Bionic multi-brain module expert thinking link agent system

Abstract

The invention discloses a bionic multi-brain module expert thinking link agent system which comprises N dialogue AI modules, N supervision AI modules, N path decision AI modules, a historical dialogue recording module, a long memory AI module and a work memory AI module. The dialogue AI module is used for carrying out direct dialogue with the user and generating a corresponding answer according to information input by the user; the historical dialogue recording module is used for independently storing historical dialogue records of the dialogue AI modules; the long memory AI module is used for storing accumulated dialogue records of all dialogue AI modules; the work memory AI module is used for storing key dialogue information and context information of the current dialogue AI module and outputting a reasoning conclusion; the supervision AI module is used for providing supervision suggestions for the dialogue AI module according to the real-time dialogue content of the dialogue AI module; and the path decision AI module is used for deciding which dialogue AI module continues to communicate with the user according to the historical dialogue record of the dialogue AI module. According to the method, the system processing capability, the dialogue continuity, the deep reasoning capability and the user experience can be improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and more specifically to a bionic multi-brain module expert thinking link intelligent body system. Background Art

[0002] Existing artificial intelligence systems mainly rely on a single AI model to interact with users. This architecture has been widely adopted in various fields in practical applications, including but not limited to:

[0003] Customer service system: used to handle users' common problems, provide information query, solve simple after-sales service problems, etc.

[0004] Virtual assistants: such as those in smart speakers (Google Assistant, Siri, etc.), used to perform daily tasks, control smart home devices, etc.

[0005] Online education: used to provide personalized learning guidance, answer questions, etc.

[0006] Medical consultation: used to provide preliminary medical advice, appointment services, etc.

[0007] This type of AI system based on a single model is relatively simple to implement and has relatively low deployment and maintenance costs, but it has problems such as limited processing power, poor conversational coherence, low deep reasoning capabilities, and poor user experience.

[0008] Therefore, how to provide an artificial intelligence system in which multiple AI modules collaborate to improve system processing capabilities, conversation coherence, deep reasoning capabilities, and user experience is an issue that technical personnel in this field urgently need to solve. Summary of the invention

[0009] In view of this, the purpose of the present invention is to provide a bionic multi-brain module expert thinking link intelligent agent system.

[0010] In order to achieve the above object, the present invention adopts the following technical solution:

[0011] A bionic multi-brain module expert thinking link intelligent agent system, comprising N dialogue AI modules, N supervision AI modules, N path decision AI modules, a historical dialogue record module, a long memory AI module and a working memory AI module; wherein N is greater than or equal to 3;

[0012] The conversational AI module is used to have direct conversations with users and generate corresponding responses based on user input information;

[0013] The historical conversation record module is used to independently store the historical conversation records of each conversation AI module;

[0014] The long memory AI module is used to store the accumulated conversation records of all conversation AI modules;

[0015] The working memory AI module is used to store the key dialogue information and context information of the current dialogue AI module and output the reasoning conclusion;

[0016] The supervisory AI module is used to provide supervisory opinions to the dialogue AI module based on the real-time dialogue content of the dialogue AI module;

[0017] The path decision AI module is used to decide which dialogue AI module will continue to communicate with the user based on the historical dialogue records of the dialogue AI module.

[0018] Preferably, each dialogue AI module includes an AI perception module, an AI execution module, a dialogue display interface and a dialogue expert AI module;

[0019] The AI ​​perception module is used to perceive and understand user input information, and send the user input information to the dialogue display interface for display;

[0020] The dialogue expert AI module is used to generate corresponding answers according to the user input information;

[0021] The AI ​​execution module is used to send the corresponding answer generated by the dialogue expert AI module to the dialogue interaction interface in a certain form for display.

[0022] Preferably, the AI ​​perception module includes a graphic perception unit, an audio and video perception unit, and an interactive perception unit; the AI ​​execution module includes a graphic output unit, an audio and video output unit, and an interactive output unit.

[0023] Preferably, the long memory AI module includes a first background setting unit, a conversation record accumulation unit, a long memory rule unit and a first relevant information extraction unit;

[0024] The first background setting unit includes pre-embedded information; the pre-embedded information includes background information, role settings, core tasks and knowledge base;

[0025] The accumulated dialogue record unit is used to store the accumulated dialogue records of all dialogue AI modules;

[0026] The long memory rule unit is used to store long memory rules, and the long memory rules are: extracting first relevant information according to the first background setting unit, the accumulated conversation records and a preset target; wherein the preset target is a target embedded in advance;

[0027] The first relevant information extraction unit is used to extract first relevant information according to the first background setting unit, the accumulated conversation records and the long memory rule.

[0028] Preferably, the working memory AI module includes a second background setting unit, a key dialogue recording unit, a working memory rule unit and a reasoning analysis unit;

[0029] The second background setting unit includes pre-embedded information; the pre-embedded information includes background information, role settings, core tasks and knowledge base;

[0030] The key dialogue recording unit is used to store the key dialogue information and context information of the current dialogue AI module;

[0031] The working memory rule unit is used to store working memory rules, and the working memory rules are: extracting second relevant information according to the information stored in the key conversation record unit and a preset target; wherein the preset target is a target embedded in advance;

[0032] The reasoning analysis unit is used to perform reasoning and give reasoning conclusions based on the second background setting unit, the information stored in the key dialogue recording unit and the working memory rules.

[0033] Preferably, the dialogue expert AI module includes a third background setting unit, a relevant information unit, a first current target unit, a working memory unit, a first implementation step unit, a dialogue rule unit, a short-term memory unit, a supervisory opinion unit, and an answer output unit;

[0034] The third background setting unit includes pre-embedded information; the pre-embedded information includes background information, role settings, core tasks and knowledge base;

[0035] The relevant information unit is used to synchronously store the first relevant information extracted by the first relevant information extraction unit;

[0036] The first current target unit is used to decompose the preset target into current small targets; wherein the preset target is a target embedded in advance;

[0037] The working memory unit is used to store the reasoning conclusions obtained by the reasoning analysis unit;

[0038] The first implementation step unit is used to store the implementation steps of the current small goal;

[0039] The dialogue rule unit is used to store the dialogue rules of the current dialogue expert AI module;

[0040] The short-term memory unit is used to store the most recent rounds of historical dialogue records of the current dialogue AI module;

[0041] The supervision opinion unit is used to store the supervision opinions provided by the supervision AI module;

[0042] The answer output unit is used to integrate the information stored in the third background setting unit, the relevant information unit, the first current target unit, the working memory unit, the first implementation step unit, the dialogue rule unit, the short-term memory unit and the supervisory opinion unit, and output an answer corresponding to the user input information.

[0043] Preferably, the supervision AI module includes a fourth background setting unit, a first historical dialogue recording unit, a second current target unit, a second implementation step unit, a supervision rule unit, and a supervision opinion output unit;

[0044] The fourth background setting unit includes pre-embedded information; the pre-embedded information includes background information, role settings, core tasks and knowledge base;

[0045] The first historical dialogue recording unit is used to store the real-time dialogue content of the current dialogue AI module;

[0046] The second current target unit is used to synchronously store the current small target obtained by disassembling the first current target unit;

[0047] The second implementation step unit is used to synchronously store the implementation steps stored by the first implementation step unit;

[0048] The supervision rule unit is used to store supervision rules, and the supervision rules are: generating supervision opinions according to the current small goal stored in the second current goal unit and the implementation steps stored in the second implementation step unit;

[0049] The supervisory opinion output unit is used to output supervisory opinions by integrating the information stored in the fourth background setting unit, the first historical dialogue record unit, the second current target unit, the second implementation step unit and the supervisory rule unit.

[0050] Preferably, the path decision AI module includes a second history dialogue recording unit, a fifth background setting unit, a path decision rule unit, and a path decision result output unit;

[0051] The second historical dialogue record unit is used to store the historical dialogue record of the current dialogue AI module;

[0052] The fifth background setting unit includes pre-embedded information; the pre-embedded information includes background information, role settings, core tasks and knowledge base;

[0053] The path decision rule unit is used to store the path decision rule; the path decision rule is determined according to actual needs;

[0054] The path decision result output unit is used to comprehensively store the information of the second historical dialogue record unit, the fifth background setting unit and the path decision rule unit to output a path decision result, and the path decision result is: which dialogue AI module will continue to communicate with the user next.

[0055] Preferably, the above-mentioned artificial intelligence system also includes an AI self-learning module;

[0056] The AI ​​self-learning module is used to perform self-learning using a dynamically updated data knowledge base, and then use the self-learned AI self-learning module to update the knowledge base stored in the third background setting unit, the current small goal stored in the first current goal unit, the implementation steps stored in the first implementation step unit, and the dialogue rules stored in the dialogue rule unit.

[0057] Preferably, the dynamically updated data knowledge base includes an external knowledge database, a public domain knowledge database and a self-precipitated knowledge database;

[0058] The self-precipitated knowledge database is extracted from the historical dialogue records of each dialogue AI module.

[0059] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a bionic multi-brain module expert thinking link intelligent agent system, which can improve the system processing capability, dialogue coherence, deep reasoning capability and user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] 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 embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0061] Figure 1 A structural block diagram of a bionic multi-brain module expert thinking link intelligent agent system provided by the present invention;

[0062] Figure 2 A connection diagram of the internal units of the conversational AI module provided by the present invention;

[0063] Figure 3 A connection relationship diagram of the long memory AI module, the working memory AI module, the dialogue expert AI module, and the historical dialogue record module provided by the present invention;

[0064] Figure 4 A connection diagram of the supervisor AI module and the dialogue expert AI module provided by the present invention;

[0065] Figure 5 A connection diagram of the path decision AI module and the historical conversation record module provided by the present invention;

[0066] Figure 6 A connection diagram of the AI ​​self-learning module, the conversation expert AI module, the dynamically updated data knowledge base, and the historical conversation record module provided by the present invention. DETAILED DESCRIPTION

[0067] 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.

[0068] like Figure 1 As shown, the embodiment of the present invention discloses a bionic multi-brain module expert thinking link intelligent agent system, including N dialogue AI modules, N supervision AI modules, N path decision AI modules, a historical dialogue record module, a long memory AI module and a working memory AI module; wherein N is greater than or equal to 3;

[0069] The conversational AI module is used to have direct conversations with users and generate corresponding responses based on user input information;

[0070] It is understandable that the conversational AI module can be a professional service role, such as health expert, psychological counselor, tax consultant, marketing expert, customer support, etc., and different conversational AI roles can be formulated according to business scenario requirements.

[0071] Specifically, the expert thinking chain is reflected in the following: experts in various fields first break down the workflow, and then different conversational AIs are responsible for their respective process stages. This division of labor mechanism ensures that the system can efficiently handle complex problems and provide accurate and professional responses through gradually in-depth interactions.

[0072] The historical conversation record module is used to independently store the historical conversation records of each conversation AI module;

[0073] The long memory AI module is used to store the accumulated conversation records of all conversation AI modules;

[0074] The working memory AI module is used to store the key dialogue information and context information of the current dialogue AI module and output the reasoning conclusion;

[0075] It can be understood that: the information stored in the long memory AI module and the working memory AI module are obtained from the historical dialogue record module;

[0076] The historical dialogue record module includes N independent storage units for storing historical dialogue records of N dialogue AI modules respectively;

[0077] In the present invention: the long memory AI module is equivalent to the long memory function of the human brain;

[0078] The working memory AI module is equivalent to the key memory function of the human brain;

[0079] The supervisory AI module is used to provide supervisory opinions to the dialogue AI module according to the real-time dialogue content of the dialogue AI module; then, the dialogue AI module adjusts the answer according to the supervisory opinions provided by the supervisory AI module;

[0080] The path decision AI module is used to decide which dialogue AI module will continue to communicate with the user based on the historical dialogue records of the dialogue AI module.

[0081] Further, such as Figure 2 As shown, each dialogue AI module includes an AI perception module, an AI execution module, a dialogue display interface and a dialogue expert AI module;

[0082] The AI ​​perception module is used to perceive and understand user input information, and send the user input information to the dialogue display interface for display;

[0083] The dialogue expert AI module is used to generate corresponding answers according to the user input information;

[0084] The AI ​​execution module is used to send the corresponding answer generated by the dialogue expert AI module to the dialogue interaction interface in a certain form for display.

[0085] It can be understood that: taking the user input information as voice as an example, the AI ​​execution module can send it to the dialogue interaction interface for display in the form of voice, text, etc.

[0086] Furthermore, the AI ​​perception module includes a graphic perception unit, an audio and video perception unit, and an interactive perception unit; the AI ​​execution module includes a graphic output unit, an audio and video output unit, and an interactive output unit.

[0087] It can be understood that: the image and text perception unit is used to collect the image and text information input by the user and process it into structured data;

[0088] The audio and video perception unit is used to collect audio and video information input by the user and process it into structured data;

[0089] The interactive perception unit is used to collect data in interactive forms such as questionnaires, 3D modeling, gestures, etc. and process them into structured processing;

[0090] The graphic output unit is used to output the answer in graphic form to the user;

[0091] The audio and video output unit is used to output the answer in audio and video form to the user;

[0092] The interactive output unit is used to output answers in interactive forms such as questionnaires, 3D modeling, gestures, etc. to the user.

[0093] Further, such as Figure 3 As shown, the long memory AI module includes a first background setting unit, a conversation record accumulation unit, a long memory rule unit and a first related information extraction unit;

[0094] The first background setting unit includes pre-embedded information; the pre-embedded information includes background information, role settings, core tasks and knowledge base;

[0095] The accumulated dialogue record unit is used to store the accumulated dialogue records of all dialogue AI modules;

[0096] The long memory rule unit is used to store long memory rules, and the long memory rules are: extracting first relevant information according to the first background setting unit, the accumulated conversation records and a preset target; wherein the preset target is a target embedded in advance;

[0097] The first relevant information extraction unit is used to extract first relevant information according to the first background setting unit, the accumulated conversation records and the long memory rule.

[0098] It can be understood that: the first relevant information includes user information, conversation stage and conversation progress, etc.;

[0099] It is understandable that the long-memory AI module is responsible for storing detailed records of all conversations between users and the system, including all questions, answers, contextual information, etc. These records can span multiple conversations and cover long-term user interaction history. By analyzing and summarizing long-term conversation records, the long-memory AI module can infer users' long-term interests and trends, and provide valuable background information and decision support for future conversations.

[0100] The long-memory AI module cumulatively summarizes all conversation records between the user and the system to form a comprehensive and detailed long-term memory. These summaries not only contain specific conversation content, but also include information such as the user's intentions, feedback, and habits. In future conversations, the long memory provided by the long-memory AI module can be used as a reference to ensure that the system can provide coherent and relevant responses based on past interactions, avoid information loss and repeated questions, and improve user experience. For example, the system can remember topics and questions that users have discussed before, and actively mention relevant information in future conversations.

[0101] Further, the working memory AI module includes a second background setting unit, a key dialogue recording unit, a working memory rule unit and a reasoning analysis unit;

[0102] The second background setting unit includes pre-embedded information; the pre-embedded information includes background information, role settings, core tasks and knowledge base;

[0103] The key dialogue recording unit is used to store the key dialogue information and context information of the current dialogue AI module;

[0104] The working memory rule unit is used to store working memory rules, and the working memory rules are: extracting second relevant information according to the information stored in the key conversation record unit and a preset target; wherein the preset target is a target embedded in advance;

[0105] The reasoning analysis unit is used to perform reasoning and give reasoning conclusions based on the second background setting unit, the information stored in the key dialogue recording unit and the working memory rules.

[0106] It can be understood that: the second relevant information includes user information, conversation stage and conversation progress, etc.;

[0107] The working memory AI module can infer the user's potential needs and intentions and form conclusions by analyzing and understanding the current conversation information. The conclusions are embedded in the prompt words of the specified conversation AI module to help the system provide more accurate services and support.

[0108] Further, the dialogue expert AI module includes a third background setting unit, a relevant information unit, a first current target unit, a working memory unit, a first implementation step unit, a dialogue rule unit, a short-term memory unit, a supervisory opinion unit, and an answer output unit;

[0109] The third background setting unit includes pre-embedded information; the pre-embedded information includes background information, role settings, core tasks and knowledge base;

[0110] The relevant information unit is used to synchronously store the first relevant information extracted by the first relevant information extraction unit;

[0111] The first current target unit is used to decompose the preset target into current small targets; wherein the preset target is a target embedded in advance;

[0112] The working memory unit is used to store the reasoning conclusions obtained by the reasoning analysis unit;

[0113] The first implementation step unit is used to store the implementation steps of the current small goal;

[0114] The dialogue rule unit is used to store the dialogue rules of the current dialogue expert AI module;

[0115] It is understandable that: dialogue rules include word limit, dialogue style, precautions...;

[0116] The short-term memory unit is used to store the most recent rounds of historical dialogue records of the current dialogue AI module;

[0117] It can be understood that the most recent several rounds of historical dialogue records of the current dialogue AI module are obtained from the historical dialogue records of the current dialogue AI module; specifically, the short-term memory unit continuously stores the most recent 6 rounds of historical dialogue records. After updating, it is implanted into the prompt words of the next dialogue AI module to maintain the continuity of the dialogue.

[0118] In the present invention, the short-term memory unit is equivalent to the short-term memory function of the human brain;

[0119] The supervision opinion unit is used to store the supervision opinions provided by the supervision AI module;

[0120] The answer output unit is used to integrate the information stored in the third background setting unit, the relevant information unit, the first current target unit, the working memory unit, the first implementation step unit, the dialogue rule unit, the short-term memory unit and the supervisory opinion unit, and output an answer corresponding to the user input information.

[0121] Further, such as Figure 4 As shown, the supervision AI module includes a fourth background setting unit, a first historical dialogue record unit, a second current target unit, a second implementation step unit, a supervision rule unit, and a supervision opinion output unit;

[0122] The fourth background setting unit includes pre-embedded information; the pre-embedded information includes background information, role settings, core tasks and knowledge base;

[0123] The first historical dialogue recording unit is used to store the real-time dialogue content of the current dialogue AI module;

[0124] The second current target unit is used to synchronously store the current small target obtained by disassembling the first current target unit;

[0125] The second implementation step unit is used to synchronously store the implementation steps stored by the first implementation step unit;

[0126] The supervision rule unit is used to store supervision rules, and the supervision rules are: generating supervision opinions according to the current small goal stored in the second current goal unit and the implementation steps stored in the second implementation step unit;

[0127] The supervisory opinion output unit is used to output supervisory opinions by integrating the information stored in the fourth background setting unit, the first historical dialogue record unit, the second current target unit, the second implementation step unit and the supervisory rule unit.

[0128] It is understandable that the supervisory AI module is specifically designed to monitor, guide and optimize the conversation process and content to ensure the efficiency, accuracy and compliance of interactions with users.

[0129] The supervisory AI module analyzes the conversation records between the dialogue AI module and the user, detects and analyzes whether the conversation deviates from the topic. Based on the analysis results, the supervisory AI module can return the conversation to the original process through a fixed dialogue mechanism.

[0130] Further, such as Figure 5 As shown, the path decision AI module includes a second history dialogue recording unit, a fifth background setting unit, a path decision rule unit, and a path decision result output unit;

[0131] The second historical dialogue record unit is used to store the historical dialogue record of the current dialogue AI module;

[0132] The fifth background setting unit includes pre-embedded information; the pre-embedded information includes background information, role settings, core tasks and knowledge base;

[0133] The path decision rule unit is used to store path decision rules; the path decision rules are determined according to actual needs; that is, different dialogue scenarios correspond to different path decision rules, and the path decision rules are rules embedded in advance.

[0134] The path decision result output unit is used to comprehensively store the information of the second historical dialogue record unit, the fifth background setting unit and the path decision rule unit to output a path decision result, and the path decision result is: which dialogue AI module will continue to communicate with the user next.

[0135] It is understandable that the path decision AI module selects the most appropriate dialogue AI module to continue the conversation based on the conversation content of the dialogue AI module and user needs.

[0136] The system uses the dialogue AI module-specific identifier returned by the path decision AI module to switch prompt words to ensure dialogue continuity.

[0137] Further, such as Figure 6 As shown, the above artificial intelligence system also includes an AI self-learning module;

[0138] The AI ​​self-learning module is used to perform self-learning using a dynamically updated data knowledge base, and then use the self-learned AI self-learning module to update the knowledge base stored in the third background setting unit, the current small goal stored in the first current goal unit, the implementation steps stored in the first implementation step unit, and the dialogue rules stored in the dialogue rule unit.

[0139] It can be understood that: the AI ​​self-learning module includes prompt word generation, professional knowledge base, vertical category model and AI learning;

[0140] in:

[0141] Prompt word generation: AI generates optimized prompt words based on conversation data;

[0142] Professional knowledge base: A vertical knowledge base is formed based on external knowledge database and internal input knowledge

[0143] Vertical model: Train and fine-tune the vertical model based on vertical conversation and feedback data;

[0144] AI learning: building and training based on internal and external data;

[0145] Furthermore, the dynamically updated data knowledge base includes an external knowledge database, a public domain knowledge database and a self-precipitated knowledge database;

[0146] The self-precipitated knowledge database is extracted from the historical dialogue records of each dialogue AI module.

[0147] The public domain knowledge database refers to the public knowledge and data on the Internet;

[0148] The external knowledge database is knowledge and data provided by experts.

[0149] Finally, some alternatives of the present invention are given:

[0150] 1: Alternatives to long-memory AI modules

[0151] 1) Distributed Memory System

[0152] Technical means: Distribute the long memory function on multiple storage nodes, and each node is responsible for storing the memory information of different users and conversations. The system integrates the memory information of each node through a distributed algorithm to ensure the continuity of the conversation.

[0153] Pros and cons comparison

[0154] characteristic Distributed Memory Network (DMN) Long Memory AI Module Context handling Strong, able to handle complex contexts and long-term dependencies In the , mainly rely on static storage and retrieval Dynamic Updates Strong, able to dynamically adapt to the conversational context Weak, mainly relying on pre-stored memory data Implementation complexity High, requires expertise and computing resources Low, relatively easy to implement and maintain Computing resource consumption High, requiring a lot of resources during training and inference Low, less resource requirements Data Dependency High, requires a lot of high-quality training data In the case of long memory, it relies on pre-stored long memory data flexibility Strong, adaptable to a variety of tasks and scenarios Medium, relatively limited flexibility Applicable scenarios Question answering system, sentiment analysis, machine translation, etc. Customer support, medical consultation, education, etc.

[0155] Use a distributed memory network (DMN): if the system needs to handle complex contexts and dynamically changing conversation situations, and has sufficient computing resources and data support.

[0156] Use long memory AI modules: If the system needs to maintain a long conversation history and knowledge, and has high requirements on resources and implementation complexity.

[0157] 2. Alternatives to the Supervisory AI Module

[0158] 1) Rule-Based Engine

[0159] Description: Use predefined rules and logic to monitor and moderate conversations.

[0160] Advantages: Simple to implement, easy to understand and maintain, suitable for fixed and clear dialogue flow.

[0161] Application scenarios: Suitable for customer service systems and appointment systems with clear processes.

[0162] 2) Template Matching System

[0163] Description: Use conversation templates and pattern matching to monitor and moderate conversations.

[0164] Advantages: Quickly matches common conversation patterns, easy to implement and expand.

[0165] Application scenarios: Applicable to FAQ systems, standardized service processes, etc.

[0166] 3) Reinforcement Learning (RL)

[0167] Description: Use reinforcement learning algorithms to train intelligent agents to monitor and control conversations, and continuously optimize conversation strategies based on feedback.

[0168] Advantages: Able to adaptively learn and optimize dialogue strategies and handle complex and dynamically changing dialogue situations.

[0169] Application scenarios: Suitable for complex dialogue systems, such as intelligent assistants, high-end customer service, etc.

[0170] 4) Supervised Learning Models

[0171] Description: Use supervised learning algorithms to train models and monitor and regulate based on historical conversation data.

[0172] Advantages: Able to process large amounts of historical data and provide accurate monitoring and control capabilities.

[0173] Application scenarios: Suitable for data-rich scenarios, such as large customer service centers, online education platforms, etc.

[0174] 5) Dialog Management System (DMS)

[0175] Description: Use a specialized dialog management framework and tools to monitor and manage the dialog flow.

[0176] Advantages: Provides complete dialogue management functions and supports complex dialogue logic and multi-round dialogues.

[0177] Application scenarios: Applicable to systems that require complex dialogue management, such as virtual assistants, voice interaction systems, etc.

[0178] Advantages and disadvantages comparison

[0179]

[0180] Rule Engine: If the conversation flow is fixed and clear, the rule engine is a simple and efficient choice.

[0181] Template matching system: suitable for scenarios with many common dialogue patterns and little variation.

[0182] Reinforcement learning: Suitable for systems with complex dialogue scenarios and requiring adaptive optimization, but has high implementation and training requirements.

[0183] Supervised learning model: Applicable to scenarios with large amounts of historical data, and can be used for accurate monitoring by learning from historical data.

[0184] Dialogue Management System: Suitable for systems that require complex dialogue management, providing comprehensive dialogue monitoring and management capabilities.

[0185] Supervisory AI module: Suitable for complex conversations and multi-tasking scenarios, providing high flexibility and adaptability to ensure the consistency and effectiveness of conversations.

[0186] 3. Alternatives to the Path Decision AI Module

[0187] 1) State Machine

[0188] Description: Manage the conversation flow and determine the conversation path by defining states and state transition rules.

[0189] Advantages: Clear structure, easy to implement and debug, suitable for clear dialogue state flow.

[0190] Disadvantages: Limited support for complex dialogue logic and not suitable for handling complex or dynamically changing dialogues.

[0191] 2) Decision Tree

[0192] Description: Determine the dialogue path and next action through a series of conditions and branches.

[0193] Advantages: Intuitive and easy to understand, suitable for processing complex decision logic.

[0194] Disadvantages: Decision trees can become large and unmanageable as the number of conditions and branches increases.

[0195] 3) Bayesian Network

[0196] Description: Use probabilistic models to decide dialogue paths, taking into account uncertainty and dependencies.

[0197] Advantages: Ability to handle uncertainty and provide probabilistic decision making.

[0198] Disadvantages: Complex to implement and understand, requires professional knowledge.

[0199] 4) Scheduler

[0200] Description: Dynamically allocate dialog tasks and paths based on task priority and resource availability.

[0201] Pros: Good for multitasking and dynamic environments.

[0202] Disadvantages: Complex to implement and maintain, and depends on an effective scheduling algorithm.

[0203] 5) Hybrid System

[0204] Description: Combines multiple technologies such as rule engine, state machine, decision tree, etc. to provide flexible path decision solutions

[0205] Advantages: Ability to comprehensively utilize multiple technical advantages and provide high flexibility and adaptability.

[0206] Disadvantages: Complex implementation, requiring effective integration of multiple technologies.

[0207] Advantages and disadvantages comparison

[0208]

[0209] State machine: suitable for finite-state dialogue process management and application scenarios of moderate complexity.

[0210] Decision tree: Suitable for complex decision logic and conditional branching requirements, and can handle more complex dialogue processes.

[0211] Bayesian Network: Suitable for scenarios that need to deal with uncertainty and dependencies.

[0212] Scheduler: Suitable for multi-tasking and dynamic environment scenarios.

[0213] Hybrid system: Suitable for scenarios that require the comprehensive use of multiple technical advantages to handle complex conversations and multi-tasking.

[0214] Path decision AI module: Suitable for complex conversations and multi-tasking scenarios, providing high flexibility and adaptability to ensure conversation continuity and user experience.

[0215] 4. Alternatives to Short-Term Memory Units

[0216] 1) Buffer Mechanism

[0217] Description: Use a buffer to save the contents of the last few rounds of conversations, and read and update them as needed.

[0218] Advantages: Simple implementation, easy to adjust buffer size.

[0219] Disadvantages: For complex dialogue scenarios, buffer management may be insufficient to provide advanced analysis and memory capabilities.

[0220] 2) Sliding Window Mechanism

[0221] Description: Use a sliding window to save a fixed number of recent conversations, and the window moves with the new conversation content.

[0222] Advantages: Flexibly adapt to changes in conversation length and maintain conversation continuity.

[0223] Disadvantages: The implementation is complex and requires effective management of window movement and data updates.

[0224] 3) Circular Buffer

[0225] Description: Use a circular buffer to save the most recent conversation content. When the buffer is full, the oldest data will be overwritten.

[0226] Advantages: Efficient use of memory resources, suitable for fixed-length conversation history.

[0227] Disadvantages: Limited ability to manage complex conversation histories, and may lose important contextual information.

[0228] Advantages and disadvantages comparison

[0229]

[0230] Buffer mechanism: suitable for simple dialogue management needs, simple to implement and easy to adjust.

[0231] Sliding window mechanism: suitable for complex dialogue management, flexibly adapting to changes in dialogue length, and maintaining dialogue continuity.

[0232] Ring buffer: Suitable for fixed-length conversation history management and efficient use of memory resources.

[0233] Short-term memory unit: Applicable to scenarios of coherent dialogue management, ensuring the coherence and contextual consistency of the dialogue in a short to medium time frame.

[0234] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0235] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A bionic multi-brain module expert thinking link intelligent agent system, characterized in that: It includes N dialogue AI modules, N supervision AI modules, N path decision AI modules, historical dialogue record module, long memory AI module and working memory AI module; where N is greater than or equal to 3; The conversational AI module is used to have direct conversations with users and generate corresponding responses based on user input information; The historical conversation record module is used to independently store the historical conversation records of each conversation AI module; The long memory AI module is used to store the accumulated conversation records of all conversation AI modules; The working memory AI module is used to store the key dialogue information and context information of the current dialogue AI module and output the reasoning conclusion; The supervisory AI module is used to provide supervisory opinions to the dialogue AI module based on the real-time dialogue content of the dialogue AI module; The path decision AI module is used to decide which dialogue AI module will continue to communicate with the user based on the historical dialogue records of the dialogue AI module.

2. According to claim 1, a bionic multi-brain module expert thinking link intelligent agent system is characterized in that: Each conversational AI module includes an AI perception module, an AI execution module, a conversation display interface, and a conversation expert AI module; The AI ​​perception module is used to perceive and understand user input information, and send the user input information to the dialogue display interface for display; The dialogue expert AI module is used to generate corresponding answers according to the user input information; The AI ​​execution module is used to send the corresponding answer generated by the dialogue expert AI module to the dialogue interaction interface in a certain form for display.

3. A bionic multi-brain module expert thinking link intelligent agent system according to claim 2, characterized in that: The AI ​​perception module includes a graphic perception unit, an audio and video perception unit, and an interactive perception unit; the AI ​​execution module includes a graphic output unit, an audio and video output unit, and an interactive output unit.

4. A bionic multi-brain module expert thinking link intelligent agent system according to claim 2 or 3, characterized in that: The long memory AI module includes a first background setting unit, a conversation record accumulation unit, a long memory rule unit and a first relevant information extraction unit; The first background setting unit includes pre-embedded information; the pre-embedded information includes background information, role settings, core tasks and knowledge base; The accumulated dialogue record unit is used to store the accumulated dialogue records of all dialogue AI modules; The long memory rule unit is used to store long memory rules, and the long memory rules are: extracting first relevant information according to the first background setting unit, the accumulated conversation records and a preset target; wherein the preset target is a target embedded in advance; The first relevant information extraction unit is used to extract first relevant information according to the first background setting unit, the accumulated conversation records and the long memory rule.

5. The bionic multi-brain module expert thinking link intelligent agent system according to claim 4 is characterized in that: The working memory AI module includes a second context setting unit, a key dialogue recording unit, a working memory rule unit, and a reasoning analysis unit; The second background setting unit includes pre-embedded information; the pre-embedded information includes background information, role settings, core tasks and knowledge base; The key dialogue recording unit is used to store the key dialogue information and context information of the current dialogue AI module; The working memory rule unit is used to store working memory rules, and the working memory rules are: extracting second relevant information according to the information stored in the key conversation record unit and a preset target; wherein the preset target is a target embedded in advance; The reasoning analysis unit is used to perform reasoning and give reasoning conclusions based on the second background setting unit, the information stored in the key dialogue recording unit and the working memory rules.

6. A bionic multi-brain module expert thinking link intelligent agent system according to claim 5, characterized in that: The dialogue expert AI module includes a third background setting unit, a relevant information unit, a first current target unit, a working memory unit, a first implementation step unit, a dialogue rule unit, a short-term memory unit, a supervisory opinion unit, and an answer output unit; The third background setting unit includes pre-embedded information; the pre-embedded information includes background information, role settings, core tasks and knowledge base; The relevant information unit is used to synchronously store the first relevant information extracted by the first relevant information extraction unit; The first current target unit is used to decompose the preset target into current small targets; wherein the preset target is a target embedded in advance; The working memory unit is used to store the reasoning conclusions obtained by the reasoning analysis unit; The first implementation step unit is used to store the implementation steps of the current small goal; The dialogue rule unit is used to store the dialogue rules of the current dialogue expert AI module; The short-term memory unit is used to store the most recent rounds of historical dialogue records of the current dialogue AI module; The supervision opinion unit is used to store the supervision opinions provided by the supervision AI module; The answer output unit is used to integrate the information stored in the third background setting unit, the relevant information unit, the first current target unit, the working memory unit, the first implementation step unit, the dialogue rule unit, the short-term memory unit and the supervisory opinion unit, and output an answer corresponding to the user input information.

7. The bionic multi-brain module expert thinking link intelligent agent system according to claim 6 is characterized in that: The supervision AI module includes a fourth background setting unit, a first historical dialogue record unit, a second current target unit, a second implementation step unit, a supervision rule unit, and a supervision opinion output unit; The fourth background setting unit includes pre-embedded information; the pre-embedded information includes background information, role settings, core tasks and knowledge base; The first historical dialogue recording unit is used to store the real-time dialogue content of the current dialogue AI module; The second current target unit is used to synchronously store the current small target obtained by disassembling the first current target unit; The second implementation step unit is used to synchronously store the implementation steps stored by the first implementation step unit; The supervision rule unit is used to store supervision rules, and the supervision rules are: generating supervision opinions according to the current small goal stored in the second current goal unit and the implementation steps stored in the second implementation step unit; The supervisory opinion output unit is used to output supervisory opinions by integrating the information stored in the fourth background setting unit, the first historical dialogue record unit, the second current target unit, the second implementation step unit and the supervisory rule unit.

8. The bionic multi-brain module expert thinking link intelligent agent system according to claim 7 is characterized in that: The path decision AI module includes a second history dialogue recording unit, a fifth background setting unit, a path decision rule unit, and a path decision result output unit; The second historical dialogue record unit is used to store the historical dialogue record of the current dialogue AI module; The fifth background setting unit includes pre-embedded information; the pre-embedded information includes background information, role settings, core tasks and knowledge base; The path decision rule unit is used to store the path decision rule; the path decision rule is determined according to actual needs; The path decision result output unit is used to comprehensively store the information of the second historical dialogue record unit, the fifth background setting unit and the path decision rule unit to output a path decision result, and the path decision result is: which dialogue AI module will continue to communicate with the user next.

9. The bionic multi-brain module expert thinking link intelligent agent system according to claim 6 is characterized in that: It also includes AI self-learning modules; The AI ​​self-learning module is used to perform self-learning using a dynamically updated data knowledge base, and then use the self-learned AI self-learning module to update the knowledge base stored in the third background setting unit, the current small goal stored in the first current goal unit, the implementation steps stored in the first implementation step unit, and the dialogue rules stored in the dialogue rule unit.

10. The bionic multi-brain module expert thinking link intelligent agent system according to claim 9 is characterized in that: The dynamically updated data knowledge base includes an external knowledge database, a public domain knowledge database and a self-precipitated knowledge database; The self-precipitated knowledge database is extracted from the historical dialogue records of each dialogue AI module.

Images