CN119537542A

CN119537542A - Interactive question-answering task processing method and related device based on multi-agent collaboration

Info

Publication number: CN119537542A
Application number: CN202411613560.4A
Authority: CN
Inventors: 黄际洲; 彭志洺; 赵世奇
Original assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Current assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Priority date: 2024-11-12
Filing date: 2024-11-12
Publication date: 2025-02-28

Abstract

The present disclosure provides an interactive question-and-answer task processing method based on multi-agent collaboration and a related device, which relates to the fields of artificial intelligence technology such as generative large language models and agents. The method includes: determining the target task to be solved according to the natural language input of the user; decomposing the target task into multiple sub-target tasks including at least an interactive question-and-answer task, and issuing each sub-target task to each sub-target agent including at least an interactive question-and-answer agent, and different sub-target agents are used to process different types of sub-tasks; controlling the interactive question-and-answer agent to present an interactive graphical user interface corresponding to the interactive question-and-answer task to the user, and presenting an answer result corresponding to the received feedback answer, and the interactive graphical user interface includes a question and an interactive answer area; controlling the associated sub-target agent among the multiple sub-target agents to generate a corresponding associated processing result based on the answer result output by the interactive question-and-answer agent, and the associated sub-target agent is a sub-target agent that depends on the answer result as part of the task input information; determining the target processing result corresponding to the target task based on the associated processing result. By applying this solution, better comprehensive task processing effects can be achieved at a lower overall cost, especially for interactive question-and-answer tasks that require users to give subjective answers to questions and verify the effects based on the answers. The interactive question-and-answer agent provides users with an interactive graphical interface that includes questions and interactive answer areas, which will enable users' task requirements to be met more quickly through interactive answering, thereby improving the task solving effect.

Description

Interactive question-answering task processing method and related device based on multi-agent cooperation

Technical Field

The disclosure relates to the technical field of data processing, in particular to the technical field of artificial intelligence such as a large language model and an intelligent agent, and especially relates to an interactive question-answering task processing method and device based on multi-intelligent agent cooperation, electronic equipment, a computer readable storage medium and a computer program product.

Background

With the rapid development and iteration of the generative large language model, it has a better ability to understand the needs of user inputs and give corresponding results.

In order to make the output result of the generative large language model more in accordance with the specific requirement, the generated large language model is taken as a base model, and the intelligent agent constructed by combining the preset role parameters appears.

How to solve the complex problem presented by the user by using the intelligent agent is still a technical problem to be solved urgently by those skilled in the art.

Disclosure of Invention

The embodiment of the disclosure provides an interactive question-answering task processing method, device, electronic equipment, computer readable storage medium and computer program product based on multi-agent cooperation.

According to the method, a target task to be solved is determined according to natural language input of a user, the target task is decomposed into a plurality of sub-target tasks at least comprising interactive question-answer tasks, the sub-target tasks are correspondingly issued to the sub-target agents at least comprising the interactive question-answer agents, different sub-agents are used for processing different types of sub-tasks, the interactive question-answer agents are controlled to present an interactable graphical user interface corresponding to the interactive question-answer tasks and present answer results corresponding to received feedback answers to the user, the interactable graphical user interface comprises a question and an interaction answer area, the associated sub-target agents in the sub-target agents are controlled to generate corresponding associated processing results based on the answer results output by the interactive question-answer agents, the associated sub-target agents are sub-target agents which depend on the answer results as part of task input information, and the target processing results corresponding to the target task are determined based on the associated processing results.

According to the second aspect, the embodiment of the disclosure provides an interactive question-and-answer task processing device based on multi-agent collaboration, which comprises a target task determining unit, a task decomposing and corresponding issuing unit and an associated processing result control generating unit, wherein the target task determining unit is configured to determine a target task to be solved according to natural language input of a user, the task decomposing and corresponding issuing unit is configured to decompose the target task into a plurality of sub-target tasks at least comprising interactive question-and-answer tasks and correspondingly issue each sub-target task to each sub-target agent at least comprising the interactive question-and-answer agents, different sub-agents are used for processing different types of sub-tasks, the interactive interface control presenting unit is configured to control the interactive question-and-answer agents to present an interactive graphical user interface corresponding to the interactive question-and-answer tasks to the user and present answer results corresponding to received feedback answers, the associated graphical user interface comprises a question and an interactive answer area, the associated processing result control generating unit is configured to control the associated sub-target agents in the plurality of sub-target agents to generate corresponding associated processing results based on answer results output by the interactive question-and answer agents, the associated sub-target agents are configured to be used as target processing result-dependent target processing result determining units, and the associated target processing results are configured to be used as target processing result-dependent and target processing results.

In a third aspect, an embodiment of the disclosure provides an electronic device, including at least one processor, and a memory communicatively connected to the at least one processor, where the memory stores instructions executable by the at least one processor to enable the at least one processor to implement the multi-agent collaboration-based interactive question-answering task processing method as described in the first aspect.

In a fourth aspect, embodiments of the present disclosure provide a non-transitory computer-readable storage medium storing computer instructions for enabling a computer to implement the multi-agent collaboration-based interactive question-answering task processing method as described in the first aspect, when executed.

In a fifth aspect, embodiments of the present disclosure provide a computer program product comprising a computer program which, when executed by a processor, is capable of implementing the steps of the multi-agent collaboration-based interactive question-answering task processing method as described in the first aspect.

According to the interactive question-answer task processing scheme based on multi-agent cooperation, a main agent determines a target task to be solved according to natural language input of a user, the target task is decomposed into a plurality of sub-target tasks at least comprising interactive question-answer tasks, each sub-target task is correspondingly issued to each sub-target agent at least comprising interactive question-answer agents, the interactive question-answer agents present interactive graphical user interfaces corresponding to the interactive question-answer tasks under the control of the main agent, answer results corresponding to received feedback answers are presented, associated sub-target agents depending on the answer results also generate corresponding associated processing results under the control of the main agent based on the answer results, and finally the main agent obtains target processing results corresponding to the target tasks according to the associated processing results.

The method and the system adopt an agent cluster formed by a pre-built main agent and a plurality of sub agents to process task demands proposed by users, wherein the main agent is responsible for understanding the task demands of the users and decomposing the whole task demands into a plurality of sub-target tasks which can be respectively executed by different sub agents, and each sub agent processes the sub-tasks matched with the main agent according to the mobilization of the main agent. The main agent and the sub agents are cooperated to process the task of different types, compared with the scheme that the single and full-functional agent is used for processing the task of different types, the scheme has better processing effect on the single type of task, and the sub agents with relatively smaller scale are convenient for flexibly increasing and changing corresponding functions, and can bring better comprehensive task processing effect under the condition of lower comprehensive cost.

Especially, aiming at the interactive question-answering task which needs the user to make subjective answers to the proposed questions and verify the effect according to the answer result, the interactive question-answering agent can enable the task requirement of the user to be met more quickly through the interactive action answer mode by providing the interactive graphical interface which comprises the questions and the interactive answer area for the user, so that the task solving effect is improved.

It should be understood that the description in this section is not intended to identify key or critical features of the embodiments of the disclosure, nor is it intended to be used to limit the scope of the disclosure. Other features of the present disclosure will become apparent from the following specification.

Drawings

Other features, objects and advantages of the present disclosure will become more apparent upon reading of the detailed description of non-limiting embodiments, made with reference to the following drawings:

FIG. 1 is an exemplary system architecture in which the present disclosure may be applied;

fig. 2 is a flowchart of an interactive question-answering task processing method based on multi-agent collaboration according to an embodiment of the present disclosure;

FIG. 3 is a flow chart of a method for determining a target task from natural language input provided by an embodiment of the present disclosure;

FIG. 4 is a flow chart of a method for decomposing a target task into a plurality of sub-target tasks including an instant messaging query class task according to a task element provided by an embodiment of the present disclosure;

FIG. 5 is a flow chart of a method of creating a sub-agent provided by an embodiment of the present disclosure;

FIG. 6 is a flow chart of a method for controlling an interactive question-answer agent to output answer results provided by embodiments of the present disclosure;

FIG. 7 is a flowchart of a method for controlling an interactive question-answer agent and an associated sub-target agent to generate an associated processing result and obtain a target processing result according to an embodiment of the present disclosure;

fig. 8 is a two-branch schematic diagram of controlling a sub-target agent to output a corresponding sub-processing result according to an embodiment of the present disclosure;

FIGS. 9-1 to 9-14 are exemplary diagrams of a multi-agent collaboration-based interactive question-answering task processing method under an application scenario provided by embodiments of the present disclosure;

Fig. 10 is a block diagram of an interactive question-answering task processing device based on multi-agent collaboration according to an embodiment of the present disclosure;

fig. 11 is a schematic structural diagram of an electronic device adapted to execute an interactive question-answering task processing method based on multi-agent collaboration according to an embodiment of the present disclosure.

Detailed Description

Exemplary embodiments of the present disclosure are described below in conjunction with the accompanying drawings, which include various details of the embodiments of the present disclosure to facilitate understanding, and should be considered as merely exemplary. Accordingly, one of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present disclosure. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness. It should be noted that, without conflict, the embodiments of the present disclosure and features of the embodiments may be combined with each other.

In the technical scheme of the disclosure, the related processes of collecting, storing, using, processing, transmitting, providing, disclosing and the like of the personal information of the user accord with the regulations of related laws and regulations, and the public order colloquial is not violated.

FIG. 1 illustrates an exemplary system architecture 100 to which embodiments of multi-agent collaboration-based interactive question-answering task processing methods, apparatus, electronic devices, and computer-readable storage media of the present disclosure may be applied.

As shown in fig. 1, a system architecture 100 may include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 is used as a medium to provide communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, among others.

The user may interact with the server 105 via the network 104 using the terminal devices 101, 102, 103 to receive or send messages or the like. Various applications for implementing information communication between the terminal devices 101, 102, 103 and the server 105, such as a complex task processing application, a browser application, an instant messaging application, and the like, may be installed on the terminal devices.

The terminal devices 101, 102, 103 and the server 105 may be hardware or software. When the terminal devices 101, 102, 103 are hardware, they may be various electronic devices with display screens, including but not limited to smartphones, tablet computers, laptop computers, desktop computers, etc., and when the terminal devices 101, 102, 103 are software, they may be installed in the above-listed electronic devices, which may be implemented as a plurality of software or software modules, or as a single software or software module, which is not particularly limited herein. When the server 105 is hardware, it may be implemented as a distributed server cluster formed by a plurality of servers, or may be implemented as a single server, and when the server is software, it may be implemented as a plurality of software or software modules, or may be implemented as a single software or software module, which is not limited herein specifically.

The server 105 can provide various services through various built-in applications, for example, a complex task processing application capable of providing one-key processing services for complex tasks, and the server 105 can achieve the following effects when running the complex task processing application, namely, firstly, receiving natural language input which is transmitted by a user through terminal equipment 101, 102 and 103 through a network 104, then determining a target task to be solved according to the natural language input, then decomposing the target task into a plurality of sub-target tasks at least comprising an interactive question-answer type task, issuing each sub-target task to each sub-target agent at least comprising an interactive question-answer type task, using different sub-agents for processing different types of sub-tasks, next, controlling the interactive question-answer agent to present an interactable graphical user interface corresponding to the interactive question-answer type task to the user, presenting an answer result corresponding to the received feedback answer, wherein the interactable graphical user interface comprises a question and an interactive action answer area, then, controlling associated sub-target agents in the plurality of sub-target agents to generate an associated sub-target task based on the output of the associated sub-target agents as the answer result corresponding to the associated target-answer result, and finally, controlling the interactive question-answer agent to present an interactive question-answer agent to be used as the associated target answer result.

Further, the server 105 may also transmit the target processing result back to the terminal devices 101, 102, 103 through the network 104, so that the terminal devices 101, 102, 103 display the received target processing result to the user.

It should be noted that the natural language input may be stored in advance in the server 105 in various ways, in addition to being acquired from the terminal apparatuses 101, 102, 103 through the network 104. Thus, when the server 105 detects that such data has been stored locally (e.g., pending tasks left until processing is initiated), it may choose to retrieve the data directly from the local, in which case the exemplary system architecture 100 may not include the terminal devices 101, 102, 103 and network 104.

Because the complex task needs to occupy more operation resources and stronger operation capability, the interactive question-answering task processing method based on multi-agent cooperation provided in the subsequent embodiments of the present disclosure is generally executed by the server 105 having stronger operation capability and more operation resources, and accordingly, the interactive question-answering task processing device based on multi-agent cooperation is also generally disposed in the server 105. However, it should be noted that, when the terminal devices 101, 102, 103 also have the required computing capability and computing resources, the terminal devices 101, 102, 103 may also complete each operation performed by the server 105 through the complex task processing application installed thereon, and further output the same result as the server 105. Particularly, in the case that a plurality of terminal devices with different computing capabilities exist at the same time, when the complex task processing application determines that the terminal device has a relatively strong computing capability and relatively more computing resources remain, the terminal device can execute the above computation, so that the computing pressure of the server 105 is properly reduced, and accordingly, the interactive question-answering task processing device based on multi-agent cooperation can also be disposed in the terminal devices 101, 102 and 103. In this case, the exemplary system architecture 100 may also not include the server 105 and the network 104.

It should be noted that the main agent and each sub-target agent may be installed on the server 105 at the same time, and each sub-target agent that is invoked and controlled by the main agent may be installed on another server or terminal device different from the server 105, which is not particularly limited herein.

It should be understood that the number of terminal devices, networks and servers in fig. 1 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for implementation.

Referring to fig. 2, fig. 2 is a flowchart of an interactive question-answering task processing method based on multi-agent collaboration according to an embodiment of the disclosure, wherein the flowchart 200 includes the following steps:

Step 201, determining a target task to be solved according to natural language input of a user;

This step aims at determining a target task that the user wants to solve, i.e. the target task corresponds to a task requirement, by fully understanding and analyzing the received natural language input of the user (e.g. the natural language input by the user through the used terminal devices 101, 102, 103 is received through the network 104 shown in fig. 1) by the execution subject of the interactive question-answer task processing method based on multi-agent collaboration (e.g. the server 105 carrying the main agent shown in fig. 1).

Specifically, the step is usually a comprehensive processing process comprising multiple steps, and involves complicated intelligent reasoning links such as language understanding, context analysis, task modeling and the like. Such as receiving input, preprocessing and normalization, semantic understanding and intent recognition, task modeling and reasoning links.

The main intelligent agent serving as an execution main body firstly receives natural language input of a user, then carries out preprocessing and normalization on the received natural language input to ensure that the natural language input can be effectively understood and processed, and then determines a target task according to user intention reflected by input information obtained after the preprocessing and normalization of contextual understanding.

The method comprises the steps of receiving a natural language input, wherein the natural language input is in a form of voice, text, image, video or other forms of language data, converting the natural language input in a non-text form into a natural language text which is convenient to identify and process by adopting a conversion technology for convenience in processing, taking various expression forms (such as direct request, indirect request and multi-step request) of the natural language input on request information into consideration, preprocessing and standardization processing comprising word segmentation and marking, noise removal, spelling correction and synonym processing into consideration for effectively understanding the actual task requirement of a user, and accurately identifying the intention of the user which is expected to be expressed by the natural language input by means of keyword extraction, entity identification and combination of context information in an intention identification stage, and finally obtaining a target task corresponding to the natural language input by adopting task modeling operation comprising task abstraction (mapping the requirement of the user to a task framework which can be identified by a system) and target task definition.

Step 202, decomposing a target task into a plurality of sub-target tasks at least comprising interactive question-answer tasks, and correspondingly issuing each sub-target task to each sub-target agent at least comprising interactive question-answer agents;

Based on step 201, this step aims at decomposing the target task into a plurality of sub-target tasks by the execution subject, and issuing each sub-target task to each sub-target agent correspondingly. The sub-target tasks decomposed in the step at least comprise interactive question-answer tasks, and the sub-target tasks correspond to the interactive question-answer tasks, and each sub-target agent at least comprises an interactive question-answer agent special for processing the interactive question-answer tasks.

The different sub-agents are used for processing different types of sub-tasks, that is, sub-agents dedicated to a single type of task are created in advance according to different task types, and the division modes of the task types are various, for example, according to the processing steps of the task, the processing mode of the task, the complexity of the task, the data types related to the task, and the like, which are not particularly limited herein, that is, the different sub-agents are applied to processing different atomic tasks (that is, the task of the minimum unit, where "minimum" is a relative concept, not an absolute concept, and is the minimum task unit that can be decomposed at present).

The interactive question-answering tasks are interactive tasks that require users to subjectively answer the proposed questions and verify effects according to answer results, such as learning and mastering tasks of knowledge points (especially digital knowledge points, such as chicken and rabbit same-cage questions), game tasks of asking for answers, and the like. Taking a task of learning a chicken and rabbit same-cage problem as an example, it is known that the task not only needs to learn the definition of the problem, but also needs to learn and master the knowledge point as much as possible by answering practice problems and test problems after learning the definition, so that the execution focus of the task is to provide an interactive problem solution in the form of an interactive problem, so that a user can solve the task requirement in the interaction process.

In this embodiment, the plurality of sub-target tasks required to be decomposed by the target task in this step should at least include an interactive question-and-answer task, and in addition, the task types of other decomposed sub-target tasks are not limited, and may be any type of task, and a sub-target agent matched with the task type and having a corresponding type of task processing capability is also required to be used.

Step 203, controlling the interactive question and answer to present an interactive graphical user interface corresponding to the interactive question and answer type task to the user, and presenting an answer result corresponding to the received feedback answer;

Based on step 202, this step aims to present the interactive graphical user interface corresponding to the interactive question-answer type task to the user by the execution subject interactive question-answer, and present the answer result corresponding to the received feedback answer. The interactive graphic user interface comprises a question and an interactive answer area, the feedback answer of the user is input in the interactive answer area, the input modes include various modes, such as clicking alternative options, inputting the answer in a blank filling area or a specific graph, and the like, and the answer result corresponding to the feedback answer is a judging conclusion of whether the feedback answer is correct or not, for example, the judging conclusion can comprise an incorrect answer and a correct answer.

204, Controlling associated sub-target intelligent agents in the plurality of sub-target intelligent agents to generate corresponding associated processing results based on the answer results output by the interactive question-answer intelligent agents;

On the basis of step 203, this step aims to control, by the execution subject, an associated sub-target agent among the plurality of sub-target agents to generate a corresponding associated processing result based on the answer result output by the interactive question-answer agent, the associated sub-target agent being a sub-target agent that depends on the answer result as part of the task input information, so that execution of the corresponding sub-target task by the associated sub-target agent is actually arranged after the interactive question-answer agent outputs the answer result.

It should be appreciated that each sub-target agent, including the interactive question-answer agent and the associated sub-target agent, should actually be a call or arrangement by the main agent that outputs sub-processing results corresponding to the issued sub-target task. Specifically, the main agent can synchronously inform the execution sequence or the execution trigger signal when correspondingly issuing each sub-target task to each sub-target agent, so that each sub-target agent executes the corresponding sub-target task according to the informed execution sequence or executes the corresponding sub-target task when judging that the current state meets the requirement of the execution trigger signal, namely each sub-target agent actively executes the corresponding sub-target task at a proper time, and the main agent can also actively invoke each sub-target agent to execute the corresponding sub-target task according to the requirement in a mode of actively invoking when determining that different sub-target agents meet the execution time, namely each sub-target agent only passively executes the corresponding sub-target task according to the invoking instruction.

Whether the above-mentioned active execution mechanism or passive execution mechanism belongs to different cooperative forms of cooperative processing tasks of the main agent and the plurality of sub-agents, which is specifically selected according to actual requirements of actual application scenarios, and is not specifically limited herein.

In addition, in the process that each sub-target agent executes the corresponding sub-target task to obtain the corresponding sub-processing result, the personalized preference of the user can be combined. The multiple sub-processing results at least comprise answer results output by the interactive question-answer agent and corresponding to the affiliated interactive question-answer tasks and associated processing results output by the associated sub-target agent. The rest sub-processing results can also be the results obtained by other agents when processing the tasks of the corresponding types, the sub-processing results can be obtained only by depending on the other sub-processing results, and the sub-processing results can also be obtained by self determination according to the task description information of the sub-target tasks without depending on the other sub-processing results.

And 205, determining a target processing result corresponding to the target task based on the associated processing result.

On the basis of step 203, this step is intended to be performed by determining a target processing result corresponding to the target task based on the associated processing result.

Specifically, in the process of processing the associated processing result to obtain the target processing result, the decomposing mode of decomposing a plurality of sub-target tasks from the target task and the multiple processing modes including de-duplication, adjustment expression, sentence adjustment and rich expression are fully referred to, so that the finally obtained target processing result not only can meet the task requirement of a user, but also can display critical requirement result information and improve information identification as much as possible.

The interactive question-answering task processing method based on multi-agent cooperation provided by the embodiment of the disclosure adopts an agent cluster formed by a pre-built main agent and a plurality of sub-agents to process task demands proposed by users, wherein the main agent is responsible for understanding task demands of the users and decomposing the overall task demands into a plurality of sub-target tasks which can be respectively executed by different sub-agents, and each sub-agent processes the sub-tasks matched with the main agent according to the mobilization of the main agent. The main agent and the sub agents are cooperated to process the task of different types, compared with the scheme that the single and full-functional agent is used for processing the task of different types, the scheme has better processing effect on the single type of task, and the sub agents with relatively smaller scale are convenient for flexibly increasing and changing corresponding functions, and can bring better comprehensive task processing effect under the condition of lower comprehensive cost.

To enhance understanding of how to determine the target task, referring also to fig. 3, fig. 3 is a flowchart of a method for determining the target task according to a natural language input according to an embodiment of the disclosure, where the flowchart 300 includes the following steps:

Step 301, converting natural language input of a user into natural language text;

The step aims at converting the natural language input of various expression forms into natural language texts which are convenient to recognize and come out by the execution main body, namely uniformly converting the various expression forms into text forms, for example, a voice-to-text technology can be adopted to convert natural voice signals of voice forms into natural language texts.

Step 302, carrying out intention recognition on the natural language text to obtain an intention recognition result comprising knowledge point learning and grasping intention;

On the basis of step 301, this step aims at performing intention recognition on the natural language text by the execution subject, and obtaining an intention recognition result including knowledge point learning intention. That is, the intention recognition result includes at least the knowledge point learning grasping intention, and may include other intentions, and is not particularly limited herein.

An implementation, including and not limited to, may include the following specific steps:

firstly, the executing body carries out semantic understanding on a natural language text to obtain a semantic understanding result, then, the executing body determines a preliminary intention (or called suspected intention) according to the semantic understanding result, and in order to improve the accuracy of intention determination, the executing body can also carry out personalized correction on the preliminary intention by utilizing the personalized preference of the user to obtain an intention recognition result comprising the requirement verification intention combined with instant information.

The implementation method comprises the steps of firstly determining the preliminary intention directly corresponding to the natural language text by using semantic understanding, and then carrying out personalized correction on the preliminary intention by using personalized preference of the user, such as screening, removing, supplementing implicit or missing information and the like, so that the finally determined intention recognition result is more in line with the actual expectation of the user.

In addition to this implementation, various technologies involved in the multiple processing links mentioned in step 201 may be adopted, and other implementations for determining the intent recognition result are built by combining with the actual application scenario, which are not listed here.

And step 303, determining a target task to be solved according to the intention recognition result.

On the basis of step 302, this step aims at determining the target task to be solved by the execution subject according to the intention recognition result. Wherein the target tasks include tasks for learning knowledge points.

Specifically, to convert the identified intent into a specific task, a task modeling operation including task abstraction (mapping the user's needs to a task framework recognizable by the system), and a target task definition may also be employed to finally obtain a target task corresponding to the natural language input.

The embodiment provides a more specific implementation manner of determining the target task according to the natural language input through steps 301-303, which includes the steps of conversion of the expression form, intention recognition and key processing for determining the target task based on the result of the intention recognition, so as to provide a more accurate implementation scheme of the determined target task with high feasibility as much as possible.

To enhance understanding of how to decompose a target task into a plurality of sub-target tasks, referring also to fig. 4, fig. 4 is a flowchart of a method for decomposing a target task into a plurality of sub-target tasks including an explicit result acquisition class task according to a task element provided in an embodiment of the disclosure, where the flowchart 400 includes the following steps:

step 401, determining a plurality of task elements constituting a target task;

the present step aims at determining a plurality of task elements constituting a target task from the execution subject, and different task elements correspond to different types of tasks.

The task elements refer to basic units or constituent parts forming target tasks, and each target task may be composed of a plurality of independent elements, and the elements are key steps, sub-tasks or resources necessary for realizing the target task. Each task element may correspond to an independent task that work cooperatively to facilitate completion of the target task.

Assuming that the target task is "develop a new software product," the target task can be broken down into multiple task elements, each task element corresponding to a different type of task:

1) Demand analysis (administrative tasks) determining user demand, market demand;

2) System design (technical task) architecture design, database design, interface design;

3) Code development (technical task) that a programmer writes codes and realizes functions;

4) Testing and quality control (technical and administrative tasks) unit testing, integration testing, quality assurance;

5) Marketing promotion (manageability and communication task) by making a marketing plan and promoting products.

Step 402, decomposing an interactive question-answering task for learning and grasping knowledge points from a target task according to a single task element which specifically requires a user to perform subjective answer on a proposed problem and checks whether the knowledge points to which the problem belongs are learned and grasped according to an answer result;

The step aims at decomposing the interactive question-answering task from the target task by the execution main body, and particularly depends on a single task element which specifically needs a user to subjectively answer the proposed problem and checks whether knowledge points to which the problem belongs are learned or not according to the answer result.

Specifically, in order to fully grasp a certain knowledge point, two interactive question-answering agents may be designed, one is a knowledge point exercise answer agent corresponding to a knowledge point exercise stage, and the other is a knowledge point grasp test agent corresponding to an exercise result inspection stage. The difference is that one provides a plurality of exercises and the other provides only a small number of test questions.

Step 403, decomposing the target task into a plurality of sub-target tasks comprising only a single task element.

The step aims at decomposing the target task into a plurality of sub-target tasks only comprising single task elements in the mode of decomposing and obtaining the interactive question-and-answer type task according to the task elements in the step 402 by the execution subject, wherein the sub-target tasks at least comprise the interactive question-and-answer type task.

In this embodiment, a task decomposition manner is provided by steps 401-403 according to task elements forming a target task and decomposing each task element into sub-target tasks, and the task decomposition can be flexibly completed according to actual situations by combining the predefining of the task elements (for example, according to a manner of constructing different sub-agents).

Further, considering that even the same task element has additional conditions, when the task is decomposed, the task needs to be decomposed into sub-target tasks carrying the additional conditions, so that when the sub-target tasks are issued to the sub-target agents, the sub-target agents are selected as the sub-target agents by combining the additional conditions.

An implementation, including but not limited to, may be to decompose a target task into a plurality of sub-target tasks that include only a single task element and corresponding additional conditions, which may include a specified task processing style and/or a specified result presentation style, in response to the presence of additional conditions corresponding to the task elements. The different task processing modes abstract different processing logics of the same type of task (for example, the same type of task can have a plurality of different processing logics and can finally obtain correct answers), and the result display form can comprise at least one of a text class, a form class, an image class, a video class and an interactable card class.

To enhance understanding of how different sub-agents are created in advance, referring also to fig. 5, fig. 5 is a flowchart of a method for creating sub-agents according to an embodiment of the present disclosure, where the flowchart 500 includes the following steps:

step 501, determining a task type set for creating sub-agents;

step 502, constructing sub-agents with task processing capacity of the corresponding type of tasks for each type of tasks in the task type set.

The present embodiment provides a scheme for constructing sub-agents with task processing capabilities of corresponding types of tasks for the determined multiple types of tasks through steps 501-502. Further, the number of sub-agents corresponding to each type of task may be only 1, or may be plural, and there may be a distinction between plural ones, or may be the same, or whether the distinction is required may depend on whether the type of task generally includes a distinction in further additional conditions.

Specifically, for a first target type task with at least two task processing modes, a sub-agent corresponding to each task processing mode can be constructed for the first target type task, and for a second target type task with at least two result display modes, a sub-agent corresponding to each result display mode can be constructed for the second target type task.

To enhance understanding of how the interactive question-answering agent outputs the answer result, referring also to fig. 6, fig. 6 is a flowchart of a method for controlling the interactive question-answering agent to output the answer result according to an embodiment of the present disclosure, where the flowchart 600 includes the following steps:

Step 601, controlling an interactive question-answering agent to determine a question and an interactive answer area based on received task input information;

And the execution body controls the interactive question-answering agent to determine a question and an interactive question-answering area based on the received task input information, wherein the received task input information is determined based on task description information which is decomposed from the target task and is related to the interactive question-answering type task. Furthermore, the received task input information can be determined together according to task description information related to the interactive question-answer task and personalized preferences of the user, which are decomposed from the target task, namely, the personalized preferences of the user are additionally introduced, so that the matching degree of the determined task input information and the actual task demands of the user is improved as much as possible.

In particular, the personalized preferences may include at least one of:

Age, educational level, area of interest, historical consumption preferences.

Taking the age in the personalized preference as an example, when the execution body controls the interactive question-answering agent to determine the problem based on the received task input information, the method is specifically implemented as the execution body controls the interactive question-answering agent to determine the problem matched with the cognition degree corresponding to the age based on the received task input information, so that the generated problem is matched with the cognition degree of the current age of the user.

Step 602, controlling the interactive question-answering agent to present an interactive graphical user interface containing questions and interactive response areas to a user;

Based on step 601, this step aims to control the interactive questioning and answering agent to present an interactable graphical user interface comprising a question and an interactive answer area to the user by the execution subject, so that the user can answer in the interactive answer area by presenting the interactable graphical user interface comprising the question and the interactive answer area, thereby facilitating the receipt of a feedback answer.

Step 603, controlling the interactive question-answering agent to present an answer result determined according to the difference degree between the feedback answer received from the interactive question-answering area and the standard answer of the question;

The step aims at controlling the interactive question-answering agent to present an answer result determined according to the degree of difference between the feedback answer received from the interactive question-answering area and the standard answer of the question by the execution subject. That is, the larger the difference between the feedback answer and the standard answer, the more the answer result is obtained as a wrong answer, and the more the answer result is obtained as a nearly correct answer.

The present step provides a specific implementation manner for controlling the interactive question-answer agent to present the interactive graphical user interface, receiving the feedback answer, and analyzing the feedback answer to obtain the answer result through steps 601-603.

To further enhance understanding of how the associated sub-target agent generates the associated processing result based on the answer result and how the target processing result is obtained based on the associated processing result, referring also to fig. 7, fig. 7 is a flowchart of a method for controlling the interactive question-answer agent, the associated sub-target agent to generate the associated processing result and obtain the target processing result according to an embodiment of the disclosure, where the flowchart 700 includes the following steps:

Step 701, responding to the feedback answer being an error answer different from the standard answer, controlling the interactive question-answering agent to regenerate an interactive graphical user interface containing a new question different from the question corresponding to the last time of answering the error and an interactive answer area matched with the new question;

The present step aims at the situation that the interactive question-answering agent determines that the current feedback answer is an incorrect answer, and on the basis, the execution main body further controls the interactive question-answering agent to regenerate an interactable graphical user interface comprising a new question different from the question corresponding to the previous answer error and an interactive answer area matched with the new question, and in order to increase the answer accuracy of the new question, the new question is further controlled to be generated based on the question corresponding to the previous user answer accuracy, namely, the new question is made to be as close to the question corresponding to the previous answer accuracy as possible, and a certain difference exists, so that the learning and the mastering of knowledge points to which the question belongs are further consolidated.

Step 702, controlling the interactive question-answering agent to output prompt information of completion of the interactive question-answering in response to obtaining a preset number of answer results;

Under the condition that the interactive question-answering agent is required to continuously initiate the preset number of questions to ask users to answer and further obtain the preset number of answer results, the execution main body controls the interactive question-answering agent to output prompt information of completion of the interactive question-answering, namely the prompt information represents that the answer of the preset number of questions is completed.

Step 703, in response to receiving the prompt message, controlling the accuracy rate calculation agent to determine the response accuracy rate according to the preset number of response results;

Based on step 702, this step aims at controlling the accuracy rate calculating agent to determine the response accuracy rate according to the preset number of response results when the executing agent receives the prompt message, that is, the accuracy rate calculating agent of this step is a specific associated sub-target agent.

Step 704, controlling learning advice agents in the plurality of sub-target agents to output matched learning advice according to the response accuracy;

based on step 703, this step aims at outputting a matched learning advice by the above-described execution subject control learning advice agent according to the answer accuracy rate.

Step 705, summarizing the number of the answer results, the answer accuracy and the learning advice to obtain a target processing result corresponding to the target task.

Based on step 704, this step aims to collect, by the execution subject, the number of answer results, the answer accuracy and the learning advice, and obtain a target processing result corresponding to the target task, which is used to comprehensively describe the situation of the current knowledge point grasp.

Based on any of the above embodiments, considering that the user may correct the content of the outputted partial sub-processing results or send new constraint information at any time in the whole process of respectively executing the corresponding sub-target tasks by the sub-target agents to output the sub-processing results, referring also to fig. 8, fig. 8 is a two-branch schematic diagram for controlling the sub-target agents to output the corresponding sub-processing results according to the embodiment of the present disclosure, the process 800 includes the following steps:

Step 801, responding to the selected state of the instruction input box, and controlling the currently executed sub-target agent to suspend outputting the corresponding sub-processing result;

The instruction input box is used for inputting instructions by a user, and is in an unselected state in the process of outputting corresponding sub-processing results by the sub-target agent. That is, under the scheme of the embodiments provided by the disclosure, once the instruction input box is in the selected state, the user may need to input some new instructions, and the original sub-processing result output process needs to be interrupted.

Step 802, in response to no new instruction generated in the process of recovering the instruction input box from the selected state to the unselected state, controlling the currently executed sub-target agent to continuously output a corresponding sub-processing result;

this step corresponds to a branching situation where no new instruction is generated in the process of restoring the instruction input box from the selected state to the unselected state, i.e. no new instruction is input to the user, and then the sub-processing result output result originally interrupted by the selected instruction input box will continue.

Step 803, in response to the new instruction being generated in the process of recovering the instruction input box from the selected state to the unselected state, extracting correction information from the new instruction;

This step corresponds to another branch situation in which a new instruction is generated during the return of the instruction input box from the selected state to the unselected state, and the execution body is required to extract correction information from the new instruction, where the correction information includes a correction to the original information and an addition to the new information.

Step 804, determining sub-target agents affected by the correction information;

and 805, controlling the affected sub-target agent to output the corresponding sub-processing result again in combination with the correction information.

Step 804 and step 805 are performed by the execution body determining the sub-target agent affected by the correction information, and then the execution body controlling the affected sub-target agent to output the corresponding sub-processing result again in combination with the correction information. Specifically, if all the sub-target agents are affected, the sub-process result may be output from the sub-target agent that is executed first based on the correction information.

To deepen understanding, the present disclosure further provides a specific implementation scheme for attempting to eliminate the prior art defect and overcome the prior art problem in combination with the prior art defect actually existing in the specific application scenario:

The existing demand meeting mode mainly adopts a multi-layer system funnel to recall, sort, mix and arrange heterogeneous results and the like of related webpages through a search engine. And each strategy funnel sorts the web pages based on the information of basic relevance, user feedback behavior, authority and the like, and cuts off and outputs the web pages to the next layer. The system has the disadvantages that the relevance can only be matched from the content level through relevance, the relevance cannot be understood from the task level and the problem can be solved for the user, in addition, the user can only use the fixed content to meet the user requirement, and when the user expresses personalized requirements and the user can further meet the requirements through multiple rounds, the high-quality and continuous meeting effect cannot be realized. Again the whole matching process is unexplainable for the user, who can only perform the last information screening.

Related large model-based helper-type products often want to try to solve this problem by a large model-based full-featured helper. However, a single large model is difficult to achieve high-quality satisfaction in different open fields and different tasks, and in professional fields and special scenes, a single assistant is difficult to establish trust degree among users and achieve expected satisfaction effects.

Namely, the related art has the following difficulties in terms of task completion in the open field:

1) How to efficiently understand and disassemble the key steps of the user expression;

2) How to find the best satisfying mode for each key step;

3) How to output the execution process and the integration result user-friendly and completely.

In order to solve the above-mentioned difficulties, the present embodiment provides a solution for completing the end-to-end task of the user demand interaction based on the multi-agent collaboration mode. The scheme is characterized in that:

Before searching, a single point type query (query word, query sentence, search word and search sentence) is used for completing a requirement, each requirement needs to be defined by the user's own thinking, the cost is too high for the user, and for a complex task, the user does not know how to think about multiple queries to search. The paradigm proposed in this embodiment is intended to thoroughly help the user complete the task, and the task completion paradigm can help the user directly complete the problem to be solved or the task to be completed in one step.

The embodiment realizes disassembly and multi-step satisfaction of the user demands based on cooperation among a plurality of agents, and finally, the method is automatically integrated into a complete solution capable of meeting the whole task. The scheme can fulfill the following functions:

1) The method comprises the following steps of understanding the key steps of demand, constructing a main intelligent agent, understanding intention of the demand expressed by a user and disassembling the main intelligent agent into key steps which can be completed by a plurality of subtask intelligent agents;

2) Subtask agent generation, wherein the main agent is responsible for generating a plurality of subtask agent candidates required for completing a current task based on understanding of key steps and agent capabilities;

3) The method comprises the steps that an interactive task is completed, a plurality of sub-task agents input according to the task, key steps and user personalized preference information, and personalized interaction steps and user interaction interfaces meeting the task are generated, so that a user can complete the whole task through interaction with the multi-agents;

4) And integrating and outputting the complete scheme, interactively generating results by the main agent integrating task disassembling process and a plurality of agents, and completing the interactive presentation of the whole scheme through dialogue arrangement.

In other words, in the embodiment, the main agent is constructed to schedule the plurality of sub-task agents to cooperate, and the main agent cooperates with the plurality of sub-task agents and cooperates with the plurality of sub-task agents to complete the whole task;

And by providing a scheduling and distributing mechanism based on an end-to-end generation type large model, the large model has stronger expression capability, can overcome the problems of insufficient effect consistency and global optimality of the traditional multi-layer ordering mechanism, and can realize the global optimal combined scheduling optimization for tasks. From the viewpoint of model capability, from the perspective of intelligent body understanding and task understanding, the end-to-end alignment training of the model is realized.

Simultaneously, a conversational arrangement task is provided, the complex task is presented through the conversational arrangement by sub-tasks of a plurality of steps according to the logic sequence under the application scene, the user experiences very smoothly in the process, and the contents are not given out in any combination or at one time, so that the information is numerous and complicated, and the user cannot digest; and because the generated completion of the whole task is personalized, the completion from task disassembly to subtask is generated according to user input and user preference, different scheme disassembly, subtask agent completion scheme and presentation effect are generated for different users, and the user satisfaction effect is improved.

The implementation block diagram of the scheme is shown in the following figure 9-1, and inputs of the system are the demands input by a user and distributable agent sets (including basic setting of the agents, mounted plugins, workflows and the like) and outputs multi-level generated contents for completing tasks. The whole system is constructed based on a multi-agent cooperative mode and mainly comprises two types of agents, namely 1) a main agent (MAIN AGENT), wherein the whole understanding, decomposition and series connection of tasks are performed, and the final result is given. 2) The subtask agent (TASK AGENT) is that the main agent calls the calling agent to cooperatively complete the task issued by the main agent.

The scheme provided by the embodiment can be widely applied to various information meeting application programs or independent products, and the specific application forms of the embodiment are shown below by taking a scene where a hundred X application program is located as an example:

The user can switch to the function use mode provided in the present embodiment by one key (the "AI button" of the bottom Bar in the following figures) while searching conventionally. The student in the primary school stage inputs the chicken and rabbit co-cage problem, the results of web search and intelligent answer are displayed (corresponding to fig. 9-2), and the interactive multi-agent meeting page of the chicken and rabbit co-cage problem is switched to through an AI button, and a graphical teaching mode conforming to the student acceptance mode in the primary school stage is generated as shown in the following fig. 9-3.

In the task completion page, an interactive interface generated by a plurality of task agents through automation is displayed, and a user completes the complete process from teaching, practice and test of the chicken and rabbit same cage problem through interaction. The intelligent body 'primary school mathematics teacher' is responsible for decomposing the whole problem and disassembling the problem to each subtask, including calling the 'problem solving method jockey' intelligent body to complete the teaching task, the 'scotch bottom teaching aid' intelligent body to complete the training task, and the 'chicken and rabbit same cage test' intelligent body to complete the testing task. The result of each task completion agent completes the subtasks through the user-friendly generated interactive UI, as in fig. 9-4 through 9-7, which in turn presents the process of "solution method jockey" agent interactive teaching task completion.

When the user completes the interactive task of an agent, the host agent's mathematics and the teacher' guides to conduct the next exercise link, and the same way, the interactive is conducted through the automatic generation UI to complete the exercise of the same cage problem of the chicken and the rabbit, as shown in figures 9-8 to 9-9, and then the host agent scheduling subtask agent's chicken and rabbit same cage test' completes the test of the knowledge point, as shown in figures 9-10 to 9-12.

The personalized effect is generated for different users, for example, the users are students in 12-14 the beginning of the year, the same generation result of the 'chicken rabbit with cage' problem is shown in comparison presented in fig. 9-13 to fig. 9-14, and it can be seen that fig. 9-14 provides a matching scheme solved by a junior knowledge point column equation.

As can be seen more specifically from the above examples, this embodiment has the following improvement points and technical effects over the prior art:

1) Task satisfaction paradigm, which is to upgrade a user's request into a complete multi-agent task completion satisfaction paradigm instead of once more information retrieval

The current search engine can search information for carrying out relevance matching on requests, and the intelligent assistant can carry out multi-step disassembly and step-by-step search for one-time tasks to meet, but the disassembly of task meeting and the complete requirement meeting of users from multi-agent cooperation are a brand new user requirement meeting mode. Besides, the disassembly and multi-agent capacity alignment training is provided, so that the task disassembly and task completion are highly consistent, and the task completion effect is greatly improved.

2) Multi-agent candidate generation based on multi-agent collaboration

According to the task disassembly result, a target cooperative agent set is generated end to end through a large model, which is a brand-new method, is different from the traditional method of sorting agents through a retrieval recall sorting system, the multi-agent cooperative task is satisfied, the combination of the multi-agent on the requirement that the task is not completed is met, the large model is required to be capable of deeply understanding the task, the capability of understanding each agent is required to carefully delineate the difficult boundary of the problem, and finally the optimal agent set under the current target is obtained in a combination optimization mode. The difficulties here include:

difficulty 1-deep understanding of the capabilities of an agent, understanding the boundaries of the capabilities of an agent and the task that an agent is adept, finely characterizing subtle differences between capabilities, such as different creation styles of drawing agents, and repairing subtle capability differences in drawings.

And the difficulty 2 is that understanding the matching relationship between the task and the intelligent capability is not semantic similarity matching, but depth matching on the task completion capability, the task type and the task boundary are required to be deeply described and understood, and matching on the intelligent capability and the open set of the task requirement is a great challenge.

And 3, under the condition of giving a multi-agent cooperative task, obtaining the optimal combination is a combination optimization problem, and combination optimization is needed.

3) The task completion process and results are satisfied by the interactions of agent end-to-end generation and automation orchestration. Currently, it is difficult for a search engine to speak the search process to a user, while an intelligent assistant presents the thinking result and the single-step result, and presents a style in which the UI is basically preset. The invention can display the results in an end-to-end generation mode, comprising the steps of displaying the content, the style and the result of each step, thanks to the task disassembly and combination of multi-agent cooperation and the completion mode of multi-subtasks. And the result of each step and the subtask completed by each agent are according to the personalized preference of the user, and the result can be further interacted, so that the end-to-end generated experience is a comprehensive innovation. Specifically, the end-to-end generation experience includes several parts:

① Constructing a main task model from task disassembly to generation of an agent set to agents, and training by a unified model;

② The end-to-end generation of the interaction is that the generation result of the agent is not single-round content, but content and a UI interface which can be interacted in a plurality of rounds are generated. Compared with the traditional method of generating single content, generating a preset graphical interactive interface (GRAPHICAL USER INTERFACE, GUI) and a meeting form of a natural language interactive interface (Language User Interface, LUI), the automatic generation and planning combined interactive interface can greatly improve the ceiling of the single agent meeting capability. In order to achieve this capability, the agent needs to complete the capability of interactive planning (UI planning) in addition to the planning of the underlying logic, and this planning capability is a significant innovation in the agent, so that the agent can plan on the thinking task, and also achieve planning on the external interactive presentation, thereby greatly improving the satisfaction capability and interactive friendliness of the agent.

③ The interactive editing type interaction satisfies that the interactive display is automatically arranged from task disassembly to task satisfaction, so that the task completion effect can be greatly improved. The method is equivalent to the automatic generation of the whole of the planning of task completion, the content of task step completion and the interactive design of step completion, and greatly improves the task scene which can be met by multiple intelligent agents. Particularly for complex tasks, multiple steps are required to accomplish the task through complex interactions. For example, knowledge point learning in teaching materials of middle and primary schools can be mastered by students through complete processes of teaching, practice and testing, the traditional education application can only solidify content, teaching can not be performed by adopting different forms aiming at different knowledge, and layering coordination of learning, practice and testing on specific knowledge points is difficult.

④ The personalized energy is added in the whole end-to-end generation process, so that the satisfaction effect is further greatly improved. Aiming at different users, the same task has great differences from the scheme to the satisfaction mode, the interactive design and the result planning, and the end-to-end generation of the personalized information can completely customize the whole effect for the user, so that the effect ceiling is greatly improved.

With further reference to fig. 10, as an implementation of the method shown in the foregoing fig. s, the present disclosure provides an embodiment of an interactive question-answering task processing device based on multi-agent collaboration, where the embodiment of the device corresponds to the embodiment of the method shown in fig. 2, and the device may be specifically applied to various electronic devices.

As shown in fig. 10, the interactive question-and-answer task processing device 1000 based on multi-agent cooperation of the present embodiment may include a target task determination unit 1001, a task decomposition and correspondence issuing unit 1002, an interactive interface control presenting unit 1003, an association processing result control generating unit 1004, and a target processing result determination unit 1005. The target task determining unit 1001 is configured to determine a target task to be solved according to natural language input of a user, the task decomposing and corresponding issuing unit 1002 is configured to decompose the target task into a plurality of sub-target tasks at least comprising an interactive question-answer type task and correspondingly issue each sub-target task to each sub-target agent at least comprising the interactive question-answer agent, wherein different sub-agents are used for processing different types of sub-tasks, the interactive interface control presenting unit 1003 is configured to control the interactive question-answer agent to present an interactive graphical user interface corresponding to the interactive question-answer type task to the user and present an answer result corresponding to a received feedback answer, wherein the interactive graphical user interface comprises a question and an interactive answer area, the associated processing result control generating unit 1004 is configured to control the associated sub-target agents in the plurality of sub-target agents to generate corresponding associated processing results based on the answer result output by the interactive question-answer agent, wherein the associated sub-target agents are sub-target agents which depend on the answer result as part of task input information, and the target processing unit 1005 is configured to correlate the task processing results corresponding to the received response result.

In the present embodiment, in the interactive question-answer task processing device 1000 based on multi-agent collaboration, specific processes and technical effects of the target task determining unit 1001, the task decomposing and corresponding issuing unit 1002, the interactive interface control presenting unit 1003, the associated processing result control generating unit 1004, and the target processing result determining unit 1005 may refer to the relevant descriptions of steps 201 to 205 in the corresponding embodiment of fig. 2, respectively, and will not be described herein.

In some optional implementations of the present embodiment, the target task determination unit 1001 may include:

a conversion subunit configured to convert the natural language input of the user into natural language text;

an intention recognition subunit configured to perform intention recognition on the natural language text to obtain an intention recognition result including knowledge point learning grasp intention;

and a target task determination subunit configured to determine a target task to be solved according to the intention recognition result, wherein the target task comprises a task for learning to grasp knowledge points.

In some optional implementations of the present embodiment, the intent recognition subunit may be further configured to:

carrying out semantic understanding on the natural language text to obtain a semantic understanding result;

determining a preliminary intention according to the semantic understanding result, and performing personalized correction on the preliminary intention by utilizing personalized preference of the user to obtain an intention recognition result including knowledge point learning grasping intention.

In some optional implementations of this embodiment, the task decomposition and correspondence issuing unit 1002 may include a task decomposition sub-unit configured to decompose the target task into a plurality of sub-target tasks including at least the interactive question-and-answer type task, and the task decomposition sub-unit may include:

A task element determination module configured to determine a plurality of task elements constituting a target task, wherein different task elements correspond to different types of tasks;

the interactive question-answering task decomposition module is configured to decompose an interactive question-answering task for learning and grasping knowledge points from a target task according to a single task element which specifically requires a user to subjectively answer a proposed problem and checks whether the knowledge points to which the problem belongs are learned and grasped according to an answer result;

The sub-target task decomposition module is configured to decompose the target task into a plurality of sub-target tasks which only comprise single task elements, wherein the sub-target tasks at least comprise interactive question-answer tasks.

In some alternative implementations of the present embodiments, the interactive question-answering agent includes a knowledge point exercise answer agent corresponding to a knowledge point exercise phase and a knowledge point mastering test agent corresponding to an exercise outcome verification phase.

In some alternative implementations of the present embodiment, the sub-objective task decomposition module may be further configured to:

And responding to the existence of additional conditions corresponding to the task elements, decomposing the target task into a plurality of sub-target tasks only comprising single task elements and corresponding additional conditions, wherein the additional conditions comprise a designated task processing mode and/or a designated result display mode.

In some optional implementations of this embodiment, the multi-agent collaboration-based interactive question and answer task processing device 1000 may further include a sub-agent constructing unit configured to construct different sub-agents, where the sub-agent constructing unit may include:

a task type set determination subunit configured to determine a task type set for creating a sub-agent;

A creation subunit configured to construct, for each type of task in the set of task types, a sub-agent having task processing capabilities for the corresponding type of task.

In some optional implementations of the present embodiment, the creation subunit may be further configured to:

Responding to the first target type task with at least two task processing modes, and constructing sub-agents corresponding to each task processing mode for the first target type task, wherein different task processing modes abstract from different processing logics of the same type task.

responding to the second target type task to have at least two result display forms, and constructing sub-agents respectively corresponding to each result display form for the second target type task, wherein the result display forms comprise at least one of characters, forms, images, videos and interactive cards.

In some optional implementations of the present embodiment, the interactive interface control presentation unit 1003 may include:

the system comprises a first control subunit, a first control subunit and a second control subunit, wherein the first control subunit is configured to control an interactive question-answer agent to determine a question and an interactive answer area based on received task input information, and the received task input information is determined and obtained based on task description information which is decomposed from a target task and is related to an interactive question-answer type task;

A second control subunit configured to control the interactive question-answering agent to present an interactable graphical user interface to the user comprising a question and an interactive answer region;

And a third control subunit configured to control the interactive question-answering agent to present an answer result determined according to a degree of difference between the feedback answer received from the interactive question-answering area and the standard answer of the question.

In some optional implementations of this embodiment, the interactive question-answering task processing device 1000 based on multi-agent collaboration may further include:

And the regeneration control unit is configured to respond to the feedback answer being an error answer different from the standard answer, and control the interactive question-answering agent to regenerate an interactive graphical user interface containing a new question different from the question corresponding to the last time of answering the error and an interactive answer area matched with the new question, wherein the new question is generated based on the question corresponding to the last time of answering the error.

The prompt information output control unit is configured to control the interactive question-answering agent to output prompt information of completion of the interactive question-answering in response to obtaining a preset number of answer results;

Correspondingly, the association processing result control generation unit 1004 is further configured to:

And in response to receiving the prompt message, controlling the accuracy rate calculation agent to determine the response accuracy rate according to the preset number of response results.

In some optional implementations of the present embodiment, the target processing result determining unit 1005 may be further configured to:

Controlling learning suggestion agents in the plurality of sub-target agents to output matched learning suggestions according to the response accuracy;

and summarizing the number of the answer results, the answer accuracy and the learning advice to obtain a target processing result corresponding to the target task.

In some optional implementations of this embodiment, the received task input information is determined according to task description information related to the interactive question-and-answer task and personalized preferences of the user, which are resolved from the target task.

In some alternative implementations of the present embodiment, the personalized preferences include at least one of:

Age, educational level, area of interest, historical consumption preferences.

In some optional implementations of this embodiment, the first control subunit may include a question determination control module configured to control the interactive question-answering agent to determine a question based on the received task input information, the question determination control module may be further configured to:

in response to the personalized preferences including age, the interactive question-answering agent is controlled to determine a question matching a degree of cognition corresponding to the age based on the received task input information.

A pause output control unit configured to control the currently executed sub-agent to pause outputting the corresponding sub-processing result in response to the instruction input box being in a selected state, wherein the instruction input box is used for inputting an instruction by a user, and the instruction input box is in an unselected state in the process of outputting the corresponding sub-processing result by the sub-agent;

A correction information extraction unit configured to extract correction information from a new instruction in response to the new instruction being generated in a process of the instruction input box being restored from the selected state to the unselected state;

an affected sub-target agent determination unit configured to determine a sub-target agent affected by the correction information;

and the re-output control unit is configured to control the affected sub-target agent to re-output the corresponding sub-processing result in combination with the correction information.

The device embodiment corresponding to the method embodiment is presented, and the interactive question-answering task processing device based on multi-agent cooperation provided by the embodiment adopts an agent cluster formed by a pre-built main agent and a plurality of sub-agents to process task demands proposed by users, wherein the main agent is responsible for understanding task demands of the users and decomposing the whole task demands into a plurality of sub-target tasks which can be respectively executed by different sub-agents, and each sub-agent processes the sub-tasks matched with the sub-agent according to mobilization of the main agent. The main agent and the sub agents are cooperated to process the task of different types, compared with the scheme that the single and full-functional agent is used for processing the task of different types, the scheme has better processing effect on the single type of task, and the sub agents with relatively smaller scale are convenient for flexibly increasing and changing corresponding functions, and can bring better comprehensive task processing effect under the condition of lower comprehensive cost.

According to an embodiment of the present disclosure, there is further provided an electronic device, including at least one processor, and a memory communicatively connected to the at least one processor, where the memory stores instructions executable by the at least one processor, so that the at least one processor, when executing, is capable of implementing the multi-agent collaboration-based interactive question-answering task processing method described in any of the above embodiments.

According to an embodiment of the present disclosure, there is further provided a readable storage medium storing computer instructions for enabling a computer to implement the interactive question-answering task processing method based on multi-agent collaboration described in any of the above embodiments when executed.

According to an embodiment of the present disclosure, there is further provided a computer program product capable of implementing the multi-agent collaboration-based interactive question-answering task processing method described in any of the above embodiments when executed by a processor.

Fig. 11 illustrates a schematic block diagram of an example electronic device 1100 that can be used to implement embodiments of the present disclosure. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular telephones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the disclosure described and/or claimed herein.

As shown in fig. 11, the apparatus 1100 includes a computing unit 1101 that can perform various appropriate actions and processes according to a computer program stored in a Read Only Memory (ROM) 1102 or a computer program loaded from a storage unit 1108 into a Random Access Memory (RAM) 1103. In the RAM 1103, various programs and data required for the operation of the device 1100 can also be stored. The computing unit 1101, ROM 1102, and RAM 1103 are connected to each other by a bus 1104. An input/output (I/O) interface 1105 is also connected to bus 1104.

Various components in the device 1100 are connected to I/O interfaces 1105, including an input unit 1106, e.g., keyboard, mouse, etc., an output unit 1107, e.g., various types of displays, speakers, etc., a storage unit 1108, e.g., magnetic disk, optical disk, etc., and a communication unit 1109, e.g., network card, modem, wireless communication transceiver, etc. The communication unit 1109 allows the device 1100 to exchange information/data with other devices through a computer network such as the internet and/or various telecommunication networks.

The computing unit 1101 may be a variety of general purpose and/or special purpose processing components having processing and computing capabilities. Some examples of computing unit 1101 include, but are not limited to, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), various specialized Artificial Intelligence (AI) computing chips, various computing units running machine learning model algorithms, a Digital Signal Processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 1101 performs the respective methods and processes described above, such as an interactive question-and-answer task processing method based on multi-agent collaboration. For example, in some embodiments, the multi-agent collaboration-based interactive question and answer task processing method may be implemented as a computer software program tangibly embodied on a machine-readable medium, such as storage unit 1108. In some embodiments, some or all of the computer programs may be loaded and/or installed onto device 1100 via ROM 1102 and/or communication unit 1109. When the computer program is loaded into the RAM 1103 and executed by the computing unit 1101, one or more steps of the above-described multi-agent collaboration-based interactive question-answering task processing method can be performed. Alternatively, in other embodiments, the computing unit 1101 may be configured to perform the multi-agent collaboration-based interactive question-and-answer task processing method in any other suitable manner (e.g., by means of firmware).

Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuit systems, field Programmable Gate Arrays (FPGAs), application Specific Integrated Circuits (ASICs), application Specific Standard Products (ASSPs), systems On Chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include being implemented in one or more computer programs that are executable and/or interpretable on a programmable system including at least one programmable processor, which may be a special or general purpose programmable processor, operable to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

Program code for carrying out methods of the present disclosure may be written in any combination of one or more programming languages. These program code may be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus such that the program code, when executed by the processor or controller, causes the functions/operations specified in the flowchart and/or block diagram to be implemented. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package, partly on the machine and partly on a remote machine or entirely on the remote machine or server.

In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user, for example, feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback), and input from the user may be received in any form, including acoustic input, speech input, or tactile input.

The systems and techniques described here can be implemented in a computing system that includes a background component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such background, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a Local Area Network (LAN), a Wide Area Network (WAN), and the Internet.

The computer system may include a client and a server. The client and server are typically remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, also called a cloud computing server or a cloud host, and is a host product in a cloud computing service system, so as to solve the defects of large management difficulty and weak service expansibility in the traditional physical host and Virtual Private Server (VPS) PRIVATE SERVER service.

According to the technical scheme, an intelligent agent cluster formed by a pre-built main intelligent agent and a plurality of sub intelligent agents is adopted to process task demands proposed by users, wherein the main intelligent agent is responsible for understanding task demands of the users and decomposing the whole task demands into a plurality of sub-target tasks which can be respectively executed by different sub intelligent agents, and each sub intelligent agent processes the sub-tasks matched with the sub-intelligent agent according to the mobilization of the main intelligent agent. The main agent and the sub agents are cooperated to process the task of different types, compared with the scheme that the single and full-functional agent is used for processing the task of different types, the scheme has better processing effect on the single type of task, and the sub agents with relatively smaller scale are convenient for flexibly increasing and changing corresponding functions, and can bring better comprehensive task processing effect under the condition of lower comprehensive cost.

It should be appreciated that various forms of the flows shown above may be used to reorder, add, or delete steps. For example, the steps recited in the present disclosure may be performed in parallel or sequentially or in a different order, provided that the desired results of the technical solutions of the present disclosure are achieved, and are not limited herein.

The above detailed description should not be taken as limiting the scope of the present disclosure. It will be apparent to those skilled in the art that various modifications, combinations, sub-combinations and alternatives are possible, depending on design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present disclosure are intended to be included within the scope of the present disclosure.

Claims

1. A method for processing interactive question-answering tasks based on multi-agent collaboration, applied to a main agent, comprising:

Determine the target task to be solved based on the user's natural language input;

Decomposing the target task into a plurality of sub-target tasks including at least an interactive question-and-answer task, and issuing each of the sub-target tasks to each sub-target agent including at least an interactive question-and-answer agent; wherein different sub-agents are used to process different types of sub-tasks;

Controlling the interactive question-and-answer agent to present an interactive graphical user interface corresponding to the interactive question-and-answer task to the user, and presenting an answer result corresponding to the received feedback answer; wherein the interactive graphical user interface includes a question and interactive answer area;

Controlling an associated sub-goal agent among the plurality of sub-goal agents to generate a corresponding associated processing result based on the answer result output by the interactive question-and-answer agent; wherein the associated sub-goal agent is a sub-goal agent that depends on the answer result as part of the task input information;

A target processing result corresponding to the target task is determined based on the associated processing result.

2. The method according to claim 1, wherein determining the target task to be solved according to the user's natural language input comprises:

Converting the user's natural language input into natural language text;

Performing intent recognition on the natural language text to obtain an intent recognition result including the intent to learn and master the knowledge point;

Determine the target task to be solved according to the intention recognition result; wherein the target task includes a task for learning and mastering knowledge points.

3. The method according to claim 2, wherein the performing of intent recognition on the natural language text to obtain an intent recognition result including the knowledge point learning and mastering intent comprises:

Performing semantic understanding on the natural language text to obtain a semantic understanding result;

A preliminary intention is determined according to the semantic understanding result, and the preliminary intention is personalized modified using the personalized preference of the user to obtain an intention recognition result including the intention of learning and mastering the knowledge point.

4. The method according to claim 2, wherein the step of decomposing the target task into a plurality of sub-target tasks including at least an interactive question-and-answer task comprises:

Determine a plurality of task elements constituting the target task; wherein different task elements correspond to different types of tasks;

According to a single task element that specifically requires the user to answer the questions subjectively and checks whether the knowledge points to which the questions belong have been learned and mastered according to the answer results, an interactive question-and-answer task for learning and mastering the knowledge points is decomposed from the target task;

The target task is decomposed into a plurality of sub-target tasks each including only a single task element; wherein the sub-target tasks at least include the interactive question-and-answer task.

5. The method according to claim 4, wherein the interactive question-and-answer agent includes a knowledge point practice answering agent corresponding to a knowledge point practice phase and a knowledge point mastery test agent corresponding to a practice result verification phase.

6. The method according to claim 4, wherein the step of decomposing the target task into a plurality of sub-target tasks each containing only a single task element comprises:

In response to the existence of additional conditions corresponding to the task elements, the target task is decomposed into multiple sub-target tasks each containing only a single task element and the corresponding additional conditions; wherein the additional conditions include: a specified task processing method and/or a specified result presentation form.

7. The method according to claim 1, wherein the process of constructing different sub-agents comprises:

Determining a set of task types for creating sub-agents;

For each type of task in the task type set, a sub-agent having the task processing capability of the corresponding type of task is constructed.

8. The method according to claim 7, wherein for each type of task in the task type set, a sub-agent having the task processing capability of the corresponding type of task is constructed, comprising:

In response to the first target type task having at least two task processing methods, a sub-agent corresponding to each task processing method is constructed for the first target type task; wherein different task processing methods are abstracted from different processing logics for the same type of tasks.

9. The method according to claim 7, wherein for each type of task in the task type set, a sub-agent having the task processing capability of the corresponding type of task is constructed, comprising:

In response to the second target type task having at least two result presentation forms, a sub-agent corresponding to each result presentation form is constructed for the second target type task; wherein the result presentation form includes: at least one of: text type, table type, image type, video type, and interactive card type.

10. The method according to claim 1, wherein the step of controlling the interactive question-and-answer agent to present an interactive graphical user interface corresponding to the interactive question-and-answer task to the user and presenting an answer result corresponding to the received feedback answer comprises:

Controlling the interactive question-and-answer agent to determine the question and the interactive answer area based on the received task input information; wherein the received task input information is determined based on the task description information related to the interactive question-and-answer task decomposed from the target task;

Controlling the interactive question-answering agent to present an interactive graphical user interface including the question and the interactive answering area to the user;

The interactive question-and-answer agent is controlled to present an answer result determined according to the degree of difference between the feedback answer received from the interactive answering area and the standard answer to the question.

11. The method according to claim 1, further comprising:

In response to the feedback answer being an incorrect answer different from the standard answer, the interactive question-and-answer agent is controlled to regenerate an interactive graphical user interface including a new question different from the question corresponding to the last incorrect answer and an interactive answering area matching the new question; wherein the new question is generated based on the question corresponding to the last correct answer.

12. The method according to claim 11, further comprising:

In response to obtaining a preset number of answer results, controlling the interactive question-and-answer agent to output prompt information indicating that the interactive question-and-answer is completed;

Correspondingly, the controlling of the associated sub-goal agent among the plurality of sub-goal agents to generate corresponding associated processing results based on the answer results output by the interactive question-and-answer agent includes:

In response to receiving the prompt information, the accuracy calculation agent is controlled to determine the answer accuracy according to the preset number of answer results.

13. The method according to claim 12, wherein the determining the target processing result corresponding to the target task based on the associated processing result comprises:

Controlling a learning suggestion agent among the plurality of sub-goal agents to output matching learning suggestions according to the answer accuracy rate;

The number of answer results, the answer accuracy rate and the learning suggestions are summarized to obtain a target processing result corresponding to the target task.

14. The method according to claim 10, wherein the received task input information is determined based on task description information related to the interactive question-and-answer task decomposed from the target task and the personalized preference of the user.

15. The method according to claim 14, wherein the personalized preference comprises at least one of the following:

Age, education level, areas of interest, historical consumption preferences.

16. The method according to 15, wherein controlling the interactive question-answering agent to determine the question based on the received task input information comprises:

In response to the personalized preference including the age, the interactive question-and-answer agent is controlled to determine questions that match the cognitive level corresponding to the age based on the received task input information.

17. The method according to any one of claims 1 to 16, further comprising:

In response to the instruction input box being in a selected state, controlling the currently executed sub-agent to pause outputting the corresponding sub-processing result; wherein the instruction input box is used for the user to input instructions, and the instruction input box is in an unselected state during the process of the sub-agent outputting the corresponding sub-processing result;

In response to a new instruction being generated in the process of the instruction input box being restored from the selected state to the unselected state, extracting correction information from the new instruction;

Determining the sub-target agent affected by the correction information;

Control the affected sub-target agent to re-output the corresponding sub-processing result in combination with the correction information.

18. An interactive question-answering task processing device based on multi-agent collaboration, applied to a main agent, comprising:

A target task determination unit, configured to determine a target task to be solved according to a natural language input of a user;

The task decomposition and corresponding distribution unit is configured to decompose the target task into a plurality of sub-target tasks including at least an interactive question-and-answer task, and distribute each of the sub-target tasks to each sub-target agent including at least an interactive question-and-answer agent; wherein different sub-agents are used to process different types of sub-tasks;

An interactive interface control and presentation unit is configured to control the interactive question-and-answer agent to present an interactive graphical user interface corresponding to the interactive question-and-answer task to the user, and present an answer result corresponding to the received feedback answer; wherein the interactive graphical user interface includes a question and interactive answer area;

an associated processing result control generating unit, configured to control an associated sub-goal agent among the plurality of sub-goal agents to generate a corresponding associated processing result based on the answer result output by the interactive question-and-answer agent; wherein the associated sub-goal agent is a sub-goal agent that depends on the answer result as part of the task input information;

The target processing result determining unit is configured to determine the target processing result corresponding to the target task based on the associated processing result.

19. The apparatus according to claim 18, wherein the target task determination unit comprises:

a conversion subunit, configured to convert the user's natural language input into a natural language text;

An intention recognition subunit is configured to perform intention recognition on the natural language text to obtain an intention recognition result including the intention of learning and mastering the knowledge point;

The target task determination subunit is configured to determine the target task to be solved according to the intention recognition result; wherein the target task includes a task for learning and mastering knowledge points.

20. The apparatus according to claim 19, wherein the intention identification subunit is further configured to:

21. The device according to claim 19, wherein the task decomposition and corresponding distribution unit comprises a task decomposition subunit configured to decompose the target task into a plurality of sub-target tasks including at least an interactive question-and-answer task, the task decomposition subunit comprising:

A task element determination module is configured to determine a plurality of task elements constituting the target task; wherein different task elements correspond to different types of tasks;

The interactive question-and-answer task decomposition module is configured to decompose the interactive question-and-answer task for learning and mastering the knowledge point from the target task according to a single task element that specifically requires the user to give a subjective answer to the question and to check whether the knowledge point to which the question belongs has been learned and mastered according to the answer result;

The sub-goal task decomposition module is configured to decompose the target task into multiple sub-goal tasks that only contain a single task element; wherein the sub-goal tasks at least include the interactive question-and-answer task.

22. The device according to claim 21, wherein the interactive question-and-answer agent comprises a knowledge point practice answering agent corresponding to a knowledge point practice phase and a knowledge point mastery testing agent corresponding to a practice achievement verification phase.

23. The apparatus according to claim 21, wherein the sub-goal task decomposition module is further configured to:

24. The apparatus according to claim 18, further comprising: a sub-agent construction unit configured to construct different sub-agents, the sub-agent construction unit comprising:

A sub-unit is created, which is configured to construct a sub-agent with task processing capability of the corresponding type of task for each type of task in the task type set.

25. The apparatus according to claim 24, wherein the creating subunit is further configured to:

26. The apparatus according to claim 24, wherein the creating subunit is further configured to:

In response to the second target type task having at least two result presentation forms, a sub-agent corresponding to each result presentation form is constructed for the second target type task; wherein the result presentation form includes: at least one of text, table, image, video, and interactive card types.

27. The device according to claim 18, wherein the interactive interface control presentation unit comprises:

A first control subunit is configured to control the interactive question-and-answer agent to determine the question and the interactive answer area based on the received task input information; wherein the received task input information is determined based on the task description information related to the interactive question-and-answer task decomposed from the target task;

A second control subunit is configured to control the interactive question-answering agent to present an interactive graphical user interface including the question and the interactive answering area to the user;

The third control subunit is configured to control the interactive question-and-answer agent to present an answer result determined according to the degree of difference between the feedback answer received from the interactive answering area and the standard answer to the question.

28. The apparatus of claim 18, further comprising:

A regeneration control unit is configured to control the interactive question-and-answer agent to regenerate an interactive graphical user interface including a new question different from the question corresponding to the last wrong answer and an interactive answer area matching the new question in response to the feedback answer being an incorrect answer different from the standard answer; wherein the new question is generated based on the question corresponding to the last correct answer.

29. The apparatus according to claim 28, further comprising:

A prompt information output control unit, configured to control the interactive question-and-answer agent to output prompt information indicating that the interactive question-and-answer is completed in response to obtaining a preset number of answer results;

Correspondingly, the association processing result control generation unit is further configured to:

30. The apparatus according to claim 29, wherein the target processing result determination unit is further configured to:

31. The device according to claim 27, wherein the received task input information is determined based on task description information related to the interactive question-and-answer task decomposed from the target task and the personalized preference of the user.

32. The apparatus of claim 31 , wherein the personalized preference comprises at least one of the following:

Age, education level, areas of interest, historical consumption preferences.

33. The apparatus according to 32, wherein the first control subunit comprises a question determination control module configured to control the interactive question-answering agent to determine the question based on the received task input information, and the question determination control module is further configured to:

34. The apparatus according to any one of claims 18 to 33, further comprising:

A pause output control unit is configured to control the currently executed sub-agent to pause outputting the corresponding sub-processing result in response to the instruction input box being in a selected state; wherein the instruction input box is used for the user to input instructions, and the instruction input box is in an unselected state during the process of the sub-agent outputting the corresponding sub-processing result;

a correction information extraction unit, configured to extract correction information from a new instruction in response to the instruction input box generating a new instruction in the process of restoring from the selected state to the unselected state;

An affected sub-target agent determining unit, configured to determine the sub-target agent affected by the correction information;

The re-output control unit is configured to control the affected sub-target agent to re-output the corresponding sub-processing result in combination with the correction information.

35. An electronic device comprising:

at least one processor; and

a memory communicatively connected to the at least one processor; wherein,

The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the interactive question-and-answer task processing method based on multi-agent collaboration as described in any one of claims 1-17.

36. A non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable the computer to execute the interactive question-answering task processing method based on multi-agent collaboration as described in any one of claims 1-17.

37. A computer program product, comprising a computer program, which, when executed by a processor, implements the steps of the interactive question-and-answer task processing method based on multi-agent collaboration according to any one of claims 1 to 17.