross-domain knowledge graph, and provide rich knowledge support for answers to complex questions. Through graph query and reasoning technology, it can accurately retrieve and present relevant information, perform sentiment detection on user questions, and dynamically adjust the tone and style of answers according to sentiment tendencies, thereby improving user satisfaction and experience, supporting multilingual input and output, and realizing language recognition and translation adaptation through a cross-language processing module to ensure that questions in different languages ​​can be understood and answered. For questions that require virtual reality to assist in understanding or involve historical hypothetical questions, it can provide virtual reality interaction and time travel simulation functions to enhance user understanding and experience.

[0122] Although 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 alterations may be made to the embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A multi-agent question-answering system based on a large language model, characterized by: It includes language understanding generation module, multi-agent coordination module, knowledge graph retrieval module, sentiment analysis adaptation module, virtual reality interaction module, time travel simulation module, cross-language processing module, and future prediction reasoning module; The language understanding generation module is used for semantic analysis and response generation; The multi-agent coordination module is used for agent allocation and agent interaction; The knowledge graph retrieval module is used for graph construction and graph query; The sentiment analysis adaptation module is used for sentiment detection and sentiment adaptation; The virtual reality interaction module is used for scene construction and interactive control; The time travel simulation module is used for historical data integration and time simulation; The cross-language processing module is used for language identification and translation adaptation; The future prediction reasoning module is used for trend analysis and hypothesis reasoning.

2. The multi-agent question-answering system based on a large language model according to claim 1, characterized in that: The language understanding generation module also includes a semantic parsing unit and a response generation unit; The semantic parsing unit uses a pre-trained language model to perform deep semantic parsing on the input question to extract key information and intent; The response generation unit is used to generate a natural and fluent answer based on a large language model.

3. The multi-agent question-answering system based on a large language model according to claim 2, characterized in that: The multi-agent coordination module also includes an agent allocation unit and an agent interaction unit; The agent allocation unit is used to dynamically allocate the most suitable agent to answer the question according to the question type and field; The agent interaction unit is used to manage the communication and collaboration between multiple agents to ensure information flow and consistency.

4. The multi-agent question-answering system based on a large language model according to claim 3, characterized in that: The knowledge graph retrieval module also includes a graph construction unit and a graph query unit; The graph construction unit is used to construct and maintain a cross-domain knowledge graph to support answers to complex questions; The graph query unit is used to retrieve relevant information in the knowledge graph and present it in a structured manner in response to user questions.

5. The multi-agent question-answering system based on a large language model according to claim 4, characterized in that: The emotion analysis adaptation module also includes an emotion detection unit and an emotion adaptation unit; The emotion detection unit is used to identify the emotional tendency in the user's question to adjust the tone and style of the answer; The emotion adaptation unit is used to dynamically adjust the answer strategy according to the emotion detection result to improve user satisfaction.

6. The multi-agent question-answering system based on a large language model according to claim 5, characterized in that: The virtual reality interaction module also includes a scene construction unit and an interaction control unit; The scene construction unit is used to explain complex concepts or processes and construct virtual reality scenes to assist understanding; The interaction control unit is used to allow the user to interact with objects in the virtual reality environment to enhance the experience.

7. The multi-agent question-answering system based on a large language model according to claim 6, characterized in that: The time travel simulation module also includes a historical data integration unit and a time simulation unit; The historical data integration unit is used to collect and integrate historical data to provide a basis for time travel simulation; The time simulation unit is based on historical data and is used to simulate the social and cultural status at a specific time point and answer hypothetical questions.

8. The multi-agent question-answering system based on a large language model according to claim 7, characterized in that: The cross-language processing module also includes a language recognition unit and a translation adaptation unit; The language recognition unit is used to automatically recognize the language type of the user input and supports multi-language input; The translation adaptation unit is used to translate non-native language questions into a language supported by the system and adapt the language habits.

9. The multi-agent question-answering system based on a large language model according to claim 8, characterized in that: The future prediction reasoning module also includes a trend analysis unit and a hypothesis reasoning unit; The trend analysis unit is used to analyze existing data, predict future trends, and provide answers to forward-looking questions; The hypothesis reasoning unit is based on logical reasoning and hypothesis analysis and is used to reason about situations that have not occurred or will not occur.

10. A multi-agent question-answering method based on a large language model, according to any one of claims 1 to 9, characterized in that: The following steps are involved: S1. User input: receiving questions raised by users in natural language, in text, voice or other forms; S2. Language understanding: Use the semantic parsing unit in the language understanding generation module to perform deep semantic analysis on the question and extract key information and intent; S3. Agent allocation and interaction: According to the problem type and domain, the most suitable agent is dynamically allocated through the agent allocation unit in the multi-agent coordination module, and multiple agents communicate and collaborate as necessary through the agent interaction unit; S4. Knowledge retrieval and reasoning: In the knowledge graph retrieval module, the graph query unit is used to retrieve relevant information in the knowledge graph. For problems that require prediction or reasoning, the future prediction and reasoning module is used for analysis. S5. Sentiment analysis and adaptation: The sentiment analysis and adaptation module detects the sentiment of user questions and adjusts the tone and style of the answer based on the results; S6.Response generation: Generate natural and fluent responses in the response generation unit based on the large language model; S7. Cross-language processing: For non-native language input, the cross-language processing module performs language recognition and translation adaptation to ensure that the system can understand and answer; S8. Virtual reality interaction: For problems that require virtual reality to assist understanding, use the virtual reality interaction module to build scenes and provide interactive experience; S9. Time Travel Simulation: For answers to historical or hypothetical questions, use the Time Travel Simulation module to simulate and reason; S10. Output feedback: Return the generated answer to the user in the form of text, voice or other forms, and collect user feedback for subsequent optimization.

Images