Text generation system and method based on multi-agent cooperation mechanism MOPAR

Abstract

The invention provides a text generation system and method based on a multi-agent cooperation mechanism MOPAR, and the system comprises a matching agent which is connected with a field multi-agent system, and is used for obtaining the writing demands of a user, generating a matching prompt word according to the writing demands, and transmitting the matching prompt word to the field multi-agent system, performing field matching with the field multi-agent system through the matching prompt word, and performing optimization according to feedback information of an reflection agent in the field multi-agent system; and the field multi-agent system comprises a plurality of multi-agents in different fields, is connected with the matched agent and is used for receiving the matched cue word, acquiring the multi-agent in the corresponding field according to the matched cue word and generating text content required by a user through the multi-agent in the corresponding field on the basis of a multi-agent coordination mechanism MOPAR. According to the method, a cooperation mechanism among multiple agents is introduced, so that a more efficient, professional and creative text generation service is provided for the user.

Description

Technical Field

[0001] The present invention relates to the technical field of text generation, and in particular to a text generation system and method based on a multi-agent collaboration mechanism MOPAR. Background Art

[0002] Text generation technology refers to the ability to create natural language text through computer programs. With the development of artificial intelligence technology, especially the rise of deep learning, text generation technology has made significant progress. Early text generation mainly relied on rule-based methods, which usually included predefined templates or grammatical structures to generate text. However, this method has limitations in dealing with complex contexts and diverse content requirements. With the development of deep learning, deep learning models are able to learn deeper language features and generate more fluent and natural text.

[0003] In recent years, the emergence of large language models (LLMs) has greatly promoted the development of text generation technology. These models can learn rich language structures and patterns by training on large-scale text data, thereby generating more natural and fluent text. For example, models based on the Transformer architecture, such as the GPT series of models, have become one of the mainstream technologies in the field of text generation. These models can not only generate general text content, but also complete various language tasks such as translation and question answering.

[0004] AI Agent refers to a highly intelligent software entity that can perform a series of complex tasks and can make decisions autonomously to a certain extent. AI Agents are designed to operate in different environments and complete specific tasks or achieve predetermined goals through self-learning and optimization. After the addition of large model technology, AI Agents have gained stronger language understanding and generation capabilities, which enables them to interact with humans more naturally and handle more complex and diverse tasks. AI Agents in the era of large models not only respond to inputs or instructions passively, but also actively plan actions according to set goals to better adapt to the environment and task requirements; in addition, AI Agents can also use various tools and services to enhance their functions, such as calling external APIs, operating hardware devices, or other software systems. The development of AI Agents represents an important direction in the field of artificial intelligence, that is, the evolution from simple chatbots to intelligent agents that can perform complex tasks.

[0005] Multi-Agent Systems (MAS) are systems composed of multiple interacting agents that can collaborate to complete complex tasks. In text generation technology, multi-agent systems can be used to simulate the interactions between different roles to generate richer and more diverse text content.

[0006] Although existing deep learning text generation technologies have performed well in many applications, they still have some challenges and problems. For example, traditional deep learning text generation algorithms often require a lot of data preparation and model fine-tuning to adapt to specific tasks or fields, which is not only time-consuming and computationally resource-intensive, but may not fully utilize professional knowledge; in addition, some other technologies use a single-agent architecture, which limits the diversity and professionalism of their generated content.

[0007] The main disadvantages of the prior art include but are not limited to:

[0008] 1. Over-reliance on model training and fine-tuning: Traditional text generation algorithms, especially those based on deep learning, often require model training and fine-tuning for specific application scenarios to achieve satisfactory performance. This process is not only time-consuming and labor-intensive, but also requires a large amount of labeled data, increasing development costs and cycles.

[0009] 2. Limitations of professional knowledge: Although large language models can generate highly natural text, due to the lack of training data in specialized fields, general large models do not perform well in generating highly specialized content. For large models dedicated to a certain field, their performance across fields is limited. Large models in the medical field are unlikely to achieve good results in the engineering field. Therefore, the model has limitations in professional knowledge and lacks flexibility, and the model may not be able to accurately provide the required information.

[0010] 3. The uniformity of generated content: When using traditional deep learning models or a single intelligent agent for text generation, the generated content may have a relatively single perspective and lack diversity. This is an obvious shortcoming in text generation scenarios that require analyzing problems from different angles or creating multi-dimensional content.

[0011] 4. Lack of flexibility: Existing text generation technologies are often limited to a single writing model or scenario, and are difficult to quickly adapt to emerging needs or changing environmental conditions. When the application scenario changes, the entire system may need to be redesigned or adjusted, and has poor flexibility.

[0012] 5. Lack of reflection mechanism: Most existing text generation methods lack an effective feedback link after generating content, and are unable to automatically evaluate the quality of the generated text and make necessary corrections, which may result in erroneous or inappropriate information being directly output to users.

[0013] In summary, although current text generation methods have greatly improved the effects of natural language processing, there is still much room for improvement in flexibility, professionalism, diversity, and self-improvement capabilities. Summary of the invention

[0014] The purpose of the present invention is to provide a text generation system and method based on a multi-agent collaboration mechanism MOPAR, aiming to solve the above-mentioned problems in the prior art.

[0015] The embodiment of the present invention provides a text generation system based on a multi-agent collaboration mechanism MOPAR, including:

[0016] A matching agent is connected to the domain multi-agent system to obtain user writing requirements, generate matching prompt words according to the writing requirements, transmit the matching prompt words to the domain multi-agent system, perform domain matching with the domain multi-agent system through the matching prompt words, and optimize according to feedback information of the reflective agent in the domain multi-agent system;

[0017] The domain multi-agent system includes multiple multi-agents in different domains, which are connected to the matching agent, and are used to receive the matching prompt words, and obtain the multi-agent in the corresponding domain according to the matching prompt words, and generate the text content required by the user through the multi-agent in the corresponding domain based on the multi-agent collaboration mechanism MOPAR.

[0018] The embodiment of the present invention provides a text generation method based on a multi-agent collaboration mechanism MOPAR, comprising:

[0019] Acquire user writing requirements through a matching agent, generate matching prompt words according to the writing requirements, transmit the matching prompt words to a domain multi-agent system, perform domain matching with the domain multi-agent system through the matching prompt words, and optimize according to feedback information of a reflective agent in the domain multi-agent system;

[0020] The matching prompt word is received through the domain multi-agent system, and a multi-agent in the corresponding domain is obtained according to the matching prompt word, and the text content required by the user is generated by the multi-agent in the corresponding domain based on the multi-agent collaboration mechanism MOPAR.

[0021] An embodiment of the present invention also provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the steps of the above-mentioned text generation method based on the multi-agent collaboration mechanism MOPAR are implemented.

[0022] An embodiment of the present invention also provides a computer-readable storage medium, on which a program for implementing information transmission is stored. When the program is executed by a processor, the steps of the above-mentioned text generation method based on the multi-agent collaboration mechanism MOPAR are implemented.

[0023] The use of the embodiments of the present invention can include the following beneficial effects: In order to overcome the problems existing in the prior art, the embodiments of the present invention propose a text generation technology based on the MOPAR multi-agent collaboration mechanism using natural language as the expression form, and by introducing a collaboration mechanism among multiple agents, it provides users with more efficient, professional and creative text generation services. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate one or more embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0025] Figure 1 Schematic diagram of a text generation system based on a multi-agent collaboration mechanism MOPAR according to an embodiment of the present invention;

[0026] Figure 2 is a flow chart of a multi-agent based MOPAR collaboration method according to an embodiment of the present invention;

[0027] Figure 3 is a schematic diagram of the relationship between intelligent agents in an embodiment of the present invention;

[0028] Figure 4 It is a flow chart of a text generation method based on a multi-agent collaboration mechanism MOPAR according to an embodiment of the present invention. DETAILED DESCRIPTION

[0029] In order to enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the following will be combined with the drawings in one or more embodiments of this specification to clearly and completely describe the technical solutions in one or more embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this document.

[0030] System Example

[0031] According to an embodiment of the present invention, a text generation system based on a multi-agent collaboration mechanism MOPAR is provided. Figure 1 Schematic diagram of a text generation system based on a multi-agent collaboration mechanism MOPAR according to an embodiment of the present invention. Figure 1 As shown, the text generation system based on the multi-agent collaboration mechanism MOPAR according to an embodiment of the present invention specifically includes:

[0032] A matching agent 10 is connected to the domain multi-agent system, and is used to obtain the user's writing requirements, generate matching prompt words according to the writing requirements, transmit the matching prompt words to the domain multi-agent system, and perform domain matching with the domain multi-agent system through the matching prompt words, and optimize according to the feedback information of the reflective agent in the domain multi-agent system;

[0033] The writing requirements include writing topics, target audiences, and content requirements;

[0034] The matching prompt words include the writing requirements and relevant information of multiple agents in various fields;

[0035] The domain multi-agent system 12 includes a plurality of multi-agents in different domains, connected to the matching agent, for receiving the matching prompt word, acquiring a multi-agent in the corresponding domain according to the matching prompt word, and generating text content required by the user through the multi-agent in the corresponding domain based on the multi-agent collaboration mechanism MOPAR, specifically including:

[0036] An outline agent is connected to the matching agent, the perspective agent and the reflection agent, and is used to receive the writing requirements, generate a writing body structure according to the writing requirements, transmit the writing requirements and the writing body structure to the perspective agent, and optimize according to the feedback information of the reflection agent; wherein the writing body structure includes several levels of writing outlines;

[0037] A perspective agent, connected to the outline agent, the retrieval agent and the reflection agent, for receiving the writing requirements and the writing subject structure, generating several types of writing perspectives for each level of the writing outline based on the writing requirements, transmitting the writing subject structure and the several types of writing perspectives to the retrieval agent, and optimizing according to the feedback information of the reflection agent;

[0038] A search agent, connected to the perspective agent, the summary agent and the reflection agent, for receiving the writing subject structure and the several types of writing perspectives, generating corresponding search statements for each writing perspective of each level of the writing outline according to the writing subject structure and the several types of writing perspectives, calling a search engine to retrieve professional knowledge in related fields through the search statements to obtain search results, transmitting the writing subject structure, the several types of writing perspectives and the search results to the summary agent, and optimizing according to feedback information from the reflection agent;

[0039] A summarizing agent, connected to the retrieval agent and the generating agent, for receiving the writing subject structure, the several types of writing perspectives and the retrieval results, summarizing each writing perspective and its corresponding professional knowledge of each level of the writing outline and eliminating redundant content to obtain a comprehensive summary content, and transmitting the comprehensive summary content to the generating agent;

[0040] A generating agent connected to the summarizing agent and the reflecting agent, configured to receive the comprehensive summary content, generate text according to the comprehensive summary content, obtain initial text content, transmit the initial text content to the reflecting agent, and optimize according to the feedback information of the reflecting agent;

[0041] A reflective agent is connected with the matching agent, the outline agent, the perspective agent, the retrieval agent and the generating agent, and is used to receive the writing requirements and the initial text content, review the initial text content according to the writing requirements, form corresponding feedback information based on the review results, and transmit the feedback information to the matching agent, the outline agent, the perspective agent, the retrieval agent and the generating agent to generate the text content required by the final user;

[0042] The review result is that the initial text content meets the user's writing needs or the initial text content does not meet the user's writing needs.

[0043] The following is a specific example of a text generation system based on a multi-agent collaboration mechanism MOPAR according to an embodiment of the present invention. Figure 2 As shown, the above technical solution of the embodiment of the present invention is described in detail.

[0044] In order to solve many limitations in the existing text generation technology, the embodiment of the present invention proposes a multi-agent collaboration mechanism MOPAR (Match-Outline-Perspective-Action-Reflect). Among them, Match: automatically matches the domain agent according to the user's writing requirements. Outline: the outline generation agent first writes an outline for the user's writing requirements. Perspective: the perspective generation agent generates a variety of writing perspectives based on the results of the outline generation agent and the user's writing requirements. Action: In the Action stage, it can be divided into three agents, responsible for retrieval (Search), summary (Summary) and generation (Generation), and these three agents jointly perform tasks (Take Action). The retrieval agent is responsible for retrieving professional domain knowledge related to the output of the outline agent and the perspective generation agent from the knowledge base, and then handing the results to the summary agent. The summary agent integrates all perspectives of all outlines and the retrieved knowledge, summarizes them, and then passes the message to the generation agent to generate the initial text. Reflect: the reflection agent, which reviews and modifies according to the results of the Action stage, and finally the user obtains the modified and optimized text content.

[0045] The implementation steps of the embodiment of the present invention specifically include:

[0046] S1. User demand reception and multi-agent matching

[0047] S1.1 The algorithm of the embodiment of the present invention takes the user's needs in natural language as input, including writing topics, target audiences, content requirements, etc. The algorithm of the embodiment of the present invention has good natural language understanding capabilities and can accurately analyze the user's intentions.

[0048] S1.2 According to user needs, the matching prompt word p_match is combined and input into a large language model (the embodiment of the present invention is not limited to a specific large language model, such as GPT-4, Tongyi Qianwen, etc.), and an intelligent matching agent (Match agent) is formed. In addition to user needs, p_match also contains relevant information such as the function description of agents in various fields, which is used to achieve multi-agent matching.

[0049] S1.3 Match agent classifies and selects user needs and determines a set of multi-agents in fields that meet the needs (such as the legal field, medical field, etc.). The multi-agent set refers to a set consisting of outline generation agents (Outline agent), perspective generation agents (Perspective agent), search agents (Search agent), summary agents (Summary agent), generation agents (Generation agent), and reflection agents (Reflection agent). These agents are executed sequentially in a pipeline manner to generate texts with deep professional knowledge and wide content coverage for users.

[0050] S2. Outline Generation

[0051] S2.1 Construct the outline generation prompt word p_outline, combine it with user needs and input it into the big model to construct the outline generation agent (Outline agent).

[0052] S2.2 Outline Agent generates a writing outline o according to the user's needs, including the outline title and writing theme, and breaks down the user's needs to form the main structure of the writing.

[0053] S3. Multi-dimensional perspective generation

[0054] S3.1 Construct the perspective generation prompt word p_perspective, combine user needs and the output o of the Outline agent and input it into the large language model to construct the perspective generation agent (Perspective agent).

[0055] S3.2 Perspective Agent uses the outline and user needs to generate diverse writing perspectives p for each level of the outline, improving the diversity of text generation. The agent can consider different writing angles to ensure diverse details of the generated text.

[0056] S4. Information Retrieval

[0057] S4.1 constructs the search hint word p_search, combines the output o of the Outline agent and the output p of the perspective generation agent, and constructs the retrieval agent (Search agent).

[0058] S4.2 Search agent can generate a search statement q for each writing perspective of each level of outline, and then the agent calls the internal knowledge base or Internet search engine to retrieve relevant professional field knowledge. Using this knowledge, the agent will generate a response to the search statement q as the search result a. The search agent of the embodiment of the present invention can ensure the accuracy and relevance of the search results, and ensure that the subsequent text generation can be based on reliable information.

[0059] S5. Information Summary

[0060] S5.1 constructs the summary prompt word p_summary, combines the output o of the Outline agent, the output p of the Perspective agent and the output a of the Search agent, and constructs the Summary agent.

[0061] S5.2 Summary Agent summarizes each writing angle and corresponding knowledge a in each level of the outline according to the requirements in the summary prompt word p_summary, extracts key information, outputs comprehensive summary knowledge s, and eliminates redundant information. The main task of this agent is to ensure the accuracy and conciseness of the information so that subsequent text generation can be based on high-quality information.

[0062] S6. Text Generation

[0063] S6.1 Text generation agent construction: The algorithm combines the text generation prompt word p_generation based on the output of the Outline agent, the output of the Perspective agent, and the output of the Summary agent, inputs it into the large language model, and constructs a text generation agent.

[0064] S6.2 First draft generation: The Generation agent generates the first draft text g based on the text generation prompt word p_generation. The goal of this agent is to generate text by combining the outputs of the Outline agent, the Perspective agent, and the Summary agent to ensure the coherence and naturalness of the text, improve the depth and breadth of text knowledge, and enable users to obtain higher quality text.

[0065] S7. Reflection and Optimization

[0066] S7.1 Construction of reflection prompt words: The algorithm combines the reflection prompt word p_reflection according to the output of the Generation agent and the needs of the user, inputs it into the large model, and constructs the reflective agent.

[0067] S7.2 Draft review: The Reflect agent reviews and evaluates the draft text according to the reflection prompt word p_reflect, and generates review opinions or revision opinions. The core goal of the Reflect agent is to consider the accuracy, coherence and degree of compliance with user needs of the text during the review process and generate high-quality review opinions.

[0068] S7.3 Feedback and revision: The Generation agent rewrites based on the opinions of the Reflect agent, and the Reflect agent reviews it to form a feedback mechanism. The feedback mechanism can be iterated multiple times to achieve the best results and ultimately generate high-quality text.

[0069] S7.4 Demand Satisfaction Judgment: The Reflect agent will also judge whether the current content meets the user's needs. If not, it will be fed back to the Match agent or other agents for regeneration or additional information supplementation. Specifically, the opinions will be fed back to the Match agent for matching multiple agents in other fields (to meet the user's cross-field writing needs), or fed back to the Outline agent, Perspective agent, Search agent, and Generation agent for regeneration or supplementation of outlines, perspectives, and knowledge.

[0070] S8. Final text output

[0071] After the final review by the Reflect agent, users obtain the optimized final text content.

[0072] like Figure 3 As shown, the intelligent agent of the embodiment of the present invention specifically includes the following contents:

[0073] 1. Multi-agent collaborative mechanism (MOPAR):

[0074] The MOPAR mechanism includes five stages: Match, Outline, Perspective, Action (Search, Summary, Generation) and Reflect. Through the division of labor and cooperation among intelligent agents, efficient and professional text generation is achieved.

[0075] The collaboration between agents, including how agents dynamically combine and perform tasks based on user needs.

[0076] 2. Dynamic matching of agents (Match):

[0077] Automatically identify user needs and match appropriate domain agents, reducing the need for fine-tuning models for specific scenarios.

[0078] 3. Outline and multi-dimensional perspective generation (Outline & Perspective):

[0079] Automatically generate a writing outline to provide a structured framework for user needs.

[0080] Generate diverse writing perspectives based on the outline to increase the diversity and professionalism of the text.

[0081] 4. Information retrieval and summary (Search & Summary):

[0082] Intelligently retrieve professional information in related fields and ensure the accuracy and relevance of the information.

[0083] Summarize the retrieved information, eliminate redundant information, and improve information quality.

[0084] 5. High-quality text generation (Generation):

[0085] Combine the results of the previous stages to generate a coherent and natural first draft text, ensuring the professionalism and breadth of the content.

[0086] 6. Feedback and optimization mechanism (Reflect):

[0087] Review the generated text and provide modification suggestions to form an iterative optimization feedback mechanism to ensure the quality of the final text generation.

[0088] In summary, the problem to be solved by the embodiments of the present invention is to improve the quality of text generation, enhance the knowledge expertise and content richness of the text. Specifically, the purposes of the embodiments of the present invention include:

[0089] (1) Improving the professionalism of generated content: In order to address the limitations of large language models in professional domain knowledge, the embodiment of the present invention proposes a multi-agent-based MOPAR collaboration method, in which the agents can automatically match, retrieve and integrate professional knowledge in related fields according to the user's specific domain requirements, ensuring that the generated text is highly professional.

[0090] (2) Increasing the diversity of generated content: The “Perspective” agent in the MOPAR mechanism can generate text content from multiple perspectives according to user requirements, while the agent in the “Action” stage further enriches the text details from different dimensions, thereby ensuring that the final generated text not only meets user needs but also has a variety of perspectives.

[0091] (3) Improve system flexibility: Different from traditional fixed models and single-scenario applications, the embodiments of the present invention achieve rapid adaptation to different application scenarios through flexible collaboration among multiple agents. Whether in the fields of medicine, law, or engineering, MOPAR can quickly adjust strategies to meet changing needs.

[0092] (4) Establishing an effective feedback and optimization mechanism: In order to solve the problem of lack of effective feedback in the prior art, the embodiment of the present invention introduces a "Reflect" intelligent agent, which is responsible for reviewing the generated content and optimizing and correcting it as needed, thereby ensuring the quality of the output text.

[0093] In summary, the embodiments of the present invention aim to provide a more flexible, professional and diversified text generation solution through the MOPAR multi-agent collaboration mechanism, so as to overcome the main shortcomings of the prior art and provide users with high-quality text generation method services. The goal of the embodiments of the present invention is to build a text generation platform that can adapt to different scenarios and has self-optimization capabilities, thereby promoting the development of text generation technology to a new level.

[0094] Method Embodiment

[0095] According to an embodiment of the present invention, a text generation method based on a multi-agent collaboration mechanism MOPAR is provided. Figure 4 is a flow chart of a text generation method based on a multi-agent collaboration mechanism MOPAR according to an embodiment of the present invention. Figure 4 As shown, the text generation method based on the multi-agent collaboration mechanism MOPAR according to an embodiment of the present invention specifically includes:

[0096] Step S401, obtaining the user's writing needs through a matching agent, generating matching prompt words according to the writing needs, transmitting the matching prompt words to a domain multi-agent system, performing domain matching with the domain multi-agent system through the matching prompt words, and optimizing according to feedback information of a reflective agent in the domain multi-agent system;

[0097] The writing requirements include writing topics, target audiences, and content requirements;

[0098] The matching prompt words include the writing requirements and relevant information of multiple agents in various fields;

[0099] Step S402, receiving the matching prompt word through the domain multi-agent system, acquiring a multi-agent in the corresponding domain according to the matching prompt word, and generating the text content required by the user through the multi-agent in the corresponding domain based on the multi-agent collaboration mechanism MOPAR, specifically including:

[0100] The outline agent in the multi-agent receives the writing requirement, generates a writing main body structure according to the writing requirement, transmits the writing requirement and the writing main body structure to the perspective agent, and optimizes the writing main body structure according to the feedback information of the reflection agent; wherein the writing main body structure includes several levels of writing outlines;

[0101] Receiving the writing requirement and the writing subject framework through a perspective agent, generating several types of writing perspectives for each level of the writing outline based on the writing requirement, transmitting the writing subject framework and the several types of writing perspectives to a retrieval agent, and optimizing according to feedback information from the reflection agent;

[0102] The writing subject structure and the several types of writing perspectives are received by a retrieval agent, and corresponding search statements are generated for each writing perspective of each level of the writing outline according to the writing subject structure and the several types of writing perspectives, a search engine is called to retrieve professional knowledge in a related field through the search statements to obtain search results, the writing subject structure, the several types of writing perspectives and the search results are transmitted to a summarizing agent, and optimization is performed according to feedback information from the reflecting agent;

[0103] The summarizing agent receives the writing subject framework, the several types of writing perspectives and the search results, summarizes each writing perspective and its corresponding professional knowledge of each level of the writing outline and removes redundant content to obtain comprehensive summary content, and transmits the comprehensive summary content to the generating agent;

[0104] Receiving the comprehensive summary content through a generating agent, generating text according to the comprehensive summary content to obtain initial text content, transmitting the initial text content to the reflecting agent, and optimizing according to feedback information from the reflecting agent;

[0105] The reflection agent receives the writing requirement and the initial text content, reviews the initial text content according to the writing requirement, forms corresponding feedback information based on the review result, and transmits the feedback information to the matching agent, the outline agent, the perspective agent, the retrieval agent and the generation agent to generate the final text content required by the user;

[0106] The review result is that the initial text content meets the user's writing needs or the initial text content does not meet the user's writing needs.

[0107] The embodiment of the present invention is a method embodiment corresponding to the above-mentioned system embodiment. The specific operation of each step can be understood by referring to the description of the system embodiment, which will not be repeated here.

[0108] In summary, the beneficial effects of the embodiments of the present invention include:

[0109] 1. Reduce the reliance on model training and fine-tuning: By designing a new multi-agent collaboration mechanism, MOPAR, the embodiments of the present invention strive to significantly reduce or even eliminate the need for model fine-tuning in specific scenarios by intelligently matching agents (Match) and multiple agents in the Action stage without sacrificing generation quality. The agents in the MOPAR mechanism can be dynamically combined according to the specific needs of users, so as to adapt to different application scenarios without changing the basic model.

[0110] 2. Improved professionalism of generated content: Through collaboration between agents, the embodiments of the present invention can automatically match, retrieve and integrate professional knowledge in related fields according to the user's specific field needs, ensuring that the generated text is highly professional. This mechanism breaks the limitations of a single model in a professional field and improves the accuracy and credibility of generated content.

[0111] 3. Increased diversity of generated content: The "Perspective" agent can generate a variety of writing perspectives, while the "Action" stage agent further enriches the text details, so that the final generated text not only meets user needs, but also has a multi-dimensional perspective, enhancing the richness and attractiveness of the text.

[0112] 4. Improved flexibility of text generation algorithm: The embodiment of the present invention realizes rapid adaptation to different application scenarios through flexible collaboration among multiple agents. Whether facing special needs in the fields of medicine, law or engineering, the system can quickly adjust strategies to meet the ever-changing needs.

[0113] 5. An effective feedback and optimization mechanism has been established: The introduction of the "Reflect" intelligent agent solves the problem of lack of effective feedback in existing technologies. It is responsible for reviewing the generated content and optimizing and correcting it as needed to ensure the quality of the output text. This mechanism improves the accuracy and reliability of text generation.

[0114] 6. High degree of automation: From demand reception, intelligent agent matching, outline generation to final text output, the entire process is highly automated, which greatly reduces the need for manual intervention and improves work efficiency.

[0115] 7. Support cross-domain applications: Through collaboration between intelligent agents, the embodiments of the present invention can easily cope with complex cross-domain tasks, achieve seamless switching between different fields, and provide users with a unified text generation solution.

[0116] 8. Improved text generation effect: The method established by the embodiment of the present invention provides a more intelligent, flexible and efficient text generation technology, which enables users to obtain higher quality and more targeted service experience, and through continuous self-optimization, it can better meet the personalized needs of users.

[0117] Device Example 1

[0118] An embodiment of the present invention provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program implements the steps described in the method embodiment when executed by the processor.

[0119] Device Example 2

[0120] An embodiment of the present invention provides a computer-readable storage medium, on which a program for implementing information transmission is stored. When the program is executed by a processor, the steps described in the method embodiment are implemented.

[0121] The computer-readable storage medium in this embodiment includes, but is not limited to, ROM, RAM, magnetic disk or optical disk, etc.

[0122] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A text generation system based on multi-agent collaboration mechanism MOPAR, characterized by include: A matching agent is connected to the domain multi-agent system to obtain user writing requirements, generate matching prompt words according to the writing requirements, transmit the matching prompt words to the domain multi-agent system, perform domain matching with the domain multi-agent system through the matching prompt words, and optimize according to feedback information of the reflective agent in the domain multi-agent system; The domain multi-agent system includes several multi-agents in different domains, which are connected to the matching agent, and are used to receive the matching prompt words, and obtain the multi-agent in the corresponding domain according to the matching prompt words, and generate the text content required by the user through the multi-agent in the corresponding domain based on the multi-agent collaboration mechanism MOPAR.

2. The system according to claim 1, characterized in that The writing requirements include writing topics, target audiences, and content requirements; The matching prompt words include the writing requirements and relevant information of multiple agents in various fields.

3. The system according to claim 2, characterized in that The domain multi-agent system specifically includes: An outline agent, connected to the matching agent, the perspective agent and the reflection agent, is used to receive the writing requirements, generate a writing body structure according to the writing requirements, transmit the writing requirements and the writing body structure to the perspective agent, and optimize according to the feedback information of the reflection agent; wherein the writing body structure includes several levels of writing outlines; A perspective agent, connected to the outline agent, the retrieval agent and the reflection agent, for receiving the writing requirements and the writing subject structure, generating several types of writing perspectives for each level of the writing outline based on the writing requirements, transmitting the writing subject structure and the several types of writing perspectives to the retrieval agent, and optimizing according to the feedback information of the reflection agent; A search agent, connected to the perspective agent, the summary agent and the reflection agent, for receiving the writing subject structure and the several types of writing perspectives, generating corresponding search statements for each writing perspective of each level of the writing outline according to the writing subject structure and the several types of writing perspectives, calling a search engine to retrieve professional knowledge in related fields through the search statements to obtain search results, transmitting the writing subject structure, the several types of writing perspectives and the search results to the summary agent, and optimizing according to feedback information from the reflection agent; A summarizing agent, connected to the retrieval agent and the generating agent, for receiving the writing subject structure, the several types of writing perspectives and the retrieval results, summarizing each writing perspective and its corresponding professional knowledge of each level of the writing outline and eliminating redundant content to obtain a comprehensive summary content, and transmitting the comprehensive summary content to the generating agent; A generating agent connected to the summarizing agent and the reflecting agent, configured to receive the comprehensive summary content, generate text according to the comprehensive summary content, obtain initial text content, transmit the initial text content to the reflecting agent, and optimize according to the feedback information of the reflecting agent; The reflective agent is connected to the matching agent, the outline agent, the perspective agent, the retrieval agent and the generating agent, and is used to receive the writing requirements and the initial text content, review the initial text content according to the writing requirements, form corresponding feedback information based on the review results, and transmit the feedback information to the matching agent, the outline agent, the perspective agent, the retrieval agent and the generating agent to generate the text content required by the final user.

4. The system according to claim 3, characterized in that The review result is that the initial text content meets the user's writing needs or the initial text content does not meet the user's writing needs.

5. A text generation method based on multi-agent collaboration mechanism MOPAR, characterized by include: Acquire user writing requirements through a matching agent, generate matching prompt words according to the writing requirements, transmit the matching prompt words to a domain multi-agent system, perform domain matching with the domain multi-agent system through the matching prompt words, and optimize according to feedback information of a reflective agent in the domain multi-agent system; The matching prompt word is received through the domain multi-agent system, and a multi-agent in the corresponding domain is obtained according to the matching prompt word, and the text content required by the user is generated by the multi-agent in the corresponding domain based on the multi-agent collaboration mechanism MOPAR.

6. The method according to claim 5, characterized in that The writing requirements include writing topics, target audiences, and content requirements; The matching prompt words include the writing requirements and relevant information of multiple agents in various fields.

7. The method according to claim 6, characterized in that The text content required by users is generated by multiple agents in the corresponding fields based on the multi-agent collaboration mechanism MOPAR, including: The outline agent in the multi-agent receives the writing requirement, generates a writing main body structure according to the writing requirement, transmits the writing requirement and the writing main body structure to the perspective agent, and optimizes the writing main body structure according to the feedback information of the reflection agent; wherein the writing main body structure includes several levels of writing outlines; Receiving the writing requirement and the writing subject framework through a perspective agent, generating several types of writing perspectives for each level of the writing outline based on the writing requirement, transmitting the writing subject framework and the several types of writing perspectives to a retrieval agent, and optimizing according to feedback information from the reflection agent; The writing subject structure and the several types of writing perspectives are received by a retrieval agent, and corresponding search statements are generated for each writing perspective of each level of the writing outline according to the writing subject structure and the several types of writing perspectives, a search engine is called to retrieve professional knowledge in a related field through the search statements to obtain search results, the writing subject structure, the several types of writing perspectives and the search results are transmitted to a summarizing agent, and optimization is performed according to feedback information from the reflecting agent; The summarizing agent receives the writing subject framework, the several types of writing perspectives and the search results, summarizes each writing perspective and its corresponding professional knowledge of each level of the writing outline and removes redundant content to obtain comprehensive summary content, and transmits the comprehensive summary content to the generating agent; Receiving the comprehensive summary content through a generating agent, generating text according to the comprehensive summary content to obtain initial text content, transmitting the initial text content to the reflecting agent, and optimizing according to feedback information from the reflecting agent; The writing requirement and the initial text content are received through the reflective agent, the initial text content is reviewed according to the writing requirement, corresponding feedback information is formed based on the review result, and the feedback information is transmitted to the matching agent, outline agent, perspective agent, retrieval agent and generation agent to generate the text content required by the final user.

8. The method according to claim 7, characterized in that The review result is that the initial text content meets the user's writing needs or the initial text content does not meet the user's writing needs.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the text generation method based on the multi-agent collaboration mechanism MOPAR as described in any one of claims 5 to 8.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores an implementation program for information transmission, and when the program is executed by the processor, it implements the steps of the text generation method based on the multi-agent collaboration mechanism MOPAR as described in any one of claims 5-8.

Images