Multi-agent online collaborative discussion system

Through the multi-agent online collaborative discussion system, a large-language model-driven agent conducts multiple rounds of discussion with users, simulates the role of students and teachers, solving the problems of high cost and low efficiency in existing online collaborative learning, and real interaction is achieved and efficient learning.

CN119090466APending Publication Date: 2024-12-06HUAZHONG NORMAL UNIV
View PDF 0 Cites 7 Cited by

Patent Information

Application Number
CN202411136712.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-19
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

In existing online collaborative learning, due to the need for professional research teams to design roles, it is cost-effective and inefficient, and lacks real interaction and effective evaluation mechanisms.

Method used

Adopt a multi-agent online collaborative discussion system, using a large language model-driven agent to conduct multiple rounds of discussion with users, simulate the role of students and teachers through preset roles, realize a real and interactive learning environment, and support online collaboration between real people and agents. study.

Benefits of technology

Save time and labor costs, improve the effect of online collaborative learning, enhance students' sense of participation and knowledge mastery, and improve the interactivity and realism of learning scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119090466A_ABST
    Figure CN119090466A_ABST
Patent Text Reader

Abstract

The invention belongs to the technical field of education, and particularly discloses a multi-agent online cooperative discussion system, which comprises a discussion module used for completing a discussion process together with a user based on a pre-created agent, the agent being driven by a large language model, and the user being driven by the large language model; the discussion process comprises the steps that a first prompt word and a second prompt word are input into a large language model, and a first type of agents and a second type of agents are determined; the first-class intelligent agent issues a discussion task; performing multiple rounds of discussion until each second-class agent and each user complete the discussion target; according to the system, on one hand, a large amount of time and labor cost can be saved, on the other hand, the online collaborative learning effect can be improved, and students are helped to better participate in discussion and grasp knowledge.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application belongs to the field of educational technology, and more specifically, relates to a multi-agent online collaborative discussion system. Background Art

[0002] In online collaborative learning with existing technologies, implementing different scaffolding settings requires a lot of manpower and time costs. Designing and implementing an effective role design experiment requires a professional research team, including educational experts, technical developers, and data analysts. In addition, the experimental cycle is usually long, and you need to wait patiently for data accumulation and analysis results, during which you may encounter various unexpected difficulties. Therefore, although role design is of great significance in collaborative learning, online collaborative learning with existing technologies requires high costs and low efficiency. Summary of the invention

[0003] In view of the defects of the prior art, the purpose of this application is to provide a multi-agent online collaborative discussion system, which aims to solve the problems of high cost and low efficiency of online collaborative learning in the prior art due to role design by research teams.

[0004] To achieve the above objectives, in a first aspect, the present application provides a multi-agent online collaborative discussion system, including a discussion module and a communication module, wherein:

[0005] The discussion module is used to complete the discussion process with the user based on the pre-created agent, wherein the agent is driven by a large language model;

[0006] The discussion process includes:

[0007] Inputting a first prompt word and a second prompt word into a large language model respectively, determining a first type of intelligent agent and a second type of intelligent agent in the pre-created intelligent agents, wherein the first prompt word includes the identity, capability parameter, discussion task and discussion goal of the first type of intelligent agent, and the second prompt word includes the identity and capability parameter of the second type of intelligent agent;

[0008] The first type of agent publishes a discussion task;

[0009] Conduct multiple rounds of discussion until each second-class agent and user completes the discussion goal. Each round of discussion includes:

[0010] Generate a third prompt word corresponding to each second-class agent based on the capability parameter and historical speech of each second-class agent, the historical speech of the first-class agent, and the historical speech of the user, and input the third prompt word into the large language model to obtain the first behavior and / or first speech content corresponding to each second-class agent output by the large language model;

[0011] Each second-type intelligent agent performs a corresponding first behavior, wherein the first behavior includes outputting a first speech content or not speaking;

[0012] Generate a fourth prompt word corresponding to the first-class agent based on the ability parameters and historical speeches of the first-class agent, the historical speeches of each second-class agent, and the historical speeches of the user, and input the fourth prompt word into the large language model to obtain the second behavior and / or second speech content of the first-class agent output by the large language model;

[0013] The first type of agent performs a corresponding second behavior and / or outputs a second speech content;

[0014] The communication module is used to realize real-time communication between the agent and the user.

[0015] This application combines a large model, prompt words and a multi-agent framework to build a multi-agent online collaborative discussion system. The system creates a more realistic and interactive learning environment by simulating different student roles and teacher roles. These agents can conduct collaborative discussions based on preset roles, and support real learners to form different groups with different student agents for online collaborative learning. On the one hand, it can save a lot of time and labor costs, and on the other hand, it can improve the effect of online collaborative learning and help students better participate in discussions and master knowledge.

[0016] According to the multi-agent online collaborative discussion system provided by the present application, the third prompt word includes a fifth prompt word, a sixth prompt word and a seventh prompt word, and the third prompt word corresponding to each second-class agent is generated based on the capability parameter and historical speech of each second-class agent, the historical speech of the first-class agent and the historical speech of the user, and the third prompt word is input into the large language model to obtain the first behavior and / or the first speech content corresponding to each second-class agent output by the large language model, including:

[0017] Based on the capability parameters and historical speeches of each second-type intelligent agent, the historical speeches of the first-type intelligent agent, and the historical speeches of the user, a fifth prompt word corresponding to each second-type intelligent agent is generated, and the fifth prompt word is input into the large language model to obtain the first thinking result corresponding to each second-type intelligent agent output by the large language model;

[0018] Generate a sixth prompt word based on the first thinking result and the capability parameter of the second type of agent, and input the sixth prompt word into the large language model to obtain the first behavior output by the large language model;

[0019] When the first behavior includes outputting the first speech content, a seventh prompt word is generated based on the first behavior, the historical speeches of each second-type intelligent agent, the historical speeches of the first-type intelligent agent and the historical speeches of the user, and the seventh prompt word is input into the large language model to obtain the first speech content output by the large language model.

[0020] During the discussion process, this application continuously updates prompt words based on historical speeches and the ability values ​​of each intelligent agent, and inputs them into the large language model to generate appropriate behaviors and speeches for the second type of intelligent agent, making the intelligent agent more human-like, thereby improving the interactivity and realism in online collaborative learning scenarios.

[0021] According to the multi-agent online collaborative discussion system provided by the present application, the fourth prompt word includes an eighth prompt word, a ninth prompt word and a tenth prompt word, and the fourth prompt word corresponding to the first-class agent is generated based on the capability parameter and historical speech of the first-class agent, the historical speech of each second-class agent and the historical speech of the user, and the fourth prompt word is input into the large language model to obtain the second behavior and / or second speech content of the first-class agent output by the large language model, including:

[0022] Generate an eighth prompt word corresponding to the first-class agent based on the ability parameters and historical speeches of the first-class agent, the historical speeches of each second-class agent, and the historical speeches of the user, and input the eighth prompt word into the large language model to obtain a second thinking result corresponding to the first-class agent output by the large language model;

[0023] Generate a ninth prompt word based on the second thinking result and the capability parameter of the first type of agent, and input the ninth prompt word into the large language model to obtain the second behavior output by the large language model;

[0024] In the case where the second behavior includes outputting the second speech content, a tenth prompt word is generated based on the second behavior, the historical speeches of each second-type intelligent agent, the historical speeches of the first-type intelligent agent, and the historical speeches of the user, and the tenth prompt word is input into the large language model to obtain the second speech content output by the large language model.

[0025] During the discussion process, this application continuously updates prompt words based on historical speeches and the ability values ​​of the first-class intelligent agents, and inputs them into the large language model to generate appropriate behaviors and speeches for the first-class intelligent agents, making the agents more human-like, thereby improving the interactivity and realism in online collaborative learning scenarios.

[0026] According to the multi-agent online collaborative discussion system provided by the present application, the second behavior includes one or more of the following:

[0027] Call on students to answer questions, change the speaking order, encourage and guide.

[0028] According to the multi-agent online collaborative discussion system provided by the present application, in the case where the second behavior is to answer a question by name, after the first type of agent performs the corresponding second behavior and / or outputs the second speech content, the discussion process further includes:

[0029] Generate an eleventh prompt word based on the ability parameters and historical speeches of the named second-category intelligent agent, the historical speeches of other second-category intelligent agents, the historical speeches of the first-category intelligent agent, and the historical speeches of the user, and input the eleventh prompt word into the large language model to obtain the third speech content output by the large language model;

[0030] The named second-type agent outputs the third speech content.

[0031] According to the multi-agent online collaborative discussion system provided by the present application, the discussion process further includes:

[0032] After each second-category agent and user completes the discussion goal, an eleventh prompt word is generated based on the ability parameters and historical speeches of the first-category agent, the historical speeches of each second-category agent, and the historical speeches of the user, and the eleventh prompt word is input into the large language model to obtain a summary speech output by the large language model;

[0033] The first type of intelligent agent outputs the summary speech, which includes the main content and highlights of the discussion process, as well as the performance evaluation of each second type of intelligent agent and user.

[0034] According to the multi-agent online collaborative discussion system provided by the present application, the first behavior includes one or more of the following:

[0035] Silence, raising opinions, questioning, asking questions, summarizing behavior.

[0036] According to the multi-agent online collaborative discussion system provided by the present application, the capability parameters of the second type of agents include one or more of the following:

[0037] Identity information parameters, prior knowledge parameters, learning motivation parameters, intra-group interaction willingness parameters, and cognitive ability parameters.

[0038] According to the multi-agent online collaborative discussion system provided by the present application, the multi-agent online collaborative discussion system further includes:

[0039] The memory module is used to store chat conversations, interaction records and feedback information during the discussion process.

[0040] According to the multi-agent online collaborative discussion system provided by the present application, the multi-agent online collaborative discussion system further includes:

[0041] The knowledge base module is used to manage educational content to ensure the timeliness and accuracy of educational content.

[0042] In general, the above technical solutions conceived by this application have the following beneficial effects compared with the prior art:

[0043] (1) By combining large models, prompt words and multi-agent frameworks, a multi-agent online collaborative discussion system is constructed. The system creates a more realistic and interactive learning environment by simulating different student roles and teacher roles. These agents can conduct collaborative discussions based on preset roles and support real learners to form different groups with different student agents for online collaborative learning. On the one hand, it can save a lot of time and labor costs, and on the other hand, it can improve the effect of online collaborative learning and help students better participate in discussions and master knowledge.

[0044] (2) During the discussion, prompt words are continuously updated based on historical speeches and the ability values ​​of various intelligent agents, and input into the large language model to generate appropriate behaviors and speeches for various intelligent agents, making the intelligent agents more human-like, thereby improving the interactivity and realism in online collaborative learning scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 It is one of the structural diagrams of the multi-agent online collaborative discussion system provided in the embodiment of the present application;

[0047] Figure 2 is a detailed schematic diagram of a character design method provided in an embodiment of the present application;

[0048] Figure 3 This is the second structural diagram of the multi-agent online collaborative discussion system provided in the embodiment of the present application. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0050] The term "and / or" in this article is a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The symbol " / " in this article indicates that the associated objects are in an or relationship, for example, A / B means A or B.

[0051] The terms "first" and "second" in the specification and claims herein are used to distinguish different objects rather than to describe a specific order of the objects. For example, a first response message and a second response message are used to distinguish different response messages rather than to describe a specific order of the response messages.

[0052] In the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0053] In the description of the embodiments of the present application, unless otherwise specified, "multiple" means two or more than two. For example, multiple processing units refer to two or more processing units, etc.; multiple elements refer to two or more elements, etc.

[0054] First, the following contents are introduced:

[0055] The information age requires learners to have collaborative learning capabilities. However, there are still some problems in online collaborative learning. For example, many learners perform unevenly during participation; learners often face topic barriers and find it difficult to find suitable discussion topics; online collaborative learning lacks an effective evaluation mechanism and cannot accurately and real-time evaluate learners' learning outcomes.

[0056] In actual online collaborative learning, implementing different scaffolding settings requires a lot of manpower and time costs. Designing and implementing an effective role design experiment requires a professional research team, including educational experts, technical developers, and data analysts. In addition, the experimental cycle is usually long, and you need to wait patiently for data accumulation and analysis results, during which you may encounter various unexpected difficulties. Therefore, although role design is of great significance in collaborative learning, in practice, the limitations of resources and time must be fully considered to ensure the feasibility and efficiency of the experiment.

[0057] In recent years, with the rapid development of artificial intelligence technology, large models, prompt words, and multi-agent framework technologies have provided new solutions for simulating discussions between students and teachers in online collaborative learning. Large models have powerful text generation capabilities and can generate more natural and intelligent conversation content; prompt words are pre-set keywords or phrases that guide learners to stay focused in discussions, provide valuable input, and help them find appropriate topics; multi-agent frameworks can simulate different student roles and teacher roles. At present, educational applications mainly focus on the interaction between single agents and learners. Although they can provide personalized learning experience and instant feedback, their main functions remain at the tool use level. The application of multi-agents in the field of education is relatively small, especially in simulating conversations between students and teachers.

[0058] Next, combine Figure 1-Figure 3 The multi-agent online collaborative discussion system provided in the embodiments of the present application is introduced.

[0059] Figure 1 is one of the structural diagrams of the multi-agent online collaborative discussion system provided in the embodiment of the present application, such as Figure 1 As shown, the system includes a discussion module 110 and a communication module 120, wherein:

[0060] The discussion module 110 is used to complete the discussion process with the user based on the pre-created agent, wherein the agent is driven by a large language model;

[0061] The discussion process includes:

[0062] Inputting the first prompt word and the second prompt word into the large language model respectively, determining the first type of intelligent agent and the second type of intelligent agent in the pre-created intelligent agent, the first prompt word including the identity, ability parameter, discussion task and discussion goal of the first type of intelligent agent, and the second prompt word including the identity and ability parameter of the second type of intelligent agent;

[0063] The first type of agent publishes discussion tasks;

[0064] Conduct multiple rounds of discussion until each second-class agent and user completes the discussion goal. Each round of discussion includes:

[0065] Generate a third prompt word corresponding to each second-class agent based on the capability parameter and historical speech of each second-class agent, the historical speech of the first-class agent, and the historical speech of the user, and input the third prompt word into the large language model to obtain the first behavior and / or first speech content corresponding to each second-class agent output by the large language model;

[0066] Each second-type intelligent agent performs a corresponding first behavior respectively, and the first behavior includes outputting a first speech content or not speaking;

[0067] Generate a fourth prompt word corresponding to the first-class agent based on the ability parameters and historical speeches of the first-class agent, the historical speeches of each second-class agent, and the historical speeches of the user, and input the fourth prompt word into the large language model to obtain the second behavior and / or second speech content of the first-class agent output by the large language model;

[0068] The first type of agent performs a corresponding second behavior and / or outputs a second speech content;

[0069] The communication module 120 is used to implement real-time communication between the agent and the user.

[0070] In online collaborative learning scenarios, the teaching process is manifested as information-rich, multi-round and task-oriented discussions, involving interactive exchanges between students and between teachers and students. Therefore, the multi-agent online collaborative discussion system provided in this application designs and develops specific application scenarios for the open source multi-agent framework. The system is driven by a large language model as a whole, and two agent roles, teachers and students, are designed and developed. The large language model serves as the brain of the agent and handles complex language interactions and cognitive tasks.

[0071] Optionally, pre-created agents are built from open source multi-agent frameworks.

[0072] Optionally, the first type of agents are agents for teacher roles, and the second type of agents are agents for student roles.

[0073] Optionally, the present application does not limit the large language model used.

[0074] Optionally, the first prompt word and the second prompt word can be input into the large language model to personalize the configuration, settings and behavior performance of the intelligent agents under different settings, making the intelligent agents more human-like, thereby improving the interactivity and realism in online collaborative learning scenarios.

[0075] Optionally, during a discussion, one first-type intelligent agent is generally set, and multiple second-type intelligent agents are generally set.

[0076] Optionally, users can participate in discussions in the system as students, explore knowledge and solve problems with the intelligent agent, and gain knowledge from it. Through the assistance and virtuality of the intelligent agent, the interactivity and effectiveness of learning are enhanced.

[0077] Multiple intelligent agents simulate the process of online collaborative learning and conduct multiple rounds of discussions. During the discussion, different intelligent agents in each round first observe and think based on the dialogue history, and then make different behavioral choices based on their thinking and their own initialization settings. Different behaviors produce different speeches, and the thinking and speech data of each round are stored as the memory of the intelligent agent. All learning goals are achieved through multiple rounds of discussions.

[0078] Figure 2 is a detailed schematic diagram of the role design method provided in the embodiment of the present application, such as Figure 2 As shown, before the discussion task begins, the pre-created agent must first be initialized, that is, it must be given corresponding roles and capabilities.

[0079] The identity, ability parameters, discussion tasks and discussion goals of the first type of intelligent agent are input into the large language model, so that a certain intelligent agent has the corresponding teacher identity and ability, such as professional field, mathematics, Chinese, science, etc., teaching experience level, novice teacher, senior teacher, etc. At the same time, it is necessary to confirm the tasks and goals of this online collaborative learning, clarify the knowledge points, and release the discussion tasks at the beginning.

[0080] The identity and ability parameters of the second type of agents are input into the large language model so that multiple agents have corresponding student identities and abilities.

[0081] After the identity and ability of the agent are determined, the first type of agent will issue a discussion task. The second type of agent will make a discussion speech according to the teacher's type agent's call or in a specified order. When it is their turn to speak, they will use the large language model to generate their own thoughts based on the discussion tasks issued by the teacher and the previous dialogue history. They will think about the discussion viewpoints of other groups and combine their own ability parameters to choose their own behavior, such as remaining silent or putting forward opinions.

[0082] After a round of discussion by students is finished, the intelligent agent in the teacher role will first evaluate the students' participation and scoring points in this round of discourse. By evaluating the students' discussion and generating their own thoughts based on their own identity settings, they can flexibly choose different behaviors, such as encouragement, guidance or comments, to promote effective teaching and learning interactions.

[0083] The system supports real-time communication methods to simulate instant communication and interaction between student and teacher agents. Real-time communication is a key element in an educational collaborative learning environment, facilitating the instant transmission of information and the rapid resolution of problems. The system ensures the privacy and security of sensitive data such as student personal information and discussions retained during use.

[0084] The multi-agent online collaborative discussion system provided by the present invention constructs a multi-agent online collaborative discussion system by combining a large model, prompt words and a multi-agent framework. The system creates a more realistic and interactive learning environment by simulating different student roles and teacher roles. These agents can conduct collaborative discussions according to preset roles, and support real learners to form different groups with different student agents for online collaborative learning. On the one hand, it can save a lot of time and labor costs, and on the other hand, it can improve the effect of online collaborative learning and help students better participate in discussions and master knowledge.

[0085] In some embodiments, the third prompt word includes a fifth prompt word, a sixth prompt word, and a seventh prompt word, and the third prompt word corresponding to each second-type agent is generated based on the capability parameter and historical speech of each second-type agent, the historical speech of the first-type agent, and the historical speech of the user, and the third prompt word is input into the large language model to obtain the first behavior and / or the first speech content corresponding to each second-type agent output by the large language model, including:

[0086] Based on the capability parameters and historical speeches of each second-type intelligent agent, the historical speeches of the first-type intelligent agent, and the historical speeches of the user, a fifth prompt word corresponding to each second-type intelligent agent is generated, and the fifth prompt word is input into the large language model to obtain the first thinking result corresponding to each second-type intelligent agent output by the large language model;

[0087] Generate a sixth prompt word based on the first thinking result and the ability parameter of the second type of agent, and input the sixth prompt word into the large language model to obtain a first behavior output by the large language model;

[0088] When the first behavior includes outputting the first speech content, a seventh prompt word is generated based on the first behavior, the historical speeches of each second-type intelligent agent, the historical speeches of the first-type intelligent agent, and the historical speeches of the user, and the seventh prompt word is input into the large language model to obtain the first speech content output by the large language model.

[0089] In some embodiments, the fourth prompt word includes an eighth prompt word, a ninth prompt word, and a tenth prompt word, and the fourth prompt word corresponding to the first type of agent is generated based on the capability parameter and historical speech of the first type of agent, the historical speech of each second type of agent, and the historical speech of the user, and the fourth prompt word is input into the large language model to obtain the second behavior and / or second speech content of the first type of agent output by the large language model, including:

[0090] Generate an eighth prompt word corresponding to the first-class agent based on the ability parameters and historical speeches of the first-class agent, the historical speeches of each second-class agent, and the historical speeches of the user, and input the eighth prompt word into the large language model to obtain a second thinking result corresponding to the first-class agent output by the large language model;

[0091] Generate a ninth prompt word based on the second thinking result and the capability parameter of the first type of agent, and input the ninth prompt word into the large language model to obtain a second behavior output by the large language model;

[0092] When the second behavior includes outputting the second speech content, a tenth prompt word is generated based on the second behavior, the historical speeches of each second-type intelligent agent, the historical speeches of the first-type intelligent agent, and the historical speeches of the user, and the tenth prompt word is input into the large language model to obtain the second speech content output by the large language model.

[0093] When it is each agent's turn to act, it needs to first generate prompt words based on its own ability parameters and all historical speeches, and input them into the large language model to obtain thinking results. Then, it generates new prompt words based on the thinking results and ability parameters. Finally, the corresponding actions and / or speech content are output through the large language model. If the action is an action that requires speech, the corresponding speech content will be output at the same time. If the action is an action that does not require speech, there is no need to output the speech content.

[0094] In some embodiments, the second behavior includes one or more of the following:

[0095] Call on students to answer questions, change the speaking order, encourage and guide.

[0096] Specifically, based on the thinking about students' participation and scores in this round, through the large language model, the teacher can dynamically change the order of students' discussion by calling on them to answer questions, changing the order of speaking, and so on to promote all students to participate in the discussion. At the same time, when students discuss the knowledge points of the collaborative task, the teacher's intelligent agent can choose to encourage the students to continue to maintain the passion of the discussion, while avoiding revealing other knowledge points to keep students motivated to explore and discover. Otherwise, the teacher's intelligent agent can take guiding actions and ask questions related to the unanswered knowledge points to help students find the right direction and stimulate their thinking.

[0097] Through multiple rounds of student discussions and timely intervention by teachers, the online collaborative learning ends when all the target knowledge points of this learning task are covered in the student discussions.

[0098] In some embodiments, when the second behavior is to answer a question by name, after the first type of agent performs the corresponding second behavior and / or outputs the second speech content, the discussion process further includes:

[0099] Generate an eleventh prompt word based on the question, the capability parameters and historical speeches of the named second-category intelligent agent, the historical speeches of other second-category intelligent agents, the historical speeches of the first-category intelligent agent, and the historical speeches of the user, and input the eleventh prompt word into the large language model to obtain the third speech content output by the large language model;

[0100] The second type of intelligent agent named outputs the third speech content.

[0101] After the teacher-type agent calls the roll, the student-type agent that is called will generate prompt words based on the question, its own ability parameters, and all historical speeches, and obtain the corresponding speech content through the large language model. The student-type agent that is called will then output the speech content.

[0102] In some embodiments, the discussion process also includes:

[0103] After each second-category agent and user completes the discussion goal, an eleventh prompt word is generated based on the ability parameters and historical speeches of the first-category agent, the historical speeches of each second-category agent, and the historical speeches of the user, and the eleventh prompt word is input into the large language model to obtain a summary speech output by the large language model;

[0104] The first type of agent outputs a summary speech, which includes the main content and highlights of the discussion process, as well as the performance evaluation of each second type of agent and user.

[0105] After the discussion task is completed, the teacher-type agent comments on the students' performance and the depth of the discussion, summarizes the main content and highlights of the discussion, evaluates each student's performance, points out the strengths and areas for improvement, and encourages students to continue their efforts and progress.

[0106] In some embodiments, the first behavior includes one or more of the following:

[0107] Silence, raising opinions, questioning, asking questions, summarizing behavior.

[0108] Student-type agents choose to remain silent, express opinions, question, raise questions, and summarize based on their thinking about tasks and dialogues and their own initialization settings. When the student role agent's learning motivation and cognitive ability settings are weak, the student agent chooses to remain silent while waiting for more information or observing peers' discussions. Silence helps simulate students' listening and thinking processes in real discussions. After the student agent forms his own opinions and has a strong setting, he will actively express opinions and share his ideas and understanding. Proposing opinions encourages students to express their own thinking and promotes the exchange and sharing of knowledge. When hearing other people's opinions or information, especially when it is different from his own prior knowledge, the student agent will raise questions to verify the accuracy and completeness of the information. Questioning behavior helps stimulate in-depth discussions and improve the quality and depth of discussions. When the student agent encounters content that he does not understand or a topic he wants to explore further, and has a strong learning motivation, he will raise questions. Raising questions encourages students to actively seek knowledge and promotes interaction between teachers and other students.

[0109] When a student-type agent is at a certain stage or end of a discussion, it will summarize and review the key points and results of the discussion. The summarizing behavior helps consolidate learning outcomes and ensures that every student understands and absorbs the content of the discussion.

[0110] In some embodiments, the capability parameters of the second type of agents include one or more of the following:

[0111] Identity information parameters, prior knowledge parameters, learning motivation parameters, intra-group interaction willingness parameters, and cognitive ability parameters.

[0112] Optionally, the capabilities of student-type agents may include identity information, prior knowledge, learning motivation, willingness to interact within a group, and cognitive ability, wherein identity information includes the student's grade, class, and background, ensuring that the agent has a clear identity in the simulation; prior knowledge covers the student's knowledge level and skill mastery related to the task, providing a basis for the agent's learning behavior; learning motivation involves the student's intrinsic and extrinsic motivation, as well as learning goals, to demonstrate the agent's enthusiasm for learning and continued participation; the setting of intra-group interaction capabilities focuses on students' group communication, collaboration, and leadership, enabling the agent to perform set interactions and cooperation in the group; the setting of cognitive abilities includes information processing, problem solving, and innovative thinking, the agent's ability to think and make decisions in discussions.

[0113] The second prompt word can be input into the large language model to set the capabilities of the student agent and diversify the student role to simulate different students in reality.

[0114] After the student-type intelligent agent clarifies its own identity and the settings of various aspects of its capabilities, each student-type intelligent agent will discuss and speak according to the teacher role intelligent agent's call or in a specified order. When it is their turn to speak, they will generate their own thoughts based on the learning tasks issued by the teacher and the previous dialogue history, think about the discussion viewpoints of students in other groups, and combine their own initialization parameters to choose their own behavior.

[0115] In some embodiments, the multi-agent online collaborative discussion system further includes:

[0116] The memory module is used to store chat conversations, interaction records and feedback information during the discussion process.

[0117] Figure 3 This is the second structural diagram of the multi-agent online collaborative discussion system provided in the embodiment of the present application, such as Figure 3 As shown, the multi-agent online collaborative discussion system also includes a memory module for storing chat dialogues, interaction records and feedback information of the online collaborative discussion so that the agents can provide continuous and consistent learning support.

[0118] In one embodiment of the present application, the database technology MySQL is used to store and manage memory data to ensure the security and reliability of the data.

[0119] In some embodiments, the multi-agent online collaborative discussion system further includes:

[0120] The knowledge base module is used to manage educational content to ensure the timeliness and accuracy of educational content.

[0121] Optionally, educational content may include course materials and reference materials, etc.

[0122] The knowledge base module supports the classification, indexing and updating of knowledge, and uses knowledge graphs to help intelligent agents better understand and associate knowledge points, simulate students' divergent thinking, and promote interdisciplinary learning and innovative thinking.

[0123] In one embodiment of the present application, the database technology MySQL is also used to store and manage knowledge base data, which facilitates rapid retrieval and updating by intelligent agents.

[0124] Through the joint collaboration of discussion module, communication module, memory module and knowledge base module, a highly interactive and intelligent online collaborative learning environment can be simulated. In this environment, the agents play the roles of teachers and students. They communicate in real time through the communication module, use the memory module to store and retrieve key information in the learning process, and support the acquisition of rich educational resources from the knowledge base module.

[0125] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions may be transmitted from a website site, a computer, a server or a data center to another website site, a computer, a server or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or a data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid state disk (SSD)), etc.

[0126] It should be understood that the various numerical numbers involved in the embodiments of the present application are only used for the convenience of description and are not used to limit the scope of the embodiments of the present application.

[0127] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A multi-agent online collaborative discussion system, It is characterized in that It includes discussion module and communication module, where: The discussion module is used to complete the discussion process with the user based on the pre-created agent, wherein the agent is driven by a large language model; The discussion process includes: Inputting a first prompt word and a second prompt word into a large language model respectively, determining a first type of intelligent agent and a second type of intelligent agent in the pre-created intelligent agents, wherein the first prompt word includes the identity, capability parameter, discussion task and discussion goal of the first type of intelligent agent, and the second prompt word includes the identity and capability parameter of the second type of intelligent agent; The first type of agent publishes a discussion task; Conduct multiple rounds of discussion until each second-class agent and user completes the discussion goal. Each round of discussion includes: Generate a third prompt word corresponding to each second-class agent based on the capability parameter and historical speech of each second-class agent, the historical speech of the first-class agent, and the historical speech of the user, and input the third prompt word into the large language model to obtain the first behavior and / or first speech content corresponding to each second-class agent output by the large language model; Each second-type intelligent agent performs a corresponding first behavior, wherein the first behavior includes outputting a first speech content or not speaking; Generate a fourth prompt word corresponding to the first-class agent based on the ability parameters and historical speeches of the first-class agent, the historical speeches of each second-class agent, and the historical speeches of the user, and input the fourth prompt word into the large language model to obtain the second behavior and / or second speech content of the first-class agent output by the large language model; The first type of agent performs a corresponding second behavior and / or outputs a second speech content; The communication module is used to realize real-time communication between the agent and the user.

2. The multi-agent online collaborative discussion system according to claim 1, It is characterized in that The third prompt words include a fifth prompt word, a sixth prompt word, and a seventh prompt word. The generating of the third prompt words corresponding to each second-class agent based on the capability parameters and historical speeches of each second-class agent, the historical speeches of the first-class agent, and the historical speeches of the user, and inputting the third prompt words into the large language model to obtain the first behavior and / or the first speech content corresponding to each second-class agent output by the large language model includes: Based on the capability parameters and historical speeches of each second-type intelligent agent, the historical speeches of the first-type intelligent agent, and the historical speeches of the user, a fifth prompt word corresponding to each second-type intelligent agent is generated, and the fifth prompt word is input into the large language model to obtain the first thinking result corresponding to each second-type intelligent agent output by the large language model; Generate a sixth prompt word based on the first thinking result and the capability parameter of the second type of agent, and input the sixth prompt word into the large language model to obtain the first behavior output by the large language model; When the first behavior includes outputting the first speech content, a seventh prompt word is generated based on the first behavior, the historical speeches of each second-type intelligent agent, the historical speeches of the first-type intelligent agent and the historical speeches of the user, and the seventh prompt word is input into the large language model to obtain the first speech content output by the large language model.

3. The multi-agent online collaborative discussion system according to claim 1, It is characterized in that The fourth prompt word includes an eighth prompt word, a ninth prompt word, and a tenth prompt word. The fourth prompt word corresponding to the first type of agent is generated based on the capability parameter and historical speech of the first type of agent, the historical speech of each second type of agent, and the historical speech of the user, and the fourth prompt word is input into the large language model to obtain the second behavior and / or second speech content of the first type of agent output by the large language model, including: Generate an eighth prompt word corresponding to the first-class agent based on the ability parameters and historical speeches of the first-class agent, the historical speeches of each second-class agent, and the historical speeches of the user, and input the eighth prompt word into the large language model to obtain a second thinking result corresponding to the first-class agent output by the large language model; Generate a ninth prompt word based on the second thinking result and the capability parameter of the first type of agent, and input the ninth prompt word into the large language model to obtain the second behavior output by the large language model; In the case where the second behavior includes outputting the second speech content, a tenth prompt word is generated based on the second behavior, the historical speeches of each second-type intelligent agent, the historical speeches of the first-type intelligent agent, and the historical speeches of the user, and the tenth prompt word is input into the large language model to obtain the second speech content output by the large language model.

4. The multi-agent online collaborative discussion system according to claim 1 or 3, It is characterized in that The second behavior includes one or more of the following: Call on students to answer questions, change the speaking order, encourage and guide.

5. The multi-agent online collaborative discussion system according to claim 4, It is characterized in that In the case where the second behavior is to answer a question by name, after the first type of agent performs the corresponding second behavior and / or outputs the second speech content, the discussion process further includes: generating an eleventh prompt word based on the question, the capability parameters and historical speeches of the named second-category intelligent agent, the historical speeches of other second-category intelligent agents, the historical speeches of the first-category intelligent agents, and the historical speeches of the user, and inputting the eleventh prompt word into the large language model to obtain the third speech content output by the large language model; The named second-type agent outputs the third speech content.

6. A multi-agent online collaborative discussion system according to any one of claims 1 to 4, It is characterized in that The discussion process also includes: After each second-category agent and user completes the discussion goal, an eleventh prompt word is generated based on the ability parameters and historical speeches of the first-category agent, the historical speeches of each second-category agent, and the historical speeches of the user, and the eleventh prompt word is input into the large language model to obtain a summary speech output by the large language model; The first type of intelligent agent outputs the summary speech, which includes the main content and highlights of the discussion process, as well as the performance evaluation of each second type of intelligent agent and user.

7. The multi-agent online collaborative discussion system according to any one of claims 1-2, It is characterized in that The first behavior includes one or more of the following: Silence, raising opinions, questioning, asking questions, summarizing behavior.

8. The multi-agent online collaborative discussion system according to claim 1, It is characterized in that The capability parameters of the second type of agents include one or more of the following: Identity information parameters, prior knowledge parameters, learning motivation parameters, intra-group interaction willingness parameters, and cognitive ability parameters.

9. The multi-agent online collaborative discussion system according to claim 1, It is characterized in that The multi-agent online collaborative discussion system also includes: The memory module is used to store chat conversations, interaction records and feedback information during the discussion process.

10. The multi-agent online collaborative discussion system according to claim 1, It is characterized in that The multi-agent online collaborative discussion system also includes: The knowledge base module is used to manage educational content to ensure the timeliness and accuracy of educational content.

Citation Information

Cited By

  • Knowledge question and answer game content generation method and device and electronic equipment

    CN120258149A

  • Multi-agent cooperative emotion support dialogue generation method and system

    CN120561255A

  • Teaching and research system and method based on large language model agent

    CN120780835A

  • A teaching and research system and method based on a large language model agent

    CN120780835B

  • Digital entertainment interaction method and device based on intelligent agent and storage medium

    CN120932640A