A large model-based agent-based question answering learning method and device

Abstract

This application discloses a question-answering learning method and apparatus based on a large-scale intelligent agent, relating to the field of artificial intelligence technology. The method includes: acquiring configuration information for question-answering learning in a target service scenario; constructing a main intelligent agent and multiple auxiliary intelligent agents based on process configuration information, template configuration information, and additional requirement information in the configuration information, wherein the main intelligent agent controls a question-answering learning process adapted to the target service scenario; when a user enters the question-answering learning interactive interface, the main intelligent agent is controlled to conduct multiple rounds of dialogue with the user according to the question-answering learning process, and during these multiple rounds of dialogue, corresponding auxiliary intelligent agents are invoked to enable the user to complete the question-answering learning task in the target service scenario. This application can enhance the flexibility of the question-answering learning process and meet the learning needs of users in different scenarios.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a question-answering learning method and apparatus based on a large model intelligent agent. Background Technology

[0002] The online assessment and quiz learning system relies on the Internet and computer technology to provide users with functions such as answering questions and learning assessments. It can meet the needs of assessment, knowledge consolidation and ability evaluation, and can provide scholars with efficient and convenient services.

[0003] Currently, traditional online assessment and quiz learning systems rely on a fixed linear process for question generation and testing. However, this fixed-process question generation and assessment model means that users can only answer questions according to the same steps in all scenarios, lacking flexibility and failing to meet the learning needs of users in different scenarios. Summary of the Invention

[0004] In view of this, this application provides a question-answering learning method and apparatus based on a large model intelligent agent, the main purpose of which is to enhance the flexibility of the question-answering learning process, so as to meet the learning needs of users in different scenarios.

[0005] According to a first aspect of this application, a question-answering learning method based on a large model agent is provided, the method comprising: Obtain the configuration information for question-answering learning in the target service scenario; Based on the process configuration information, template configuration information, and additional requirement information in the configuration information, a main intelligent agent and multiple auxiliary intelligent agents are constructed respectively. The main intelligent agent is used to control the question-answering learning process adapted to the target service scenario. When a user enters the question-and-answer learning interaction interface, the main intelligent agent is controlled to conduct multiple rounds of dialogue with the user according to the question-and-answer learning process. During the multiple rounds of dialogue, the corresponding auxiliary intelligent agents are called to enable the user to complete the question-and-answer learning task under the target service scenario.

[0006] Optionally, the plurality of auxiliary intelligent agents include a question-generating intelligent agent, a review intelligent agent, and a summary intelligent agent. The construction of a main intelligent agent and multiple auxiliary intelligent agents based on the process configuration information, template configuration information, and additional requirement information in the configuration information includes: The process configuration information, as well as the scenario information and / or knowledge background in the additional requirements information, are injected into the first intelligent agent to obtain the main intelligent agent; The question-generating requirements in the additional requirements information and the question-generating templates in the template configuration information are injected into the second intelligent agent to obtain the question-generating intelligent agent; The review requirements in the additional requirements information and the review templates in the template configuration information are injected into the third intelligent agent to obtain the review intelligent agent; The summary requirement in the additional requirement information and the summary template in the template configuration information are injected into the fourth intelligent agent to obtain the summary intelligent agent.

[0007] Optionally, the process configuration information includes at least one of the following: adding an opening statement and knowledge presentation segment, adding learning objectives for answering questions, providing feedback after each round of defense, and allowing users to ask questions freely or engage in casual conversation during the answering process.

[0008] Optionally, the control of the main intelligent agent involves multiple rounds of dialogue with the user based on the question-answering learning process, and during these multiple rounds of dialogue, the corresponding auxiliary intelligent agent is invoked to enable the user to complete the question-answering learning task under the target service scenario, including: After each round of conversation, the main agent is controlled to record the current question-answering learning state, and based on the current question-answering learning state and the question-answering learning process, the corresponding auxiliary agent is invoked in the next round of conversation to enable the user to complete the question-answering learning task under the target service scenario. The current question-answering learning state includes the current question-answering learning stage, the conversation context, the user's question-answering status, and the number of questions that the user has answered.

[0009] Optionally, when the target service scenario is a learning scenario, and the process configuration information includes adding an opening and knowledge demonstration segment, adding answering learning objectives, providing feedback after each round of defense, and allowing users to ask questions or engage in casual conversation during the answering process, the step of controlling the main agent to record the current answering learning state after each round of conversation, and based on the current answering learning state and the answering learning process, calling the corresponding auxiliary agent in the next round of conversation, includes: When a user enters the interactive interface for answering questions and learning, the main intelligent agent is controlled to display an opening message, a knowledge point outline, and the learning objectives on the interactive interface for answering questions and learning. When the user replies "start" on the question-and-answer learning interaction interface, the main intelligent agent is controlled to record the opening remarks and knowledge display segment, as well as the conversation context of the opening remarks and knowledge display segment. The main intelligent agent is then controlled to call the question-generating intelligent agent to generate and push questions to the user based on the question-and-answer learning process and the opening remarks and knowledge display segment. When the user asks a question or engages in casual conversation about the pushed topic, the main AI agent is controlled to respond to the user's question or conversation. After the user answers the pushed question, the main agent records the question-generating process, the corresponding conversation context, and the number of questions the user has answered. The main agent then calls the review agent to review the user's answer based on the question-answering learning process and the question-generating process. The user's answering behavior is recorded by the main agent as training trajectory data and stored in the database. The main intelligent agent is controlled to obtain the total number of questions from the preset question bank of the database; If the number of questions answered by the current user is not equal to the total number of questions, then the question-generating agent will continue to be invoked to generate and push questions to the user. If the number of questions answered by the current user is equal to the total number of questions, then the main agent is controlled to record the review session and the corresponding conversation context, and the main agent is controlled to call the summarizing agent to generate a learning summary and display it to the user based on the question-answering learning process and the review session.

[0010] Optionally, the control of the main intelligent agent, based on the question-answering learning process and the opening remarks and knowledge display stage, calls the question-generating intelligent agent to generate and push questions to the user, including: The main intelligent agent controls the question-generating intelligent agent to retrieve questions and options from the preset question bank based on the question-answering learning process and the opening remarks and knowledge display stage. The question-generating agent is invoked to filter the questions and options based on the question-generating requirements and the question-generating template, and then generates the push questions to be displayed to the user based on the filtered questions and options.

[0011] Optionally, the step of invoking the question-generating intelligent agent to filter the questions and options based on the question-generating requirements and the question-generating template, and generating the push question to be displayed to the user based on the filtered questions and options, includes: The main intelligent agent is controlled to call the question-generating intelligent agent to obtain historical training trajectory data from the database, and to perform data analysis based on the historical training trajectory data to obtain analysis results; The question-generating agent is invoked to filter the questions and options based on the analysis results, the question-generating requirements, and the question-generating template, and then generates the push questions to be displayed to the user based on the filtered questions and options.

[0012] Optionally, the control of the main intelligent agent, based on the question-answering learning process and the question-generating stage, calls the review intelligent agent to provide feedback on the user's answers, including: The main intelligent agent controls the main intelligent agent to retrieve standard answers and explanations from the preset question bank and retrieve historical training trajectory data from the database, based on the question-answering learning process and the question-generating stage. Based on the standard answer, the analysis, and the historical training trajectory data, the review agent is invoked to generate review information and display it to the user using the review requirements and the review template.

[0013] Optionally, the control of the main intelligent agent, based on the question-answering learning process and the review stage, calls the summarizing intelligent agent to generate a learning summary and display it to the user, including: The main intelligent agent sends the user's answer status, as well as the conversation context of the opening remarks and knowledge presentation, the conversation context of the question-setting, and the conversation context of the commentary to the summarizing intelligent agent. The main intelligent agent is controlled to call the summarizing intelligent agent to retrieve historical training trajectory data from the database, and to analyze the historical training trajectory data to obtain the analysis results; Based on the analysis results and the user's answers, as well as the conversation context of the opening remarks and knowledge presentation, the conversation context of the question-generating, and the conversation context of the commentary, the summary agent is invoked to generate the learning summary and display it to the user using the summary template and the summary requirements.

[0014] Optionally, the control mechanism involves the main agent invoking the summarizing agent to retrieve historical training trajectory data from the database, and analyzing the historical training trajectory data to obtain analysis results, including: The main intelligent agent is controlled to call the summarizing intelligent agent to obtain historical training trajectory data from the database, and based on the historical training trajectory data, the answer accuracy rate, question bank coverage rate and weak knowledge points are calculated respectively. The analysis results are determined based on the accuracy of the responses, the coverage of the question bank, and the weak knowledge points.

[0015] Optionally, when the target service scenario is an examination scenario, and the process configuration information includes adding an opening and knowledge demonstration segment, and adding a question-and-answer learning objective, the step of controlling the main agent to record the current question-and-answer learning state after each round of conversation, and based on the current question-and-answer learning state and the question-and-answer learning process, calling the corresponding auxiliary agent in the next round of conversation, includes: When a user enters the interactive interface for answering questions and learning, the main intelligent agent is controlled to display an opening message, a knowledge point outline, and the learning objectives on the interactive interface for answering questions and learning. When the user replies "start" on the question-and-answer learning interaction interface, the main intelligent agent is controlled to record the opening remarks and knowledge display segment, as well as the conversation context of the opening remarks and knowledge display segment. The main intelligent agent is then controlled to call the question-generating intelligent agent to generate and push questions to the user based on the question-and-answer learning process and the opening remarks and knowledge display segment. After the user answers the pushed question, the main intelligent agent is controlled to record the question-generating process, the conversation context corresponding to the question-generating process, and the number of questions that the user has answered. The main intelligent agent is controlled to obtain the total number of questions from a preset question bank in the database; If the number of questions answered by the current user is not equal to the total number of questions, then the question-generating agent will continue to be invoked to generate and push questions to the user. If the number of questions answered by the current user is equal to the total number of questions, then the main agent is controlled to call the summarizing agent to generate a learning summary and display it to the user, based on the question-answering learning process and the question-generating stage.

[0016] According to a second aspect of this application, a question-answering learning device based on a large model intelligent agent is provided, comprising: The acquisition unit is used to acquire configuration information for answering questions and learning in the target service scenario; The construction unit is used to construct a main intelligent agent and multiple auxiliary intelligent agents based on the process configuration information, template configuration information and additional requirement information in the configuration information, wherein the main intelligent agent is used to control the question-answering learning process adapted to the target service scenario; The control unit is used to control the main intelligent agent to conduct multiple rounds of dialogue with the user according to the question-and-answer learning process when the user enters the question-and-answer learning interactive interface, and to call the corresponding auxiliary intelligent agent during the multiple rounds of dialogue so that the user can complete the question-and-answer learning task under the target service scenario.

[0017] Optionally, the construction unit is specifically used to inject the process configuration information and the scenario information and / or knowledge background in the additional requirement information into a first intelligent agent to obtain the main intelligent agent; to inject the question-generating requirement in the additional requirement information and the question-generating template in the template configuration information into a second intelligent agent to obtain the question-generating intelligent agent; to inject the review requirement in the additional requirement information and the review template in the template configuration information into a third intelligent agent to obtain the review intelligent agent; and to inject the summary requirement in the additional requirement information and the summary template in the template configuration information into a fourth intelligent agent to obtain the summary intelligent agent.

[0018] Optionally, the process configuration information includes at least one of the following: adding an opening statement and knowledge presentation segment, adding learning objectives for answering questions, providing feedback after each round of defense, and allowing users to ask questions freely or engage in casual conversation during the answering process.

[0019] Optionally, the control unit is specifically configured to control the main agent to record the current answer learning state after each round of conversation, and based on the current answer learning state and the answer learning process, to call the corresponding auxiliary agent in the next round of conversation so that the user can complete the answer learning task under the target service scenario. The current answer learning state includes the current answer learning stage, the conversation context, the user's answering status, and the number of questions that the user has answered.

[0020] Optionally, when the target service scenario is a learning scenario, and the process configuration information includes adding an opening and knowledge presentation segment, adding learning objectives for answering questions, providing feedback after each round of defense, and allowing users to ask questions or engage in casual conversation during the response process, the control unit includes: a first control module, a second control module, a third control module, a fourth control module, and a fifth control module. The first control module is used to control the main intelligent agent to display an opening, a knowledge point outline, and answer learning objectives on the question-and-answer learning interactive interface when the user enters the interface. The second control module is used to control the main intelligent agent to record the opening remarks and knowledge display segment, as well as the conversation context of the opening remarks and knowledge display segment, after the user replies "start" on the question-and-answer learning interaction interface. The main intelligent agent is also used to control the main intelligent agent to call the question-generating intelligent agent to generate and push questions to the user based on the question-and-answer learning process and the opening remarks and knowledge display segment. The third control module is used to control the main intelligent agent to respond to the user's free questions or casual knowledge chat when the user asks questions or engages in knowledge chat about the pushed questions. The fourth control module is used to control the main intelligent agent to record the question-generating stage, the corresponding conversation context, and the number of questions that the user has answered after the user answers the pushed question. The main intelligent agent is also used to call the review intelligent agent to review the user's answer based on the answer learning process and the question-generating stage. The user's answer is recorded by the main intelligent agent as training trajectory data and stored in the database. The second control module is also used to control the main intelligent agent to obtain the total number of questions from the preset question bank of the database; The fifth control module is used to continue to call the question-generating agent to generate and push questions to the user if the number of questions answered by the current user is not equal to the total number of questions. The fifth control module is further configured to, if the number of questions answered by the current user is equal to the total number of questions, control the main agent to record the review session and the corresponding conversation context of the review session, and control the main agent to call the summarizing agent to generate a learning summary and display it to the user based on the question-answering learning process and the review session.

[0021] Optionally, the second control module includes: a first control submodule and a generation submodule. The first control submodule is used to control the main intelligent agent to call the question-generating intelligent agent to obtain questions and options from the preset question bank according to the question-answering learning process and the opening remarks and knowledge display stage; The generation submodule is used to call the question-generating agent to filter the questions and options based on the question-generating requirements and the question-generating template, and generate the push questions to be displayed to the user based on the filtered questions and options.

[0022] Optionally, the generation submodule is specifically used to control the main agent to call the question-generating agent to obtain historical training trajectory data from the database, and to perform data analysis based on the historical training trajectory data to obtain analysis results; to call the question-generating agent to filter the questions and options based on the analysis results, the question-generating requirements, and the question-generating template, and to generate the push questions to be displayed to the user based on the filtered questions and options.

[0023] Optionally, the fourth control module is specifically used to control the main intelligent agent to call the review intelligent agent to obtain standard answers and explanations from the preset question bank and obtain historical training trajectory data from the database, based on the question-answering learning process and the question-generating stage; and to call the review intelligent agent to generate review information and display it to the user based on the standard answers, the explanations and the historical training trajectory data, using the review requirements and the review template.

[0024] Optionally, the fifth control module includes: a second control submodule, a third control submodule, and a fourth control submodule. The second control submodule is used to control the main agent to send the user's answer status, as well as the conversation context of the opening remarks and knowledge display, the conversation context of the question-setting, and the conversation context of the commentary to the summarizing agent; The third control submodule is used to control the main agent to call the summarizing agent to obtain historical training trajectory data from the database, and to analyze the historical training trajectory data to obtain analysis results; The fourth control submodule is used to, based on the analysis results and the user's answer status, as well as the conversation context of the opening remarks and knowledge display stage, the conversation context corresponding to the question-generating stage, and the conversation context corresponding to the commentary stage, call the summary agent to generate the learning summary and display it to the user using the summary template and the summary requirements.

[0025] Optionally, the third control submodule is specifically used to control the main agent to call the summarizing agent to obtain historical training trajectory data from the database, and based on the historical training trajectory data, to respectively calculate the response accuracy, question bank coverage, and weak knowledge points; and to determine the analysis results based on the response accuracy, the question bank coverage, and the weak knowledge points.

[0026] Optionally, when the target service scenario is an examination scenario, and the process configuration information includes adding an opening and knowledge display segment, and adding answer learning objectives, the control unit includes: a sixth control module, a seventh control module, an eighth control module, and a ninth control module. The sixth control module is used to control the main intelligent agent to display the opening remarks, knowledge point outline and answer learning objectives on the question-and-answer learning interaction interface when the user enters the question-and-answer learning interaction interface. The seventh control module is used to control the main intelligent agent to record the opening remarks and knowledge display segment, as well as the conversation context of the opening remarks and knowledge display segment, after the user replies to start on the question-and-answer learning interaction interface. The main intelligent agent is also used to control the main intelligent agent to call the question-generating intelligent agent to generate and push questions to the user based on the question-and-answer learning process and the opening remarks and knowledge display segment. The eighth control module is used to control the main intelligent agent to record the question-generating stage, the conversation context corresponding to the question-generating stage, and the number of questions that the user has answered at present after the user answers the pushed question. The seventh control module is also used to control the main intelligent agent to obtain the total number of questions from the preset question bank of the database; The ninth control module is used to continue calling the question-generating agent to generate and push questions to the user if the number of questions answered by the current user is not equal to the total number of questions. The ninth control module is further configured to, if the number of questions answered by the current user is equal to the total number of questions, control the main intelligent agent to call the summarizing intelligent agent to generate a learning summary and display it to the user, based on the question-answering learning process and the question-generating stage.

[0027] According to a third aspect of this application, a storage medium is provided on which a computer program is stored, which, when executed by a processor, implements the above-described question-answering learning method based on a large model intelligent agent.

[0028] According to a fourth aspect of this application, a computer device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor executes the computer program to implement the above-described question-answering learning method based on a large model intelligent agent.

[0029] By employing the above technical solution, this application provides a question-answering learning method and apparatus based on a large-scale intelligent agent. Compared with existing technologies, this method constructs a main intelligent agent and multiple auxiliary intelligent agents through process configuration information, template configuration information, and additional requirement information in the configuration information. This enables the main intelligent agent to control the question-answering learning process that matches the service scenario. Simultaneously, based on a natural language interaction mode, the main intelligent agent is controlled to conduct multiple rounds of dialogue with the user on the question-answering learning interaction interface. During these multiple rounds of dialogue, the corresponding auxiliary intelligent agents are invoked according to the question-answering learning process, enabling the user to complete the question-answering learning task in the corresponding service scenario. Since the question-answering learning process of this application can be flexibly changed through the main intelligent agent, and the user can freely interact with the system during the question-answering learning process, the question-answering learning process of this application is more flexible and personalized, which can meet the learning needs of users in different service scenarios, and also enhances the user's interaction ability with the system during the question-answering learning process.

[0030] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0031] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 The diagram illustrates a flowchart of a question-answering learning method based on a large model agent provided in an embodiment of this application. Figure 2 A schematic diagram of the question-answering learning system framework provided in an embodiment of this application is shown; Figure 3 This illustration shows a data flow diagram provided in an embodiment of this application; Figure 4 This illustration shows a question-answering learning process in a learning scenario provided by an embodiment of this application; Figure 5 This illustration shows a schematic diagram of the answering and learning process in an examination scenario provided by an embodiment of this application; Figure 6 This document illustrates a training process example provided in an embodiment of this application. Figure 7 A schematic diagram of background knowledge in the interactive question-and-answer learning interface provided in an embodiment of this application is shown; Figure 8 This illustration shows a schematic diagram of the question-and-answer learning interactive interface provided in an embodiment of this application, which includes the pushing of questions and the option to ask questions freely; Figure 9 This illustration shows a diagram of the comment content in the interactive question-and-answer learning interface provided in an embodiment of this application; Figure 10 This illustration shows a schematic diagram of the summarized content in the interactive question-and-answer learning interface provided in an embodiment of this application; Figure 11 The diagram shows a structural schematic of a question-answering learning device based on a large model intelligent agent provided in an embodiment of this application. Detailed Implementation

[0032] The present application will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of the present application can be combined with each other.

[0033] The fixed question-setting and assessment model of existing technologies means that users can only answer questions according to the same process in all scenarios, which lacks flexibility and makes it difficult to meet the learning needs of users in different scenarios.

[0034] To overcome the aforementioned technical problems, this embodiment provides a question-answering learning method based on a large model intelligent agent, such as... Figure 1 As shown, the method includes: Step 10: Obtain the configuration information for answering questions and learning in the target service scenario.

[0035] The target service scenario can be any of the following: safety compliance learning and examinations, disciplinary system learning and examinations, policy and regulation learning and examinations, service skills learning and examinations, special job learning and examinations, qualification and on-the-job learning and examinations in industries such as banking, hospitals, schools, and enterprises. Furthermore, the configuration information is provided by a user with a different identity than the user taking the quiz. For example, if the user taking the quiz is a trainee in banking service skills training, the person providing the configuration information is the trainer of banking service skills training. The configuration information specifically includes: process configuration information, template configuration information, and additional requirements information.

[0036] To address the lack of flexibility in the question-answering learning process of existing intelligent question-answering systems, embodiments of the present invention provide a system architecture for question-answering learning, such as... Figure 2 As shown, the system architecture includes a presentation layer, an agent service layer, a tool service layer, and a data layer. The presentation layer specifically includes a question-and-answer learning interaction interface and API service interfaces. The question-and-answer learning interaction interface is essentially a Language User Interface (LUI), through which users can interact with the system. The presentation layer is mainly used for natural language interaction and interfacing with external systems. The agent service layer includes a main agent and multiple auxiliary agents, including a question-generating agent, a commenting agent, and a summarizing agent. The main agent is primarily responsible for interacting with the user and controlling the question-and-answer learning process, while the auxiliary agents assist the main agent in collaboratively completing tasks. The tool service layer includes data tools and computing tools. Data tools support the interaction between auxiliary agents and the database, such as obtaining question bank information and recording user answers. Computing tools support the auxiliary agents in analyzing user answers and calculating relevant metrics. The data layer includes a question bank, user historical answer data (historical training trajectory), and configuration information. The data layer provides input, historical records, and personalization support for the overall system.

[0037] Based on the above system architecture, the overall data flow of this embodiment of the invention is as follows: Figure 3 As shown, the question-answering learning system of this invention consists of a data layer, a tool service layer, an intelligent agent service layer, and a language user interface (presentation layer) connected in sequence. Under the scheduling of the main intelligent agent, multiple auxiliary intelligent agents cooperate to complete the task. Figure 3The functions of each module are as follows: The question bank stores structured questions, options, standard answers, and explanations, providing basic data for the system to generate and analyze questions. The question bank is pre-uploaded by users or generated by connecting to an intelligent question generation system; the training trajectory records users' answering performance, including answering time, correctness of answers, and answering strategies. It is mainly used for targeted question generation to optimize users' learning paths and for subsequent summary and analysis; configuration information is provided by users with different identities than the users answering questions, specifically including: process configuration information, template configuration information, and additional requirement information; the main intelligent agent is the core control unit of the system, used to interact with users and responsible for control. The system establishes a question-and-answer learning process, invoking corresponding auxiliary agents. The question-generating agent uses data tools to retrieve data and options from the question bank, generating and pushing questions. The feedback agent uses data tools to retrieve answers and explanations from the question bank, providing feedback on the user's responses. The summary agent uses data tools to retrieve the user's training trajectory from the database and uses computational tools to analyze the training trajectory, generating a learning report summary. The question-and-answer learning interactive interface (language user interface) is used for user learning to answer questions. In the language user interface, the main agent acts as an assistant, interacting with the user in multiple rounds, including pushing questions, receiving answers, providing explanations and feedback.

[0038] This system architecture of the present invention enables an end-to-end closed loop from information configuration, process instance generation, multi-round interactive question answering to summary. Secondly, it supports multi-round free questioning and knowledge expansion during the question-answering process, achieving a deep integration of examination and learning. Furthermore, in the system architecture of this invention, each functional module can be executed independently by each intelligent agent, thereby reducing interference from single-model multi-task operations and ensuring content quality and logical stability. In addition, the system architecture of this invention is more personalized and adaptable; that is, by analyzing the user's historical training trajectory, questions can be adjusted and learning paths optimized.

[0039] In specific application scenarios, before a user begins answering questions and learning, a user with a different identity pre-fills configuration information into the system. This configuration information, after being retrieved by the system, is stored in the database. Specifically, this configuration information includes process configuration information, template configuration information, and additional requirement information. For example, if the user answering questions is trainee B in banking service skills training, and the configuration information is provided by trainer A in banking service skills training, trainer A will pre-enter the configuration information into the system.

[0040] The process configuration information specifically includes: whether to add an opening and knowledge presentation segment, whether to add learning objectives for answering questions, whether to provide feedback after each round of defense, and whether to allow users to ask questions or engage in casual conversation during the response process. The process configuration information varies in different service scenarios. For example, in a learning scenario, the process configuration information includes adding an opening and knowledge presentation segment, adding learning objectives for answering questions, providing feedback after each round of defense, and allowing users to ask questions or engage in casual conversation during the response process. In an examination scenario, the process configuration information only includes adding an opening and knowledge presentation segment and adding learning objectives for answering questions, but does not provide feedback to users, nor does it allow users to ask questions or engage in casual conversation. The template configuration specifically includes question templates, feedback templates, and summary templates. Question templates include constraints on the format of the questions, question creation methods, question types, specific knowledge points, and difficulty range. Feedback templates include specific formats, expectations, level of detail, and methods of knowledge supplementation for feedback. Summary templates include the format, structure, and key content of the summary report. Question templates, feedback templates, and summary templates can be configured according to the user's actual service needs. Additional requirements include scenario information, background knowledge, question-generating requirements, commentary requirements, and summary requirements. Scenario information and background knowledge can specifically include industry cases and job-related background. Adding scenario information and background knowledge can enhance the relevance of the questions to the service content. Users can add additional requirements for presentation, question generation, commentary, and summary in the additional requirements information. For example, question generation requirements include question prediction strategies and answer modification rules. When the main agent calls the question generation agent, commentary agent, and summary agent to generate questions, provide commentary, and summarize, it will process the additional requirements information according to the user.

[0041] Step 20: Based on the process configuration information, template configuration information and additional requirement information in the configuration information, construct the main intelligent agent and multiple auxiliary intelligent agents respectively.

[0042] The process configuration information includes at least one of the following: adding an opening and knowledge presentation segment, adding learning objectives for answering questions, providing feedback after each round of Q&A, and allowing users to ask questions or engage in casual conversation during the answering process. The main intelligent agent controls the question-and-answer learning process adapted to the target service scenario and interacts with the user. Multiple auxiliary intelligent agents include a question-generating agent, a feedback agent, and a summarizing agent. These agents primarily assist the main intelligent agent in enabling users to complete the question-and-answer learning task. Both the main intelligent agent and the multiple auxiliary agents are based on a large language model.

[0043] After obtaining the configuration information, the question-answering learning system of this invention injects the configuration information into the corresponding intelligent agent, thereby constructing a main intelligent agent and multiple auxiliary intelligent agents. The configuration information is used as parameters for the main intelligent agent and auxiliary intelligent agents to perform tasks, thereby ensuring that the interactive content, question generation, comments and summaries meet specific needs.

[0044] The embodiments of the present invention enable the question-and-answer learning process to be more flexible and personalized through the process configuration step, thereby meeting the learning needs of users in different service scenarios, supporting process customization in different scenarios, and enhancing system adaptability.

[0045] When constructing the main intelligent agent, the question-generating intelligent agent, the review intelligent agent, and the auxiliary intelligent agent, the process configuration information, as well as the scenario information and / or knowledge background in the additional requirement information, are injected into the first intelligent agent to obtain the main intelligent agent; the question-generating requirement in the additional requirement information, as well as the question-generating template in the template configuration information, are injected into the second intelligent agent to obtain the question-generating intelligent agent; the review requirement in the additional requirement information, as well as the review template in the template configuration information, are injected into the third intelligent agent to obtain the review intelligent agent; and the summary requirement in the additional requirement information, as well as the summary template in the template configuration information, are injected into the fourth intelligent agent to obtain the summary intelligent agent.

[0046] Specifically, if the additional requirements information includes scenario information and / or background knowledge, the process configuration information, scenario information, and / or background knowledge are injected into the first intelligent agent to construct the main intelligent agent. When the user enters the question-and-answer learning interaction interface, the main intelligent agent will display the scenario information and / or background knowledge to the user and conduct multiple rounds of dialogue with the user according to the question-and-answer learning process corresponding to the process configuration information. Furthermore, if the additional requirements information includes question-generating requirements, commenting requirements, and summarizing requirements, these requirements, along with their templates, will be injected into the corresponding intelligent agents to construct question-generating, commenting, and summarizing intelligent agents. Under the scheduling of the main intelligent agent, the question-generating, commenting, and summarizing intelligent agents will generate questions, provide commentary, and summarize according to the corresponding templates and requirements.

[0047] In this embodiment of the invention, each functional step is executed by an independent intelligent agent, which can reduce interference from multiple tasks in a single model, thereby enabling specialized division of labor, improving the processing accuracy of each step, and ensuring content quality and logical stability.

[0048] Step 30: When the user enters the question-and-answer learning interaction interface, the main intelligent agent is controlled to conduct multiple rounds of dialogue with the user according to the question-and-answer learning process. During the multiple rounds of dialogue, the corresponding auxiliary intelligent agent is called to enable the user to complete the question-and-answer learning task under the target service scenario.

[0049] In this embodiment of the invention, after each round of conversation, the main intelligent agent is controlled to record the current question-answering learning state, and based on the current question-answering learning state and the question-answering learning process, the corresponding auxiliary intelligent agent is invoked in the next round of conversation to enable the user to complete the question-answering learning task under the target service scenario. The current question-answering learning state includes the current question-answering learning stage, the conversation context, the user's question-answering status, and the number of questions the user has answered. The question-answering learning stage specifically includes an opening and knowledge presentation stage, a free question or knowledge chat stage, a question-setting stage, a commentary stage, and a summary stage.

[0050] This invention, by recording the current question-answering learning state and, based on the current question-answering learning state and the question-answering learning process, calls the corresponding auxiliary intelligent agent in the next round of the session, thereby achieving cross-round context continuity and avoiding repeated interactions.

[0051] In a learning scenario, users are allowed to ask questions freely or engage in casual conversation, with feedback provided after each answer. The process configuration can include adding an opening and knowledge demonstration, setting learning objectives, providing feedback after each round of responses, and allowing users to ask questions or engage in casual conversation during the answering process. The scheduling process for the main agent and multiple auxiliary agents is as follows: Figure 4 As shown, it includes: Step 31: When the user enters the question-and-answer learning interaction interface, control the main intelligent agent to display the opening remarks, knowledge point outline and question-and-answer learning objectives on the question-and-answer learning interaction interface.

[0052] Specifically, the learning objectives for answering questions can include the accuracy rate and the number of questions answered by the user.

[0053] In the embodiments of the present invention, in a learning scenario, if the process configuration information includes adding an opening statement and a knowledge display section, and adding a question-and-answer learning objective, then after the user enters the question-and-answer learning interaction interface (language user interface), the system will control the main intelligent agent to prioritize displaying the opening statement, knowledge point outline, and question-and-answer learning objective to the user according to the question-and-answer learning process corresponding to the process configuration information.

[0054] Step 32: After the user replies "start" on the question-and-answer learning interaction interface, control the main intelligent agent to record the opening remarks and knowledge display segment, as well as the conversation context of the opening remarks and knowledge display segment. Then, control the main intelligent agent to call the question-generating intelligent agent to generate and push questions to the user based on the question-and-answer learning process and the opening remarks and knowledge display segment.

[0055] In this embodiment of the invention, after the user begins to reply, the main AI agent records the current question-and-answer learning stage (opening remarks and knowledge display stage) and the conversation content with the user. Then, based on the question-and-answer learning process and the current opening remarks and knowledge display stage, the main AI agent determines that the next stage is the question-generating stage. Then, the main AI agent calls the question-generating AI agent to generate and push questions for the user to answer.

[0056] When generating push questions, the main intelligent agent, based on the question-answering learning process and the opening remarks and knowledge display stage, calls the question-generating intelligent agent to obtain questions and options from the preset question bank. Then, the question-generating intelligent agent is called to filter the questions and options based on the question-generating requirements and the question-generating template, and based on the filtered questions and options, the push question is generated and displayed to the user.

[0057] Specifically, the main intelligent agent can call the question-generating intelligent agent. The question-generating intelligent agent filters questions and options based on the question-generating method, question type, specific knowledge points and difficulty range in the question-generating template, and generates corresponding push questions according to the question display format in the question-generating template. These questions are then displayed in the answer learning interaction interface (language user interface) for users to answer.

[0058] In some embodiments, the question-generating agent can also analyze the user's historical answer history to generate targeted questions, thereby optimizing the user's learning path. Based on this, the method further includes: controlling the main agent to call the question-generating agent to obtain historical training trajectory data from the database, and performing data analysis based on the historical training trajectory data to obtain analysis results; calling the question-generating agent to filter the questions and options based on the analysis results, the question-generating requirements, and the question-generating template, and generating the pushed questions to be displayed to the user based on the filtered questions and options.

[0059] Specifically, the controlling agent calls the question-generating agent to retrieve historical training trajectory data from the database. The question-generating agent analyzes and calculates the historical training trajectory data to obtain the user's answer accuracy, question bank coverage, and weak knowledge points. Then, based on the user's answer accuracy, question bank coverage, and weak knowledge points, the question-generating agent generates more targeted questions, thereby optimizing the user's learning path.

[0060] Step 33: When the user asks a question or engages in casual conversation about the pushed topic, the main AI agent is controlled to respond to the user's question or conversation. In the embodiments of the present invention, in a learning scenario, users can freely ask questions or engage in casual conversation about the pushed questions according to the question-and-answer learning process. The main intelligent agent can conduct multiple rounds of dialogue with the user during this stage, answer the user's questions, and then guide the user to continue answering the pushed questions on the question-and-answer learning interactive interface (language user interface).

[0061] Step 34: After the user answers the pushed question, the main agent is controlled to record the question-generating stage, the corresponding conversation context, and the number of questions the user has answered. The main agent is then controlled to call the review agent to review the user's answer based on the answer learning process and the question-generating stage. The user's answer is recorded by the main agent as training trajectory data and stored in the database.

[0062] In this embodiment of the invention, after a user answers a question on the interactive learning interface, the main AI will invoke a review AI to provide feedback on the user's response. Based on this, step 34 specifically includes: controlling the main AI to, according to the question-answering learning process and the question-generating stage, invoke the review AI to obtain standard answers and explanations from the preset question bank, and to obtain historical training trajectory data from the database; based on the standard answers, the explanations, and the historical training trajectory data, invoking the review AI to generate review information using the review requirements and the review template, and displaying it to the user.

[0063] Specifically, after the user responds, the controlling agent records the current question-and-answer learning stage (question generation stage), the context of the question generation stage, and the number of questions the user has already answered. Then, based on the question-and-answer learning process and the question generation stage, the controlling agent determines the next stage to be the feedback stage. Next, the controlling agent calls the feedback agent, which generates feedback information for the user's response based on the standard answer, explanation, and historical training trajectory data. Simultaneously, the controlling agent stores the user's current answer status as training trajectory data in the database. This current answer status includes the answer time, whether the answer is correct, and the answering strategy.

[0064] Step 35: Control the main intelligent agent to obtain the total number of questions from the preset question bank of the database.

[0065] In this embodiment of the invention, the total number of questions can be stored in a preset question bank, and the main intelligent agent can obtain the total number of questions from the preset question bank.

[0066] Step 36a: If the number of questions answered by the current user is not equal to the total number of questions, then the question-generating agent is invoked again to generate and push questions to the user.

[0067] In this embodiment of the invention, if the number of questions the current user has answered is not equal to the total number of questions, it means the user has not yet completed all the questions. In this case, the question-generating agent is invoked again to generate and push questions for the user to answer. After each question is generated and the user answers it, the review agent is invoked to provide feedback.

[0068] Step 36b: If the number of questions answered by the current user is equal to the total number of questions, then control the main agent to record the review session and the corresponding conversation context, and control the main agent to call the summarizing agent to generate a learning summary and display it to the user based on the question-answering learning process and the review session.

[0069] In this embodiment of the invention, if the number of questions answered by the current user is equal to the total number of questions, it means that the user has completed all the questions. At this time, the main intelligent agent records the current question-answering learning stage (commenting stage) and the corresponding conversation context. Then, based on the question-answering learning process and the commenting stage, the main intelligent agent determines that the next stage is the summary stage. Subsequently, the main intelligent agent calls the summary intelligent agent to generate a learning summary and display it to the user.

[0070] It should be noted that, in addition to completing all the questions, users can also actively enter "end training" in the question-and-answer learning interaction interface. At this time, the question-and-answer session will also end, and the summary agent will be invoked to generate a learning summary to be displayed to the user.

[0071] When generating a learning summary, the main agent sends the user's answers, the conversation context of the opening remarks and knowledge presentation, the conversation context of the question-setting, and the conversation context of the commentary to the summary agent. The main agent then calls the summary agent to retrieve historical training trajectory data from the database and analyzes the data to obtain analysis results. Based on the analysis results, the user's answers, the conversation context of the opening remarks and knowledge presentation, the conversation context of the question-setting, and the conversation context of the commentary, the summary agent uses the summary template and the summary requirements to generate the learning summary and display it to the user.

[0072] When analyzing training data, the main agent is controlled to call the summarizing agent to obtain historical training trajectory data from the database, and based on the historical training trajectory data, the response accuracy, question bank coverage, and weak knowledge points are calculated respectively; the analysis results are determined according to the response accuracy, the question bank coverage, and the weak knowledge points.

[0073] In this embodiment of the invention, multiple rounds of natural language interaction and knowledge supplementation are interspersed during the question-and-answer learning process, and the question types and difficulty are dynamically adjusted according to the training trajectory data, thereby forming a truly integrated testing and learning process.

[0074] In an exam-oriented service scenario, users are not allowed to ask questions or engage in casual conversation. No feedback is provided after each user's answer. The process configuration can include adding an opening and knowledge demonstration, and setting learning objectives. In this case, the scheduling process for the main agent and multiple auxiliary agents is as follows: Figure 5 As shown, it includes: Step 36: When the user enters the question-and-answer learning interaction interface, control the main intelligent agent to display the opening remarks, knowledge point outline and question-and-answer learning objectives on the question-and-answer learning interaction interface.

[0075] Specifically, the learning objectives for answering questions can include the accuracy rate and the number of questions answered by the user.

[0076] In an embodiment of the present invention, in an examination scenario, if the process configuration information includes adding an opening statement and a knowledge display section, and adding answer learning objectives, then after the user enters the answer learning interaction interface (language user interface), the system will control the main intelligent agent to prioritize displaying the opening statement, knowledge outline, and answer learning objectives to the user according to the answer learning process corresponding to the process configuration information.

[0077] Step 37: After the user replies "start" on the question-and-answer learning interaction interface, control the main intelligent agent to record the opening remarks and knowledge display segment, as well as the conversation context of the opening remarks and knowledge display segment, and control the main intelligent agent to call the question-generating intelligent agent to generate and push questions to the user based on the question-and-answer learning process and the opening remarks and knowledge display segment.

[0078] In this embodiment of the invention, after the user begins to reply, the main AI agent records the current question-and-answer learning stage (opening remarks and knowledge display stage) and the conversation content with the user. Then, based on the question-and-answer learning process and the current opening remarks and knowledge display stage, the main AI agent determines that the next stage is the question-generating stage. Then, the main AI agent calls the question-generating AI agent to generate and push questions to the user for answering.

[0079] When generating push questions, the main intelligent agent, based on the question-answering learning process and the opening remarks and knowledge display stage, calls the question-generating intelligent agent to obtain questions and options from the preset question bank. Then, the question-generating intelligent agent is called to filter the questions and options based on the question-generating requirements and the question-generating template, and based on the filtered questions and options, the push question is generated and displayed to the user.

[0080] Specifically, the main intelligent agent can call the question-generating intelligent agent. The question-generating intelligent agent filters questions and options based on the question-generating method, question type, specific knowledge points and difficulty range in the question-generating template, and generates corresponding push questions according to the question display format in the question template. These questions are then displayed in the answer learning interaction interface (language user interface) for users to answer.

[0081] Step 38: After the user answers the pushed question, control the main agent to record the question-generating process, the conversation context corresponding to the question-generating process, and the number of questions that the user has answered.

[0082] Step 39: Control the main intelligent agent to obtain the total number of questions from the preset question bank in the database.

[0083] In this embodiment of the invention, the total number of questions can be stored in a preset question bank, and the main intelligent agent can obtain the total number of questions from the preset question bank.

[0084] Step 310a: If the number of questions answered by the current user is not equal to the total number of questions, then the question-generating agent is invoked again to generate and push questions to the user.

[0085] In this embodiment of the invention, if the number of questions answered by the current user is not equal to the total number of questions, it means that the user has not completed all the questions. In this case, the question-generating agent is called again to generate and push questions for the user to answer.

[0086] Step 310b: If the number of questions answered by the current user is equal to the total number of questions, then control the main agent to call the summarizing agent to generate a learning summary and display it to the user, based on the question-answering learning process and the question-generating stage.

[0087] In this embodiment of the invention, if the number of questions answered by the current user is equal to the total number of questions, it means that the user has completed all the questions. At this time, the main intelligent agent records the current question-answering learning stage (question-generating stage) and the corresponding conversation context. Then, based on the question-answering learning process and the question-generating stage, the main intelligent agent determines that the next stage is the summary stage. Subsequently, the main intelligent agent calls the summary intelligent agent to generate a learning summary and display it to the user.

[0088] When generating a learning summary, the main agent sends the user's answers, the conversation context of the opening remarks and knowledge presentation, and the conversation context of the question-generating phase to the summary agent. The main agent then calls the summary agent to retrieve historical training trajectory data from the database and performs combined analysis on the historical training trajectory data to obtain analysis results. Based on the analysis results, the conversation context of the opening remarks and knowledge presentation, and the conversation context of the question-generating phase, the summary agent is invoked to generate the learning summary and display it to the user using the summary template and the summary requirements.

[0089] When analyzing training data, the main agent is controlled to call the summarizing agent to obtain historical training trajectory data from the database, and based on the historical training trajectory data, the response accuracy, question bank coverage, and weak knowledge points are calculated respectively; the analysis results are determined according to the response accuracy, the question bank coverage, and the weak knowledge points.

[0090] To make the technical implementation of the embodiments of the present invention clearer, the specific implementation process of this solution will now be described in detail using a training scenario as an example. For the training scenario, users are divided into trainer users (organizers of training, hereinafter referred to as user A) and trainees (trainees, hereinafter referred to as user B).

[0091] like Figure 6 As shown, User A is primarily responsible for creating and configuring training tasks. First, User A imports a question bank into the quiz learning system. This question bank includes structured questions, selection options, standard answers, and explanations. Next, User A configures the quiz learning process. User A can flexibly configure the order and content of the entire quiz learning process, such as whether to add an opening and knowledge presentation segment, whether to add quiz learning objectives or the number of questions to be answered, whether to provide feedback after each round of presentations, and whether to allow users to ask questions freely or engage in casual conversation during the response process. Then, User A creates templates for different segments (such as question creation, feedback, and summary), specifically including question creation templates, feedback templates, and summary templates. User A can also add additional requirement information, such as background knowledge and scenario information. After user A configures the process configuration information, template configuration information, and additional requirements information, the question-and-answer learning system will construct a training scenario based on the process configuration information, template configuration information, and additional requirements information. That is, it will generate a process instance that can be directly used by user B to answer questions and learn. This process instance needs to be completed by the cooperation of the main intelligent agent, the question-generating intelligent agent, the commenting intelligent agent, and the summarizing intelligent agent.

[0092] like Figure 6As shown, when user B enters the system's question-and-answer learning interface, the main agent, acting as an assistant, guides the user into the learning environment, explains the learning objectives, and displays relevant background knowledge to prepare for subsequent question-and-answer sessions. Figure 7 As shown in the diagram. After the user begins replying, the main AI agent calls the question-generating AI agent to filter and select questions from the question bank, generate push questions, and display them to user B. Before replying, user B can freely ask questions and engage in multiple rounds of knowledge expansion exchanges with the main AI agent. The main AI agent will respond to the user's questions in real time, such as... Figure 8 As shown. After user B answers, the main AI will call the review AI to retrieve the standard answer and explanation from the question bank, and provide feedback on user B's answer. Then, it will continue to push the next question, as shown. Figure 9 As shown, simultaneously, user B's answers are stored in the database as training trajectory data. User B's answering and feedback process can be repeated N times. In an exam scenario, the number of questions is usually fixed, while in a learning scenario, it can be a multi-round loop, or a free-questioning session can be inserted during the feedback process, until the learning task ends or the user actively ends it. The main agent then calls the summarizing agent to generate a learning summary report, which is displayed to the user. The report includes specific details such as scores, shortcomings, and improvement suggestions. Figure 10 As shown.

[0093] The above-described question-and-answer learning process can be applied in training and learning scenarios; that is, the question-and-answer learning process includes... Figure 6 The solid and dashed boxes in the text refer to the learning process in training and examination scenarios, while the answering and learning process may only include... Figure 6 The solid-line box indicates that, through user A's flexible configuration in the process control section, the question-and-answer learning process can be adapted to user groups in different service scenarios.

[0094] It should be noted that the above-described question-and-answer learning process is only a usage example and can be modified according to the service scenario in actual applications.

[0095] This invention, through a collaborative mechanism between a primary and auxiliary intelligent agent, fully integrates question generation, real-time Q&A, process feedback, and learning summary. Trainers can inject process templates, reminder rules, evaluation criteria, and optional interactive elements all at once during the task construction phase, enabling natural language configuration. Compared to existing technologies, this invention not only generates and parses questions but also dynamically adjusts question types, difficulty, and knowledge supplementation methods during execution, thereby achieving true assessment-learning integration and adaptive learning.

[0096] This embodiment provides a question-answering learning method based on a large-scale intelligent agent. By configuring process information, template information, and additional requirement information in the configuration information, a main intelligent agent and multiple auxiliary intelligent agents are constructed. The main intelligent agent can control a question-answering learning process that matches the service scenario. Simultaneously, based on a natural language interaction mode, the main intelligent agent engages in multi-round conversations with the user on the question-answering learning interface. During these conversations, the main intelligent agent calls the corresponding auxiliary intelligent agents according to the question-answering learning process, enabling the user to complete the question-answering learning task in the corresponding service scenario. Because the question-answering learning process in this embodiment can be flexibly changed through the main intelligent agent, and the user can freely interact with the system during the learning process, the question-answering learning process in this embodiment is more flexible and personalized, meeting the learning needs of users in different service scenarios, and also enhancing the user's interaction ability with the system during the learning process.

[0097] Furthermore, as Figure 1 , Figure 4 and Figure 5 The specific implementation of the method shown in this embodiment provides a question-answering learning device based on a large model agent, such as... Figure 11 As shown, the device includes: an acquisition unit 101, a construction unit 102, and a control unit 103.

[0098] The acquisition unit 101 can be used to acquire configuration information for answering questions and learning in the target service scenario.

[0099] The construction unit 102 can be used to construct a main intelligent agent and multiple auxiliary intelligent agents based on the process configuration information, template configuration information and additional requirement information in the configuration information. The main intelligent agent is used to control the question-answering learning process adapted to the target service scenario.

[0100] The control unit 103 can be used to control the main intelligent agent to conduct multiple rounds of dialogue with the user according to the question-and-answer learning process when the user enters the question-and-answer learning interactive interface, and to call the corresponding auxiliary intelligent agent during the multiple rounds of dialogue so that the user can complete the question-and-answer learning task under the target service scenario.

[0101] In some embodiments, the construction unit 102 may be specifically used to inject the process configuration information and the scenario information and / or knowledge background in the additional requirement information into a first intelligent agent to obtain the main intelligent agent; inject the question-generating requirement in the additional requirement information and the question-generating template in the template configuration information into a second intelligent agent to obtain the question-generating intelligent agent; inject the commenting requirement in the additional requirement information and the commenting template in the template configuration information into a third intelligent agent to obtain the commenting intelligent agent; and inject the summary requirement in the additional requirement information and the summary template in the template configuration information into a fourth intelligent agent to obtain the summary intelligent agent.

[0102] In some embodiments, the process configuration information includes at least one of the following: adding an opening statement and knowledge presentation segment, adding learning objectives for answering questions, providing feedback after each round of defense, and allowing users to ask questions freely or engage in casual conversation during the response process.

[0103] In some embodiments, the control unit 103 is specifically configured to control the main agent to record the current answer learning state after each round of conversation, and based on the current answer learning state and the answer learning process, to call the corresponding auxiliary agent in the next round of conversation so that the user can complete the answer learning task under the target service scenario. The current answer learning state includes the current answer learning stage, the conversation context and the user's answering status, as well as the number of questions that the user has answered.

[0104] In some embodiments, when the target service scenario is a learning scenario, and the process configuration information includes adding an opening and knowledge display segment, adding learning objectives for answering questions, providing feedback after each round of defense, and allowing users to ask questions freely or engage in casual conversation during the response process, the control unit 103 includes: a first control module, a second control module, a third control module, a fourth control module, and a fifth control module.

[0105] The first control module can be used to control the main intelligent agent to display an opening, a knowledge point outline, and answer learning objectives on the question-and-answer learning interactive interface when the user enters the interface.

[0106] The second control module can be used to control the main intelligent agent to record the opening remarks and knowledge display segment, as well as the conversation context of the opening remarks and knowledge display segment, after the user replies to the start of the question-and-answer learning interaction interface. The main intelligent agent can also be controlled to call the question-generating intelligent agent to generate and push questions to the user based on the question-and-answer learning process and the opening remarks and knowledge display segment.

[0107] The third control module can be used to control the main intelligent agent to respond to the user's free questions or casual chat about the pushed topics.

[0108] The fourth control module can be used to control the main agent to record the question-generating stage, the corresponding conversation context, and the number of questions the user has answered after the user answers the pushed question. It can also control the main agent to call the review agent to review the user's answer based on the answer learning process and the question-generating stage. The user's answering status is recorded by the main agent as training trajectory data and stored in the database.

[0109] The second control module can also be used to control the main intelligent agent to obtain the total number of questions from the preset question bank of the database.

[0110] The fifth control module can be used to continue calling the question-generating agent to generate and push questions to the user if the number of questions answered by the current user is not equal to the total number of questions.

[0111] The fifth control module can also be used to control the main agent to record the review session and the corresponding conversation context if the number of questions answered by the current user is equal to the total number of questions, and to control the main agent to call the summarizing agent to generate a learning summary and display it to the user based on the question-answering learning process and the review session.

[0112] In some embodiments, the second control module includes a first control submodule and a generation submodule.

[0113] The first control submodule can be used to control the main intelligent agent to call the question-generating intelligent agent to obtain questions and options from the preset question bank according to the question-answering learning process and the opening remarks and knowledge display stage.

[0114] The generation submodule can be used to call the question-generating agent to filter the questions and options based on the question-generating requirements and the question-generating template, and generate the push questions to be displayed to the user based on the filtered questions and options.

[0115] In some embodiments, the generation submodule may be specifically used to control the main agent to call the question-generating agent to obtain historical training trajectory data from the database, and to perform data analysis based on the historical training trajectory data to obtain analysis results; to call the question-generating agent to filter the questions and options based on the analysis results, the question-generating requirements and the question-generating template, and to generate the push questions to be displayed to the user based on the filtered questions and options.

[0116] In some embodiments, the fourth control module is specifically used to control the main intelligent agent to call the review intelligent agent to obtain standard answers and explanations from the preset question bank and obtain historical training trajectory data from the database, based on the question-answering learning process and the question-generating stage; and to call the review intelligent agent to generate review information and display it to the user based on the standard answers, the explanations and the historical training trajectory data, using the review requirements and the review template.

[0117] In some embodiments, the fifth control module includes: a second control submodule, a third control submodule, and a fourth control submodule.

[0118] The second control submodule can be used to control the main agent to send the user's answer status, as well as the conversation context of the opening remarks and knowledge display, the conversation context of the question-setting, and the conversation context of the commentary to the summarizing agent.

[0119] The third control submodule can be used to control the main agent to call the summarizing agent to obtain historical training trajectory data from the database, and to analyze the historical training trajectory data to obtain analysis results.

[0120] The fourth control submodule can be used to, based on the analysis results and the user's answer status, as well as the conversation context of the opening remarks and knowledge display stage, the conversation context corresponding to the question-generating stage, and the conversation context corresponding to the commentary stage, call the summary agent to generate the learning summary and display it to the user using the summary template and the summary requirements.

[0121] In some embodiments, the third control submodule may be specifically used to control the main agent to call the summarizing agent to obtain historical training trajectory data from the database, and based on the historical training trajectory data, to respectively calculate the response accuracy, question bank coverage and weak knowledge points; and to determine the analysis results based on the response accuracy, the question bank coverage and the weak knowledge points.

[0122] In some embodiments, when the target service scenario is an examination scenario, and the process configuration information includes adding an opening and knowledge display segment and adding a question-answering learning objective, the control unit includes: a sixth control module, a seventh control module, an eighth control module, and a ninth control module.

[0123] The sixth control module can be used to control the main intelligent agent to display an opening, knowledge point outline and answer learning objectives on the question-and-answer learning interaction interface when the user enters the interface.

[0124] The seventh control module can be used to control the main intelligent agent to record the opening remarks and knowledge display segment, as well as the conversation context of the opening remarks and knowledge display segment, after the user replies to the start of the question-and-answer learning interaction interface. It can also control the main intelligent agent to call the question-generating intelligent agent to generate and push questions to the user based on the question-and-answer learning process and the opening remarks and knowledge display segment.

[0125] The eighth control module can be used to control the main intelligent agent to record the question-generating process, the conversation context corresponding to the question-generating process, and the number of questions that the user has answered at present after the user answers the pushed question.

[0126] The main intelligent agent is controlled to obtain the total number of questions from a preset question bank in the database; The ninth control module can be used to continue calling the question-generating agent to generate and push questions to the user if the number of questions answered by the current user is not equal to the total number of questions. The ninth control module can also be used to control the main intelligent agent to call the summarizing intelligent agent to generate a learning summary and display it to the user if the number of questions answered by the current user is equal to the total number of questions.

[0127] It should be noted that other corresponding descriptions of the functional units involved in the question-answering learning device based on a large model intelligent agent provided in this embodiment can be found in [reference]. Figure 1 , Figure 4 and Figure 5 The corresponding descriptions in [the document] will not be repeated here.

[0128] Based on the above, Figure 1 , Figure 4 and Figure 5 Accordingly, this application also provides a storage medium storing a computer program, which, when executed by a processor, implements the above-described method. Figure 1 , Figure 4 and Figure 5The example shown is a question-answering learning method based on a large model agent.

[0129] Based on this understanding, the technical solution of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as CD-ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause a computer device (such as personal computer, server, or network device, etc.) to execute the methods of various implementation scenarios of this application.

[0130] Based on the above, Figure 1 , Figure 4 and Figure 5 The method shown, and Figure 11 To achieve the above objectives, the present application also provides a computer device, specifically a tablet computer, smartphone, smartwatch, smart bracelet, or other network device, as shown in the virtual device embodiment. This computer device includes a storage medium and a processor; the storage medium stores a computer program; the processor executes the computer program to achieve the above-described objectives. Figure 1 , Figure 4 and Figure 5 The method for generating transparent images is shown.

[0131] Optionally, the aforementioned physical devices may also include a user interface, a network interface, a camera, radio frequency (RF) circuitry, sensors, audio circuitry, and a Wi-Fi module. The user interface may include a display screen, input units such as a keyboard, and optional user interfaces may also include USB interfaces, card reader interfaces, etc. The network interface may optionally include standard wired interfaces, wireless interfaces (such as Wi-Fi interfaces), etc.

[0132] Those skilled in the art will understand that the computer device structure provided in this embodiment does not constitute a limitation on such physical device, and may include more or fewer components, or combine certain components, or have different component arrangements.

[0133] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the two physical devices mentioned above, supporting the operation of information processing programs and other software and / or programs. The network communication module is used to enable communication between the various components within the storage medium, as well as communication with other hardware and software in the information processing physical device.

[0134] This invention, through process configuration information, template configuration information, and additional requirement information in the configuration information, constructs a main intelligent agent and multiple auxiliary intelligent agents. The main intelligent agent can control a question-and-answer learning process that matches the service scenario. Simultaneously, based on a natural language interaction mode, the main intelligent agent engages in multiple rounds of dialogue with the user on the question-and-answer learning interface. During these rounds, it calls the corresponding auxiliary intelligent agents according to the question-and-answer learning process, enabling the user to complete the question-and-answer learning task for the corresponding service scenario. Because the question-and-answer learning process in this invention can be flexibly modified through the main intelligent agent, and the user can freely interact with the system during the learning process, the question-and-answer learning process in this invention is more flexible and personalized, meeting the learning needs of users in different service scenarios and enhancing the user's interaction with the system during the learning process.

[0135] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of a preferred embodiment, and the modules or processes shown in the drawings are not necessarily essential for implementing this application. Those skilled in the art will understand that the modules in the apparatus of the embodiment can be distributed within the apparatus of the embodiment as described, or can be modified to be located in one or more apparatuses different from this embodiment. The modules of the above-described embodiment can be combined into one module, or further divided into multiple sub-modules.

[0136] The serial numbers in this application are for descriptive purposes only and do not represent the superiority or inferiority of any particular implementation scenario. The above disclosures are merely a few specific implementation scenarios of this application; however, this application is not limited thereto, and any variations conceived by those skilled in the art should fall within the protection scope of this application.

Claims

1. A question-answering learning method based on a large model intelligent agent, characterized in that, include: Obtain the configuration information for question-answering learning in the target service scenario; Based on the process configuration information, template configuration information, and additional requirement information in the configuration information, a main intelligent agent and multiple auxiliary intelligent agents are constructed respectively. The main intelligent agent is used to control the question-answering learning process adapted to the target service scenario. When a user enters the question-and-answer learning interaction interface, the main intelligent agent is controlled to conduct multiple rounds of dialogue with the user according to the question-and-answer learning process. During the multiple rounds of dialogue, the corresponding auxiliary intelligent agents are called to enable the user to complete the question-and-answer learning task under the target service scenario.

2. The method according to claim 1, characterized in that, The multiple auxiliary intelligent agents include a question-generating intelligent agent, a review intelligent agent, and a summary intelligent agent. Based on the process configuration information, template configuration information, and additional requirement information in the configuration information, a main intelligent agent and multiple auxiliary intelligent agents are constructed, including: The process configuration information, as well as the scenario information and / or knowledge background in the additional requirements information, are injected into the first intelligent agent to obtain the main intelligent agent; The question-generating requirements in the additional requirements information and the question-generating templates in the template configuration information are injected into the second intelligent agent to obtain the question-generating intelligent agent; The review requirements in the additional requirements information and the review templates in the template configuration information are injected into the third intelligent agent to obtain the review intelligent agent; The summary requirement in the additional requirement information and the summary template in the template configuration information are injected into the fourth intelligent agent to obtain the summary intelligent agent.

3. The method according to claim 1, characterized in that, The process configuration information includes at least one of the following: adding an opening statement and knowledge presentation segment, adding learning objectives for answering questions, providing feedback after each round of defense, and allowing users to ask questions freely or engage in casual conversation during the answering process.

4. The method according to claim 2, characterized in that, The control mechanism, which involves the main intelligent agent engaging in multiple rounds of dialogue with the user according to the question-answering learning process, and invoking corresponding auxiliary intelligent agents during these rounds of dialogue, enables the user to complete the question-answering learning task within the target service scenario. This includes: After each round of conversation, the main agent is controlled to record the current question-answering learning state, and based on the current question-answering learning state and the question-answering learning process, the corresponding auxiliary agent is invoked in the next round of conversation to enable the user to complete the question-answering learning task under the target service scenario. The current question-answering learning state includes the current question-answering learning stage, the conversation context, the user's question-answering status, and the number of questions that the user has answered.

5. The method according to claim 4, characterized in that, When the target service scenario is a learning scenario, and the process configuration information includes adding an opening and knowledge demonstration segment, adding quiz learning objectives, providing feedback after each round of defense, and allowing users to ask questions or engage in casual conversation during the response process, the process of controlling the main agent to record the current quiz learning state after each round of conversation, and based on the current quiz learning state and the quiz learning process, calling the corresponding auxiliary agent in the next round of conversation, including: When a user enters the interactive interface for answering questions and learning, the main intelligent agent is controlled to display an opening message, a knowledge point outline, and the learning objectives on the interactive interface for answering questions and learning. When the user replies "start" on the question-and-answer learning interaction interface, the main intelligent agent is controlled to record the opening remarks and knowledge display segment, as well as the conversation context of the opening remarks and knowledge display segment. The main intelligent agent is then controlled to call the question-generating intelligent agent to generate and push questions to the user based on the question-and-answer learning process and the opening remarks and knowledge display segment. When the user asks a question or engages in casual conversation about the pushed topic, the main AI agent is controlled to respond to the user's question or conversation. After the user answers the pushed question, the main agent records the question-generating process, the corresponding conversation context, and the number of questions the user has answered. The main agent then calls the review agent to review the user's answer based on the question-answering learning process and the question-generating process. The user's answering behavior is recorded by the main agent as training trajectory data and stored in the database. The main intelligent agent is controlled to obtain the total number of questions from the preset question bank of the database; If the number of questions answered by the current user is not equal to the total number of questions, then the question-generating agent will continue to be invoked to generate and push questions to the user. If the number of questions answered by the current user is equal to the total number of questions, then the main agent is controlled to record the review session and the corresponding conversation context, and the main agent is controlled to call the summarizing agent to generate a learning summary and display it to the user based on the question-answering learning process and the review session.

6. The method according to claim 5, characterized in that, The control mechanism, based on the question-answering learning process and the opening remarks and knowledge demonstration phase, calls the question-generating intelligence to generate and push questions to the user, including: The main intelligent agent controls the question-generating intelligent agent to retrieve questions and options from the preset question bank based on the question-answering learning process and the opening remarks and knowledge display stage. The question-generating agent is invoked to filter the questions and options based on the question-generating requirements and the question-generating template, and then generates the push questions to be displayed to the user based on the filtered questions and options.

7. The method according to claim 6, characterized in that, The step of invoking the question-generating intelligent agent to filter the questions and options based on the question-generating requirements and the question-generating template, and generating the push question to be displayed to the user based on the filtered questions and options, includes: The main intelligent agent is controlled to call the question-generating intelligent agent to obtain historical training trajectory data from the database, and to perform data analysis based on the historical training trajectory data to obtain analysis results; The question-generating agent is invoked to filter the questions and options based on the analysis results, the question-generating requirements, and the question-generating template, and then generates the push questions to be displayed to the user based on the filtered questions and options.

8. The method according to claim 5, characterized in that, The control agent, based on the question-answering learning process and the question-generating stage, invokes the review agent to provide feedback on the user's answers, including: The main intelligent agent controls the main intelligent agent to retrieve standard answers and explanations from the preset question bank and retrieve historical training trajectory data from the database, based on the question-answering learning process and the question-generating stage. Based on the standard answer, the analysis, and the historical training trajectory data, the review agent is invoked to generate review information and display it to the user using the review requirements and the review template.

9. The method according to claim 5, characterized in that, The control mechanism, based on the question-and-answer learning process and the feedback phase, calls the summarizing agent to generate a learning summary and display it to the user, including: The main intelligent agent sends the user's answer status, as well as the conversation context of the opening remarks and knowledge presentation, the conversation context of the question-setting, and the conversation context of the commentary to the summarizing intelligent agent. The main intelligent agent is controlled to call the summarizing intelligent agent to retrieve historical training trajectory data from the database, and to analyze the historical training trajectory data to obtain the analysis results; Based on the analysis results and the user's answers, as well as the conversation context of the opening remarks and knowledge presentation, the conversation context of the question-generating, and the conversation context of the commentary, the summary agent is invoked to generate the learning summary and display it to the user using the summary template and the summary requirements.

10. The method according to claim 9, characterized in that, The control mechanism calls the main agent to retrieve historical training trajectory data from the database, and analyzes the historical training trajectory data to obtain analysis results, including: The main intelligent agent is controlled to call the summarizing intelligent agent to obtain historical training trajectory data from the database, and based on the historical training trajectory data, the answer accuracy rate, question bank coverage rate and weak knowledge points are calculated respectively. The analysis results are determined based on the accuracy of the responses, the coverage of the question bank, and the weak knowledge points.

11. The method according to claim 4, characterized in that, When the target service scenario is an examination scenario, and the process configuration information includes adding an opening and knowledge demonstration segment, and adding a question-answering learning objective, the process involves controlling the main agent to record the current question-answering learning state after each round of conversation, and based on the current question-answering learning state and the question-answering learning process, invoking the corresponding auxiliary agent in the next round of conversation, including: When a user enters the interactive interface for answering questions and learning, the main intelligent agent is controlled to display an opening message, a knowledge point outline, and the learning objectives on the interactive interface for answering questions and learning. When the user replies "start" on the question-and-answer learning interaction interface, the main intelligent agent is controlled to record the opening remarks and knowledge display segment, as well as the conversation context of the opening remarks and knowledge display segment. The main intelligent agent is then controlled to call the question-generating intelligent agent to generate and push questions to the user based on the question-and-answer learning process and the opening remarks and knowledge display segment. After the user answers the pushed question, the main intelligent agent is controlled to record the question-generating process, the conversation context corresponding to the question-generating process, and the number of questions that the user has answered. The main intelligent agent is controlled to obtain the total number of questions from a preset question bank in the database; If the number of questions answered by the current user is not equal to the total number of questions, then the question-generating agent will continue to be invoked to generate and push questions to the user. If the number of questions answered by the current user is equal to the total number of questions, then the main agent is controlled to call the summarizing agent to generate a learning summary and display it to the user, based on the question-answering learning process and the question-generating stage.

Images