Information processing method and device, storage medium and computer device

Abstract

The application discloses an information processing method and device, a storage medium and a computer device. The method comprises the following steps: a main intelligent agent performs intention recognition on user context information to obtain a target task to be executed, wherein the target task comprises at least one subtask; the main intelligent agent determines a sub intelligent agent for executing the subtask and a task execution authority of the sub intelligent agent based on task content of the subtask and description information of the sub intelligent agent; the main intelligent agent sets a task identifier for the subtask, and sends the task identifier, the task content and the task execution authority of the subtask to the sub intelligent agent; the sub intelligent agent determines a task execution range based on the task content of the subtask and the task execution authority, executes the subtask in the task execution range, and obtains a task execution result; and the main intelligent agent generates reply information based on the task execution range and the task execution result of the sub intelligent agent, and sends the reply information to a client. The above method can reduce the training cost of a model and improve the problem solving ability of the model.

Description

Technical Field

[0001] This application relates to the field of Internet technology, and in particular to an information processing method, apparatus, storage medium and computer equipment. Background Technology

[0002] In the e-commerce field, large-scale modeling technology has been widely applied in various scenarios such as information recommendation, information search, and intelligent customer service to improve user experience and operational efficiency. In existing technologies, large-scale models primarily utilize pre-trained knowledge bases or invoke pre-defined tools to respond to user questions. However, the questions raised by users during actual interactions are often highly diverse and specialized, which places very high demands on the generalization ability and service adaptability of large-scale models.

[0003] Currently, although large models have acquired a certain knowledge coverage capability through training on massive amounts of data, their training cost increases exponentially with the expansion of knowledge coverage. However, their problem-solving ability is limited by the timeliness of static knowledge bases and the flexibility of tool interfaces, making it difficult to meet the high-frequency update needs of users in e-commerce scenarios. In addition, when user questions involve cross-domain knowledge, large models often cannot effectively connect multi-dimensional information, resulting in deviations between their response results and user expectations, thus affecting user experience. Summary of the Invention

[0004] In view of this, embodiments of this application provide an information processing method, apparatus, storage medium, and computer equipment, the main purpose of which is to solve the technical problems of high model training costs and limited problem-solving capabilities.

[0005] According to a first aspect of this application, an information processing method is provided, the method comprising:

[0006] The user context information is obtained, and the main intelligent agent performs intent recognition on the user context information to obtain the target task to be executed, wherein the target task includes at least one subtask.

[0007] The main agent determines the sub-agent to execute the sub-task and the task execution permissions of the sub-agent based on the task content of the sub-task and the description information of the sub-agent. The task execution permissions include full execution permissions and partial execution permissions.

[0008] The main intelligent agent sets a task identifier for the subtask and sends the task identifier, task content, and task execution permissions of the subtask to the sub-intelligent agent executing the subtask;

[0009] The sub-agent determines a task execution range based on the task content of the sub-task and the task execution authority, and executes the sub-task in the task execution range to obtain a task execution result; wherein, when executing the sub-task, if it is determined that there is a missing task parameter, the sub-agent sends parameter information of the missing task parameter to the main agent; the main agent calls a parameter collection tool based on the parameter information, and obtains the missing task parameter through the parameter collection tool, and sends the task parameter to the sub-agent; the sub-agent executes the sub-task based on the task parameter to obtain a task execution result;

[0010] The main agent generates reply information based on the task execution range and the task execution result of the sub-agent, and sends the reply information to the client.

[0011] According to a second aspect of the present application, an information processing method is provided, which comprises:

[0012] In response to receiving user input information, the user input information is sent to a server to enable the server to obtain user context information corresponding to the user input information, and an intent of the user context information is recognized by a main agent to obtain a target task to be executed, wherein the target task comprises at least one sub-task; the main agent determines a sub-agent for executing the sub-task and a task execution authority of the sub-agent based on the task content of the sub-task and description information of the sub-agent, wherein the task execution authority comprises full execution authority and partial execution authority; the main agent sets a task identifier for the sub-task, and sends the task identifier, the task content and the task execution authority of the sub-task to the sub-agent for executing the sub-task; the sub-agent determines a task execution range based on the task content of the sub-task and the task execution authority, and executes the sub-task in the task execution range to obtain a task execution result; wherein, when executing the sub-task, if it is determined that there is a missing task parameter, the sub-agent sends parameter information of the missing task parameter to the main agent; the main agent calls a parameter collection tool based on the parameter information, and obtains the missing task parameter through the parameter collection tool, and sends the task parameter to the sub-agent; the sub-agent executes the sub-task based on the task parameter to obtain a task execution result; the main agent generates reply information based on the task execution range and the task execution result of the sub-agent;

[0013] The reply information is received and displayed.

[0014] According to a third aspect of the present application, an information processing device is provided, which comprises:

[0015] a main intelligent agent configured to acquire user context information, perform intent recognition on the user context information, and obtain a target task to be executed, wherein the target task comprises at least one subtask;

[0016] The main intelligent agent is further configured to determine a subintelligent agent for executing the subtask and a task execution authority of the subintelligent agent based on task content of the subtask and description information of the subintelligent agent, wherein the task execution authority comprises full execution authority and partial execution authority.

[0017] The main intelligent agent is further configured to set a task identifier for the subtask, and send the task identifier, the task content, and the task execution authority of the subtask to the subintelligent agent for executing the subtask.

[0018] The subintelligent agent is configured to determine a task execution range based on the task content of the subtask and the task execution authority, execute the subtask in the task execution range, and obtain a task execution result. When executing the subtask, if it is determined that there is a missing task parameter, the subintelligent agent sends parameter information of the missing task parameter to the main intelligent agent. The main intelligent agent calls a parameter collection tool based on the parameter information, acquires the missing task parameter through the parameter collection tool, and sends the task parameter to the subintelligent agent. The subintelligent agent executes the subtask based on the task parameter, and obtains a task execution result.

[0019] The main intelligent agent is further configured to generate reply information based on the task execution range and the task execution result of the subintelligent agent, and send the reply information to a client.

[0020] According to a fourth aspect of the present application, an information processing apparatus is provided, which comprises:

[0021] The information sending module is configured to, in response to receiving user input information, send the user input information to a server, so that the server acquires user context information corresponding to the user input information, and performs intent recognition on the user context information by a main intelligent agent to obtain a target task to be executed, wherein the target task includes at least one subtask; the main intelligent agent determines a subintelligent agent for executing the subtask and a task execution authority of the subintelligent agent based on task content of the subtask and description information of the subintelligent agent, wherein the task execution authority includes full execution authority and partial execution authority; the main intelligent agent sets a task identifier for the subtask, and sends the task identifier, the task content and the task execution authority of the subtask to the subintelligent agent for executing the subtask; the subintelligent agent determines a task execution range based on the task content of the subtask and the task execution authority, and executes the subtask in the task execution range to obtain a task execution result; wherein when executing the subtask, if it is determined that there is a missing task parameter, the subintelligent agent sends parameter information of the missing task parameter to the main intelligent agent; the main intelligent agent calls a parameter collection tool based on the parameter information, and acquires the missing task parameter by using the parameter collection tool, and sends the task parameter to the subintelligent agent; the subintelligent agent executes the subtask based on the task parameter to obtain the task execution result; and the main intelligent agent generates reply information based on the task execution range and the task execution result of the subintelligent agent.

[0022] The information display module is configured to receive and display the reply information.

[0023] According to a fifth aspect of the present application, a storage medium having a computer program stored thereon is provided, and the program is executed by a processor to implement the information processing method.

[0024] According to a sixth aspect of the present application, a computer device is provided, which includes a storage medium, a processor and a computer program stored in the storage medium and executable on the processor, and the processor implements the information processing method when executing the program.

[0025] By means of the technical solutions, the information processing method, device, storage medium and computer equipment provided by the embodiment of the application can perform intention recognition and task decomposition on user context information by using the main intelligent agent, can decompose the core demand of the user into operation units that can be independently executed. Moreover, by assigning suitable sub-intelligent agents to each sub-task, assigning task execution authority to each sub-intelligent agent, and determining the task execution range by the sub-intelligent agent, the execution of the task can be accurately evaluated before the task is executed, so as to improve the accuracy and controllability of the task execution result. Finally, by using the task execution range and the task execution result of the sub-intelligent agent to generate the reply content, the problem solving ability of different sub-intelligent agents in their respective fields can be utilized, the ability of the model to solve problems across fields and the ability to associate multi-dimensional information are improved, so as to reduce the training difficulty of the main intelligent agent. The above method can effectively improve the problem solving ability and task execution ability of the model, so that the model has higher flexibility and timeliness, thereby improving the experience of the user.

[0026] The above description is only a summary of the technical solutions of the application. In order to enable the technical means of the application to be more clearly understood, the application can be implemented according to the content of the description, and in order to enable the above and other purposes, characteristics and advantages of the application to be more obvious and easy to understand, the following specific embodiments of the application are described. BRIEF DESCRIPTION OF DRAWINGS

[0027] The drawings described herein are used to provide further understanding of the application, and form a part of the application. The schematic embodiments of the application and the description thereof are used to explain the application, and do not constitute an improper limitation on the application. In the drawings:

[0028] Figure 1 A flowchart of an information processing method provided by an embodiment of the application is shown;

[0029] Figure 2 A flowchart of another information processing method provided by an embodiment of the application is shown;

[0030] Figure 3 A scene diagram of an information processing method provided by an embodiment of the application is shown;

[0031] Figure 4 A flowchart of another information processing method provided by an embodiment of the application is shown;

[0032] Figure 5 A scene diagram of another information processing method provided by an embodiment of the application is shown;

[0033] Figure 6 A flowchart of the interaction between the main intelligent agent and the sub-intelligent agent in the execution of the first embodiment of an information processing method provided by an embodiment of the application is shown.

[0034] Figure 7 Fig. 1 shows an interaction flow diagram of the main agent and the sub-agent when the information processing method according to the embodiment of the present application is executed in the second implementation manner;

[0035] Figure 8 Fig. 2 shows an interaction flow diagram of the main agent and the sub-agent when the information processing method according to the embodiment of the present application is executed in the third implementation manner;

[0036] Figure 9 Fig. 3 shows an interaction flow diagram of the main agent and the sub-agent when the information processing method according to the embodiment of the present application is executed in the fourth implementation manner;

[0037] Figure 10 Fig. 4 shows an interaction flow diagram of the main agent and the sub-agent when the information processing method according to the embodiment of the present application is executed in the fifth implementation manner;

[0038] Figure 11 Fig. 5 shows a structure diagram of a memory system according to the embodiment of the present application;

[0039] Figure 12 Fig. 6 shows a structure diagram of an information processing apparatus according to the embodiment of the present application;

[0040] Figure 13 Fig. 7 shows a structure diagram of another information processing apparatus according to the embodiment of the present application. DETAILED DESCRIPTION

[0041] The present application will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments. It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0042] In one embodiment, as shown in Fig. 1, an information processing method is provided, which is taken as an example of being applied to a computer device such as a server, and includes the following steps: Figure 1

[0043] Step 101, the client sends the user input information to the server in response to receiving the user input information.

[0044] ​Specifically, the user can input information through any interface of the client. For example, the user can send conversation information through the client by initiating a conversation, or can send search keywords and other information through the client by inputting search keywords. In this embodiment, the purpose of the user inputting information is to solve a certain problem, such as an information search problem, a query problem, a consultation problem, a complaint problem, and the like, and the application scenarios are very wide, and thus the user identity can also be set according to the actual scenario. For example, in a shopping scenario, the user can be a shopping customer; in a merchant service scenario, the user can be a merchant.

[0045] In step 102, the server obtains user context information and performs intent recognition on the user context information through a main agent to obtain a target task to be executed, the target task including at least one subtask.

[0046] The user context information refers to historical behavior data and real-time request content generated by the user in the process of interacting with the e-commerce platform. For example, the user context information can include user search keywords, current conversation content, and user labels, and other multi-dimensional information. The main agent refers to an intelligent decision-making module with global task scheduling capability, which can be implemented through a pre-trained information processing model. The information processing model refers to an artificial intelligence model capable of processing a certain amount of tasks and a certain amount of data, such as a large language model (LLM), a computer vision model (CV), a multi-modal model (LMMs), and a small model formed through model distillation technology.

[0047] Further, the intent recognition refers to a process of analyzing the core demands of the user by the natural language processing capability provided by the information processing model of the main agent. The target task refers to a complete solution target corresponding to the user problem, and the subtask refers to a specific operation unit that needs to be executed for the target task. The target task can be a single task (i.e., only including one subtask), or can be decomposed into multiple subtasks. The subtasks can be in parallel relationship, or can be in logical sequence.

[0048] Specifically, when the server receives the user input information, the main agent with the natural language processing function can obtain the user context information, and perform intent recognition on the user context information to obtain the core appeal of the user, that is, the target task to be executed. Then, the main agent can decompose the target task into one or more sub-tasks according to actual needs. Through the above steps, the complex user demand can be converted into a more easily processed modular operation unit by performing intent recognition on the user context information and decomposing the recognized target task, so as to facilitate more detailed and accurate understanding of the user's intention.

[0049] In step 103, the main agent determines the sub-agent that executes the sub-task and the task execution authority of the sub-agent based on the task content of the sub-task and the description information of the sub-agent.

[0050] The sub-agent refers to an intelligent decision-making module with problem-solving capability in a certain professional field. The intelligent decision-making module can also be implemented by a pre-trained information processing model, but in model training, the sub-agent focuses more on using data in the corresponding field for vertical training, so that the model has more accurate and efficient problem-solving capability in a certain field. Correspondingly, the description information of the sub-agent refers to the explanatory information that is used to describe the functional boundaries and technical parameters of the sub-agent. Further, the task execution authority refers to the operation authority level of the sub-agent for processing the sub-task, which is given by the main agent after intent recognition and task decomposition. Specifically, it can include full execution authority and partial execution authority, etc.

[0051] Specifically, after obtaining the task content of each sub-task and obtaining the description information of each sub-agent, the main agent can match the task content of the sub-task and the description information of the sub-agent, and determine the sub-agent that executes each sub-task and the task execution authority of each sub-agent for the sub-task based on the matching result. In this embodiment, when the target task is not split (that is, the target task only contains one sub-task), the main agent can directly assign a sub-agent to the target task according to the task content of the target task, and determine the task execution authority of the sub-agent for the target task; when the target task is split into multiple sub-tasks, the main agent can assign a sub-agent to each sub-task according to the task content of each sub-task, and determine the task execution authority of each sub-agent for the sub-task.

[0052] In this embodiment, the master agent can split the target task according to the description information of the sub-agent, that is, the master agent can split the target task according to the field to which the task content belongs, instead of splitting the target task according to the number of tasks. In this way, the sub-agent can execute the corresponding sub-task of the field and obtain a more accurate execution result. In this process, if a suitable sub-agent cannot be matched for the sub-task, the task allocation can not be performed, and the master agent directly completes the execution of the corresponding sub-task and obtains the task execution result. The above steps can accurately evaluate the execution of the task before the execution of the task by allocating suitable sub-agents to each sub-task and allocating task execution authority to each sub-agent, thereby improving the accuracy and controllability of the task execution result of the sub-task.

[0053] In step 104, the master agent sets a task identifier for the sub-task, and sends the task identifier, task content and task execution authority of the sub-task to the sub-agent executing the sub-task.

[0054] The task identifier refers to a unique identifier generated by the master agent for the sub-task, which is used to track the execution status of the sub-task throughout the process. Specifically, after the task is disassembled, the master agent can set a unique task identifier for the sub-task, and send the set task identifier, disassembled task content and task execution authority of the sub-agent to the sub-agent executing the sub-task, so as to facilitate subsequent processing of the sub-agent. In this embodiment, the master agent can allocate a task identifier to the sub-task executed by the sub-agent, so as to facilitate the tracking of the task execution status, and also facilitate the storage and backtracking of the task data; as for the sub-task directly executed by the master agent, it can be determined as appropriate, that is, the task identifier can be set or not set.

[0055] In step 105, the sub-agent determines the task execution range based on the task content and the task execution authority of the sub-task, and executes the sub-task in the task execution range to obtain the task execution result.

[0056] The task execution range refers to the task execution degree re-determined by the sub-agent within the task execution authority after evaluation, which can specifically include full execution, partial execution and no execution. Specifically, after receiving the task content and task execution authority of the sub-task, the sub-agent can re-evaluate the task content of the sub-task, and determine the task execution range within the task execution authority determined by the master agent, and then execute the sub-task within the task execution range to obtain the task execution result of the sub-task.

[0057] In this embodiment, the task execution range determined by the sub-agent is equal to or lower than the task execution authority determined by the main agent for the sub-agent. For example, when the task execution authority is full execution authority, the task execution range can be full execution, partial execution or no execution; when the task execution authority is partial execution authority, the task execution range can only be partial execution or no execution. In this way, it can be ensured that the sub-agent can accurately execute the target task or part of the target task within the capability range.

[0058] In step 106, the main agent generates reply information based on the task execution range and the task execution result of the sub-agent, and sends the reply information to the client.

[0059] The reply information refers to the final response content generated by the main agent based on the task execution result of the sub-agent, which can include text content, visual charts, interactive components, etc. Specifically, after receiving the task execution range and the task execution result sent by the sub-agent, the main agent can first determine the completion degree of the target task based on the task execution range, and then directly output the task execution result of the sub-task based on the completion degree of the target task, or re-integrate, supplement or call other tools to execute the target task or supplement the execution content to obtain the final reply information sent to the client.

[0060] In this embodiment, the main agent can process the task execution result of the sub-agent to generate the reply content according to the different task execution ranges of the sub-agent. For example, when the task execution range of the sub-agent is full execution, the main agent can directly generate the reply content according to the feedback result of the sub-agent to reduce the data processing amount of the main agent; when the task execution range of the sub-agent is partial execution, the main agent can arrange, integrate or call other tools to supplement the feedback result of the sub-agent to generate the reply information, thereby perfecting the reply content; when the task execution range of the sub-agent is no execution, the main agent can determine the task execution steps and the tools to be called according to the task content of the target task, and generate the reply information, thereby ensuring the accuracy of the reply content.

[0061] In step 107, the client receives and displays the reply information.

[0062] Specifically, after receiving the reply information, the client can display the reply information through various components. Specifically, the client can display the reply information through a text stream, a picture-text display component, and an interactive component in multiple dimensions, wherein the various components can be displayed in a streaming manner according to actual conditions. For example, when the reply information contains text information, chart information, and action point information, the client can first display the chart through the picture-text display component, then display the text stream information, and finally output the action point card. Based on this, the user can first view the chart, then view the text information, and finally decide whether to adopt the solution provided by the action point card. If it is decided to adopt, the action point adoption information can be submitted through the action point card.

[0063] The above embodiments can decompose the core demand of the user into operation units that can be independently executed by using the main agent to recognize the intention of the user context information and decompose the task. Moreover, by assigning suitable sub-agents to each sub-task, assigning task execution authority to each sub-agent, and determining the task execution range by the sub-agent, the execution of the task can be accurately evaluated before the task is executed, thereby improving the accuracy and controllability of the task execution result. Finally, by using the task execution range and the task execution result of the sub-agent to generate the reply content, the problem solving ability of different sub-agents in their respective fields can be utilized to improve the ability of the model to solve problems across fields and the ability to associate multi-dimensional information, thereby reducing the training difficulty of the main agent. The above method can effectively improve the problem solving ability and task execution ability of the model, making the model more flexible and timely, thereby improving the user experience.

[0064] In the above embodiments, as shown in Figure 2 The information processing method further includes the following steps:

[0065] Step 201, when the sub-agent executes the sub-task, if it is determined that there is a missing task parameter, the parameter information of the missing task parameter is sent to the main agent.

[0066] Step 202, the main agent calls the parameter collection tool based on the parameter information of the missing task parameter, and sends the parameter information of the task parameter to the client through the parameter collection tool.

[0067] Step 203, the client displays the parameter collection component, wherein the parameter collection component displays the parameter information of the task parameter.

[0068] Step 204, the client receives the task parameter through the parameter collection component, and sends the task parameter to the main agent, and the main agent sends the task parameter to the sub-agent.

[0069] In step 205, the sub-agent executes the sub-task based on the task parameter to obtain a task execution result.

[0070] Specifically, steps 201-205 are further supplements to the above embodiments. In this embodiment, the sub-agent can determine whether there is a missing task parameter in the task execution process when determining the task execution range based on the task content of the sub-task and the task execution authority. If it is determined that there is a missing task parameter, the parameter collection tool can be called by the main agent, and the main agent can ask the client for the task parameter. After the main agent obtains the task parameter, the main agent can send the task parameter to the sub-agent, so that the sub-agent can continue to execute the sub-task and obtain the task execution result of the sub-task. In this embodiment, when the sub-agent requests the task parameter from the client through the main agent, the task identifier needs to be carried for data transmission, so as to facilitate data backtracking, storage, and task state tracking.

[0071] For example, in the "flash shopping merchant service scenario", it is assumed that the main agent assigns the sub-task "how is the recent good comment rate" to a sub-agent. When the sub-agent executes the sub-task, if it is determined that the necessary time range information is missing, the sub-agent can request the time range information from the client through the above method. At this time, the client displays the corresponding parameter collection component to collect the missing task parameter. After the user feeds back the time range information "20XX0801-20XX0831" through the client, the main agent can feed back the time range information to the sub-agent, so that the sub-agent can execute the query task of the good comment rate. Figure 3

[0072] The above embodiments can use the main agent to interact with the client and the sub-agent respectively, so that the main agent can be used as an access to interact with the client, thereby avoiding repeated development of the interaction interface of each sub-agent, reducing the development workload of the system, and reducing the complexity of the system.

[0073] In one embodiment, step 106 can be implemented in the following manner: the main agent calls at least one information display tool based on the reply information, and sends the reply information to the client through the information display tool, wherein the information display tool can include at least one of a text and image display tool, a text stream display tool, and an action point display tool. Correspondingly, in step 107, the client can display the reply information fed back by the main agent through at least one preset component, wherein the preset component can include at least one of a text and image display component, a text stream display component, and an action point display component.

[0074] ​In the above embodiments, after generating the reply information, the main intelligent agent can call the corresponding information display tool based on the content contained in the reply information to perform multi-dimensional display on the reply information. For example, referring to Figure 5 , the reply information can simultaneously contain information query results, reply script information, and action point information. At this time, the main intelligent agent can first call the graphic-text display tool to send the information query results, then call the text flow display tool to send the script information, and finally call the action point display tool to send the action point information. Correspondingly, after receiving the above information, the client can first display the query information through the graphic-text display component, then display the script information through the text flow display component, and finally display the action point information through the action point display component.

[0075] The above embodiments can make the user obtain multi-dimensional information such as graphs, text flows, and action points by using multiple information display tools to send the reply information and using multiple preset components to display the reply information. Meanwhile, other tools and components can be nested in these tools and components to further enrich the form of information display. For example, in the graphic-text display component, multiple graphs can be displayed simultaneously, and corresponding explanatory text of the graphs can be displayed; in the text flow display component, in addition to the script information, other content such as links and flowcharts can be displayed; in the action point display component, a solution can be directly provided for the user to facilitate the user to send the action point adoption information by clicking a button, or an action point setting portal can be provided for the user to facilitate the user to directly perform function setting through the action point card. Through the above manner, on the one hand, the manner of information display can be enriched, and on the other hand, the efficiency of problem solving can be improved.

[0076] In one embodiment, as shown in Figure 4 , the above information processing method further includes the following steps:

[0077] Step 301: The client displays the action point information through the action point display component.

[0078] Step 302: The client receives the action point adoption information through the action point display component and sends the action point adoption information to the server.

[0079] Step 303: The server performs action point processing based on the action point adoption information in response to receiving the action point adoption information, and sends a message that the action point execution is successful to the client.

[0080] Step 304: The client receives and displays the message that the action point execution is successful.

[0081] In the above embodiment, when the reply information generated by the main intelligent agent contains the action point information, the action point display tool can be invoked to send the action point information to the client, so that the client displays the action point information through the action point display component. The action point information refers to an executable action plan provided by the main intelligent agent or the sub-intelligent agent for the user based on the target task, and belongs to part of the task execution result. Based on the information, the user can decide whether to adopt it, and if so, the action point adoption information can be sent to the server through the action point display component, so that the server performs the corresponding operation, and then sends the action point execution success message to the client, so that the client displays the action point execution success message, so that the user can know the execution status of the action point in time.

[0082] In the embodiment, after the server sends the action point execution success message to the client, the client can first forward the message to the main intelligent agent, and the main intelligent agent can invoke the information display tool based on the information and send the action point execution success message to the client through the information display tool, so that the client can display the action point execution success message in the form of a component, thereby improving the user experience. In addition, for some action point information that needs to change the original user settings, the information display tool can be invoked to send the action point execution confirmation information to the client again before execution, and after receiving the user confirmation message, the server can process the action point, so as to avoid the user's disturbance caused by the action point misprocessing.

[0083] For example, in the "flash shopping merchant service scene", referring to Figure 5 , assuming that the reply information generated by the main intelligent agent contains the action point information of "starting order receiving reminder and automatic order receiving", the main intelligent agent can display the action point information of "starting order receiving reminder" and "starting automatic order receiving" through the action point display component after receiving the information. The user can select the action point to be executed according to the user's needs. When the user wants to start "automatic order receiving", the user can click the start button corresponding to the "automatic order receiving" component to send the action point adoption information. The server can start "automatic order receiving" based on the information, and send the action point execution success message to the client, so that the user can know the execution status of the action point.

[0084] In an embodiment, the task execution authority determined by the master agent for the sub-agent can include full execution authority and partial execution authority, and the task execution range determined by the sub-agent based on the task execution authority is lower than or equal to the task execution authority assigned by the master agent, and can specifically include full execution, partial execution, and no execution. Based on the above task allocation and execution manner, five kinds of task processing manners can be generated for different task execution authorities and task execution ranges, which are "full execution authority and full execution", "full execution authority but partial execution", "full execution authority but no execution", "partial execution authority and partial execution", and "partial execution authority but no execution". Taking the "flash shopping merchant service scene" as an example, the implementation manners of the above five kinds of task processing manners are specifically described.

[0085] Implementation manner one: full execution authority and full execution. As shown in the following table, in this task processing manner, the interaction mode between the master agent and the sub-agent is as follows: Figure 6

[0086] Step 1031, the master agent determines the sub-agent that executes the sub-task based on the task content of the sub-task and the description information of the sub-agent, and determines the task execution authority of the sub-agent as full execution authority.

[0087] Step 1041, the master agent sets a task identifier for the sub-task, and sends the task identifier, task content and full execution authority of the sub-task to the sub-agent that executes the sub-task.

[0088] Step 1051, when the sub-agent determines that the task execution range is full execution based on the task content of the sub-task and the full execution authority, the sub-agent executes the full content of the sub-task to obtain the task execution result.

[0089] Step 1061, the master agent generates reply information based on the task execution result of the sub-agent.

[0090] ​For example, in the "flash shopping merchant service scenario", when the user inputs the information "recent traffic is not very good, how to improve", the main agent determines the task content of the target task as "view the recent traffic and explain how to improve" according to the user context information. Then, the main agent can determine that a certain sub-agent can completely solve the task based on the description information of the sub-agent, and assign the "full execution authority" to the sub-agent. Then, the main agent can send the task identifier, task content and task execution authority of the task to the sub-agent. After analysis, the sub-agent thinks that it can directly answer the current question, but lacks the time range information for viewing traffic. Then, the time confirmation information is sent to the main agent, the main agent calls the tool to request the time range information from the client, and sends the requested time range information "20XX0801-20XX0831" to the sub-agent. The sub-agent queries the traffic based on the above time range information, and then generates the reply merchant script information and provides the action point. Then, the above processing result is sent to the main agent. After receiving the above information, the main agent can call multiple tools to send the above information to the client, so that the client can display the traffic card, the reply merchant script and the action point card in turn through multiple components, so as to help the merchant understand the reason for the problem and obtain the corresponding solution.

[0091] In the above embodiment, when the main agent determines that the sub-agent can completely solve the problem, it can assign the "full execution authority" to the sub-agent. After re-evaluation, the sub-agent thinks that it can indeed solve the current problem, and will feedback a task execution range to the main agent, and continue to execute the task within the range. When the main agent receives the "full execution" task execution range feedback by the sub-agent, it can give the sub-agent the full authority to solve the problem, and directly output the reply information using the processing result of the sub-agent, so as to avoid repeated processing of the task, save computing resources, and at the same time make the task get more professional processing.

[0092] Embodiment two: full execution authority but partial execution. As shown in Figure 7 The interaction mode between the main agent and the sub-agent in this task processing mode is as follows:

[0093] Step 1032, the main agent determines the sub-agent that executes the sub-task based on the task content of the sub-task and the description information of the sub-agent, and determines the task execution authority of the sub-agent as full execution authority.

[0094] Step 1042, the main agent sets the task identifier for the sub-task, and sends the task identifier, task content and full execution authority of the sub-task to the sub-agent that executes the sub-task.

[0095] In step 1052, the sub-agent determines, based on the task content of the sub-task and the full execution authority, that the task execution range is partial execution, executes at least part of the content of the sub-task, and obtains a task execution result.

[0096] In step 1062, the main agent processes the task execution result of the sub-agent to generate reply information, or determines a task execution tool based on the task execution result of the sub-agent and the target task, and generates reply information based on the task execution result of the sub-agent and the execution result of the task execution tool.

[0097] For example, in the "flash shopping merchant service scenario", when the user inputs the information "how is my recent rating", the main agent can determine, according to the user context information, that the task content of the target task is "recent rating query and evaluation". Then, the main agent determines that a certain sub-agent can completely solve the problem based on the description information of the sub-agent, and assigns the "full execution authority" to the sub-agent, and then sends the task identifier, task content and task execution authority of the task to the sub-agent. The sub-agent analyzes that it cannot directly answer the current question, but can provide part of the answer, and at the same time lacks the necessary time range information, and then sends the time confirmation information to the main agent, the main agent calls the tool to request the time range information from the client, and sends the requested time range information "20XX0801-20XX0831" to the sub-agent, the sub-agent performs rating query based on the above time range information, and then generates "rating query result", and sends "rating query result" to the main agent. After receiving the above information, the main agent can further evaluate the user's rating based on the "rating query result", and generate reply information to "how is the recent rating", and finally call the tool to send the generated reply information to the client, so that the client displays the rating card and the reply merchant's text through the preset component, so that the merchant can understand the problem in time, so as to obtain the corresponding solution.

[0098] In the above embodiment, when the sub-agent determines that it cannot completely solve the current problem after re-evaluation for the "full execution authority" assigned by the main agent, it will also feedback a task execution range to the main agent and continue to execute the task within the range. When the main agent receives the "partial execution" task execution range feedback by the sub-agent, it can assign part of the problem solving authority to the sub-agent, and then generate reply content on the processing result of the sub-agent, or call other tools based on the processing result of the sub-agent to obtain the final reply information. Through the above method, the processing results of multiple agents can be integrated, the calculation amount of the main agent can be saved, and the professionalism of task execution can be ensured.

[0099] Implementation three: full execution authority but not execution. As shown in the figure, in this task processing mode, the interaction mode between the main agent and the sub-agent is as follows: Figure 8

[0100] Step 1033, the main agent determines the sub-agent that executes the sub-task based on the task content of the sub-task and the description information of the sub-agent, and determines the task execution authority of the sub-agent as full execution authority.

[0101] Step 1043, the main agent sets a task identifier for the sub-task, and sends the task identifier, task content and full execution authority of the sub-task to the sub-agent that executes the sub-task.

[0102] Step 1053, the sub-agent generates a refusal to execute information as the task execution result when determining that the task execution range is not execution based on the task content of the sub-task and the full execution authority.

[0103] Step 1063, the main agent determines the task execution tool based on the target task, and generates a reply information based on the execution result of the task execution tool.

[0104] For example, in the "flash shopping merchant service scene", when the user inputs the information "what is my recent store rating", the main agent determines the task content of the target task as "what is the recent store rating" according to the user context information. Then, the main agent determines that a certain sub-agent can completely solve the problem based on the description information of the sub-agent, and gives the sub-agent "full execution authority", and then sends the task identifier, task content and task execution authority of the task to the sub-agent. After analysis, the sub-agent thinks that it cannot directly answer the current question, nor can it provide a partial answer, and then feeds back the refusal to execute information as the task execution result to the main agent. The main agent selects to call other tools to solve the current task based on the refusal to execute information fed back by the sub-agent. Then, the main agent selects to call the store rating query tool to query the current store rating based on the task content of the target task, and generates a reply information to "what is the recent store rating" according to the query result, and finally calls the tool to send the generated reply information to the client, so that the client displays the reply merchant's text through the preset component, so that the merchant can understand the question that needs to be consulted in time, so as to obtain the corresponding solution.

[0105] ​In the above embodiment, when the sub-agent considers that it cannot solve the current problem or provide a partial answer after re-evaluation for the "full execution authority" given to the main agent, the sub-agent can feed back the execution refusal information to the main agent. After receiving the "non-execution" of the task execution range fed back by the sub-agent, the main agent can reselect other tools (at least one tool can be called according to the execution order) based on the task content and obtain the final reply information. Through the above manner, the sub-agent can accurately evaluate its task execution capability, avoid the illusion problem of the sub-agent, and ensure the professionalism of task execution.

[0106] Embodiment four: partial execution authority and partial execution. As shown in the following table, in this task processing manner, the interaction manner between the main agent and the sub-agent is as follows: Figure 9

[0107] Step 1034, the main agent determines the sub-agent that executes the sub-task based on the task content of the sub-task and the description information of the sub-agent, and determines the task execution authority of the sub-agent as a partial execution authority.

[0108] Step 1044, the main agent sets a task identifier for the sub-task, and sends the task identifier, task content and partial execution authority of the sub-task to the sub-agent that executes the sub-task.

[0109] Step 1054, when the sub-agent determines that the task execution range is partial execution based on the task content of the sub-task and the partial execution authority, the sub-agent executes at least part of the content of the sub-task and obtains the task execution result.

[0110] Step 1064, the main agent processes the task execution result of the sub-agent to generate the reply information, or determines the task execution tool based on the task execution result of the sub-agent and the target task, and generates the reply information based on the task execution result of the sub-agent and the execution result of the task execution tool.

[0111] ​For example, in the "flash shopping merchant service scenario", when the user inputs the information "why did my recent evaluation rate decrease", the main agent can decompose the target task into two sub-tasks "how is the recent evaluation rate" and "reasons for the decrease in the evaluation rate" according to the user context information, and assign the sub-task "how is the recent evaluation rate" to a sub-agent and give the sub-agent "partial execution authority". Subsequently, the main agent can send the task identifier, task content and task execution authority of the sub-task to the sub-agent. After analysis, the sub-agent believes that it cannot directly answer the current question, but can provide part of the answer, and at the same time lacks the necessary time range information, then sends the time confirmation information to the main agent, the main agent calls the tool to request the time range information from the client, and sends the requested time range information "20XX0801-20XX0831" to the sub-agent, the sub-agent queries the evaluation rate based on the above time range information, and generates "the query result of the evaluation rate". Subsequently, the sub-agent sends the "query result of the evaluation rate" to the main agent, and the main agent calls the knowledge query retrieval tool to query "reasons for the decrease in the evaluation rate" after receiving the above information, and obtains the knowledge retrieval result of "reasons for the decrease in the evaluation rate". Finally, according to the "query result of the evaluation rate" and the knowledge retrieval result of "reasons for the decrease in the evaluation rate", the reply information to "why did the recent evaluation rate decrease" is generated, and the tool is called to send the generated reply information to the client, so that the client displays the evaluation rate card and the reply of the merchant through the preset component, so that the merchant can understand the problem to be consulted in time, so as to obtain the corresponding solution.

[0112] In the above embodiment, for the "partial execution authority" given by the main agent, when the sub-agent believes that it can provide part of the answer after re-evaluation, it will feedback a task execution range to the main agent, and continue to execute the task within the range. When the main agent receives the "partial execution" task execution range feedback by the sub-agent, it can give the sub-agent part of the authority to solve the problem, and then generate reply content on the processing result of the sub-agent, or call other tools based on the processing result of the sub-agent to obtain the final reply information. Through the above method, the processing results of multiple agents can be integrated, the calculation amount of the main agent can be saved, and the professionalism of task execution can be ensured.

[0113] Embodiment five: partial execution authority but not execution. As shown in Figure 10 The interaction mode between the main agent and the sub-agent in this task processing mode is as follows:

[0114] Step 1035, the main agent determines the sub-agent that executes the sub-task based on the task content of the sub-task and the description information of the sub-agent, and determines the task execution authority of the sub-agent as partial execution authority.

[0115] Step 1045, the main intelligent agent sets a task identifier for the subtask, and sends the task identifier, task content and partial execution permission of the subtask to the sub-intelligent agent executing the subtask.

[0116] Step 1055, the sub-intelligent agent determines that the task execution range is not execution based on the task content and partial execution permission of the subtask, and generates a refusal to execute information as the task execution result.

[0117] Step 1065, the main intelligent agent determines the task execution tool based on the target task, and generates a reply information based on the execution result of the task execution tool.

[0118] For example, in the "flash shopping merchant service scene", when the user inputs the information "how to improve the delivery speed of the store", the main intelligent agent will split the target task into two subtasks "what is the delivery time of the store" and "how to improve the delivery speed" according to the user context information, and assign the subtask "what is the delivery time of the store" to a sub-intelligent agent and give the sub-intelligent agent "partial execution permission". Subsequently, the main intelligent agent sends the task identifier, task content and task execution permission of the subtask to the sub-intelligent agent. After analysis, the sub-intelligent agent believes that it cannot directly answer the current question, nor can it provide partial answers, and then feeds back the refusal to execute information as the task execution result to the main intelligent agent. Based on the refusal to execute information fed back by the sub-intelligent agent, the main intelligent agent selects to call other tools to solve the current task, the main intelligent agent first queries the "current delivery time of the store" through the delivery time query tool, then calls the knowledge query retrieval tool to query "how to improve the delivery speed", obtains the knowledge retrieval result of "how to improve the delivery speed", and finally generates the reply information of "how to improve the delivery speed of the store" according to the query result of "current delivery time of the store" and the knowledge retrieval result of "how to improve the delivery speed", and calls the tool to send the generated reply information to the client, so that the client displays the reply merchant script through the preset component, so that the merchant can understand the problem to be consulted in time, so as to obtain the corresponding solution.

[0119] In the above embodiment, for the "partial execution permission" given by the main intelligent agent, when the sub-intelligent agent believes that it cannot solve the current problem after re-evaluation, nor can it provide partial answers, it can feed back the refusal to execute information to the main intelligent agent. After receiving the "non-execution" task execution range fed back by the sub-intelligent agent, the main intelligent agent can reselect to call other tools (at least one tool can be called according to the execution order) based on the task content, and obtain the final reply information. Through the above-mentioned manner, the sub-intelligent agent can accurately evaluate its task execution ability, avoid the illusion problem of the sub-intelligent agent, and ensure the professionalism of task execution.

[0120] In one embodiment, the information processing method further comprises the following step: storing the updated information in the memory system in response to an information update request, wherein the information covered by the information update request comprises at least one of the user context information, the task parameter, the reply information, the action point adoption information, and the action point execution success message.

[0121] In the above embodiments, during the execution of the information processing method described in each of the above embodiments, the information updated by each of the executors on the task execution state can be automatically stored in the memory system. The information stored in the memory system is subject to the prior authorization of the user, and the stored information can include the user input information received by the client, the user context information obtained by the main agent, the task parameter requested by the main agent or the sub-agent to the client, the reply information generated by the main agent, the action point adoption information received by the client, and the action point execution success message sent by the server, etc. By updating the above information to the memory system in a timely manner, the demand information of the user can be updated in a timely manner and understood in more detail, so that the user can be served in a targeted manner, thereby improving the user experience.

[0122] In one embodiment, as shown in Figure 11 The memory system comprises a session data layer, the session data layer comprises at least one target data layer, the target data layer comprises at least one task data layer, the task data layer comprises at least one model data layer and at least one behavior data layer, the user context information is stored in the task data layer, the task parameter is stored in one model data layer of the task data layer, and the reply information, the action point adoption information, and the action point execution success message are stored in one behavior data layer of the task data layer.

[0123] In the above embodiments, by storing different information in different data layers of the memory system in a hierarchical manner, the data can be classified and arranged, thereby improving the efficiency and accuracy of data query. For example, in the current round of dialogue, the sub-agent requests the time range information of information query from the client through the main agent in the process of task execution. Then, in the subsequent dialogue process, if the user again proposes an information query request, the main agent or the sub-agent can directly query the time range information obtained in the previous round in the corresponding data layer of the memory system, and directly query the information based on the time range information. In this way, on the one hand, the user experience can be improved by avoiding the user from repeatedly providing the same information, and on the other hand, the data query efficiency and the problem solving efficiency can be improved.

[0124] The technical solutions of the present application can be applied to transactions, delivery services, etc. of instant e-commerce platforms, such as Taobao flash shopping, Taofxian, Eleme takeout, and retail, etc.

[0125] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties. In addition, the labels corresponding to the steps in the above embodiments only play an identifying role and do not limit the execution order of the steps. The execution order of the steps in each embodiment can be set according to the actual situation.

[0126] Further, as Figures 1 to 11 The specific implementation of the method, the embodiments of the present application provide an information processing device, as shown in the figure, the device includes a main agent and at least one sub-agent, wherein: Figure 12 The main agent 41 can be used to obtain user context information, and perform intent recognition on the user context information to obtain a target task to be executed, wherein the target task includes at least one subtask;

[0127] The main agent 41 can also be used to determine a sub-agent for executing the subtask and a task execution authority of the sub-agent based on the task content of the subtask and the description information of the sub-agent;

[0128] The main agent 41 can also be used to set a task identifier for the subtask, and send the task identifier, task content and task execution authority of the subtask to the sub-agent for executing the subtask;

[0129] The sub-agent 42 can be used to determine a task execution range based on the task content of the subtask and the task execution authority, and execute the subtask in the task execution range to obtain a task execution result;

[0130] The main agent 41 can also be used to generate a reply information based on the task execution range and the task execution result of the sub-agent, and send the reply information to the client.

[0131]

[0132] ​In a specific application scenario, the task execution authority includes full execution authority and partial execution authority; the sub-agent 42 can be specifically used for, under the full execution authority, determining, based on the task content, that the task execution range is full execution, executing full content of the sub-task to obtain a task execution result; and / or, under the full execution authority, determining, based on the task content, that the task execution range is partial execution, executing at least part of the content of the sub-task to obtain a task execution result; and / or, under the full execution authority, determining, based on the task content, that the task execution range is no execution, generating a refusal to execute information as a task execution result; and / or, under the partial execution authority, determining, based on the task content, that the task execution range is partial execution, executing at least part of the content of the sub-task to obtain a task execution result; and / or, under the partial execution authority, determining, based on the task content, that the task execution range is no execution, generating a refusal to execute information as a task execution result.

[0133] In a specific application scenario, the apparatus further includes a parameter collection tool 43, and the sub-agent 42 is further used for, when executing the sub-task, if it is determined that there is a missing task parameter, sending parameter information of the missing task parameter to the master agent; the master agent 41 is further used for calling the parameter collection tool 43 based on the parameter information, and acquiring the missing task parameter through the parameter collection tool 43, and sending the task parameter to the sub-agent; the sub-agent 42 is further used for executing the sub-task based on the task parameter to obtain a task execution result.

[0134] In a specific application scenario, the apparatus further includes a task execution tool 44, and the master agent 41 is further used for, when the task execution range of the sub-agent is full execution, generating a reply information based on the task execution result of the sub-agent; and / or, when the task execution range of the sub-agent is partial execution, processing the task execution result of the sub-agent to generate a reply information; and / or, when the task execution range of the sub-agent is partial execution, determining a task execution tool 44 based on the task execution result of the sub-agent and the target task, and generating a reply information based on the task execution result of the sub-agent and an execution result of the task execution tool 44; and / or, when the task execution range of the sub-agent is no execution, determining a task execution tool 44 based on the target task, and generating a reply information based on an execution result of the task execution tool 44.

[0135] In a specific application scenario, the apparatus further comprises an information display tool 45, and the main agent 41 is further configured to invoke at least one information display tool 45 based on the reply information, and send the reply information to the client through the information display tool 45, wherein the information display tool comprises at least one of a text display tool, a text flow display tool, and an action point display tool.

[0136] In a specific application scenario, the information processing apparatus further comprises an action point server 46, wherein when the action point display tool is included in the invoked information display tool, the action point server 46 is configured to, in response to receiving action point adoption information, perform action point processing based on the action point adoption information, and send an action point execution success message to the client.

[0137] In a specific application scenario, the information processing apparatus further comprises a memory system 47, wherein the information processing apparatus can store updated information in the memory system 47 in response to an information update request, wherein the information covered by the information update request comprises at least one of user context information, task parameters, reply information, action point adoption information, and an action point execution success message.

[0138] In a specific application scenario, the memory system 47 comprises a session data layer, the session data layer comprises at least one target data layer, the target data layer comprises at least one task data layer, the task data layer comprises at least one model data layer and at least one behavior data layer, wherein the user context information is stored in the task data layer, the task parameters are stored in a model data layer of the task data layer, and the reply information, the action point adoption information, and the action point execution success message are stored in a behavior data layer of the task data layer.

[0139] It should be noted that other corresponding descriptions of the functions of the information processing apparatus provided by the embodiments of the present application can be referred to the corresponding descriptions of the functions in the method of the present application, which will not be repeated here. Figures 1 to 11

[0140] Further, as a specific implementation of the method, the embodiments of the present application provide an information processing apparatus, as shown in the figure, the apparatus comprises: Figures 1 to 11 Figure 13

[0141] ​​​The information sending module 51 can be configured to, in response to receiving user input information, send the user input information to a server, so that the server acquires user context information corresponding to the user input information, and performs intent recognition on the user context information by a master intelligent agent to obtain a target task to be executed, wherein the target task includes at least one subtask; the master intelligent agent determines a sub-intelligent agent for executing the subtask and a task execution authority of the sub-intelligent agent based on task content of the subtask and description information of the sub-intelligent agent; the master intelligent agent sets a task identifier for the subtask, and sends the task identifier, the task content and the task execution authority of the subtask to the sub-intelligent agent for executing the subtask; the sub-intelligent agent determines a task execution range based on the task content of the subtask and the task execution authority, and executes the subtask in the task execution range to obtain a task execution result; and the master intelligent agent generates reply information based on the task execution range and the task execution result of the sub-intelligent agent.

[0142] The information display module 52 can be configured to receive and display the reply information.

[0143] In a specific application scenario, the information display module 52 can be specifically configured to display the reply information through at least one preset component, wherein the preset component includes at least one of a text and image display component, a text stream display component and an action point display component.

[0144] In a specific application scenario, when the action point display component is included in the displayed preset component, the information sending module 51 can be further configured to receive action point adoption information through the action point display component, and send the action point adoption information to the server; and the information display module 52 can be further configured to receive and display a message that an action point is executed successfully.

[0145] In a specific application scenario, the information display module 52 can be further configured to display a parameter collection component, wherein the parameter collection component displays parameter information to be collected; and the information sending module 51 can be further configured to receive a task parameter through the parameter collection component, and send the task parameter to the server.

[0146] It should be noted that other corresponding descriptions of the functions of the information processing device provided in the embodiments of the present application can be referred to the corresponding descriptions in the method, which will not be repeated here. Figures 1 to 11

[0147] ​The embodiments of the present application further provide a computer device, which can be a personal computer, a server, a network device, etc. The computer device comprises a bus, a processor, a memory and a communication interface, and can further comprise an input / output interface and a display device. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The database of the computer device is configured to store location information. The network interface of the computer device is configured to communicate with an external terminal through a network connection. The computer program is configured to be executed by the processor to implement the steps in the method embodiments.

[0148] Those skilled in the art can understand that the structure of the computer device described above is only part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. Specifically, the computer device can comprise more or fewer components, or combine certain components, or have a different arrangement of components.

[0149] In one embodiment, a computer readable storage medium is provided, which can be non-volatile or volatile, and has stored thereon a computer program. The computer program is configured to be executed by a processor to implement the steps in the method embodiments described above.

[0150] In one embodiment, a computer program product is provided, which comprises a computer program. The computer program is configured to be executed by a processor to implement the steps in the method embodiments described above.

[0151] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments of the methods can be included. Any reference to memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without being limited thereto.

[0152] Any combination of the technical features of the above embodiments can be made. In order to make the description simple, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combination of the technical features does not exist, it should be considered as the scope of the present application.

[0153] The above embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent of the present application. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are within the scope of protection of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. An information processing method characterized by comprising: The method comprises: obtaining user context information and performing intention recognition on the user context information by a master agent to obtain a target task to be executed, wherein the target task comprises at least one subtask; the master agent determines a subagent for executing the subtask and a task execution authority of the subagent based on the task content of the subtask and the description information of the subagent, wherein the task execution authority comprises full execution authority and partial execution authority; the master agent sets a task identifier for the subtask and sends the task identifier, task content and task execution authority of the subtask to the subagent executing the subtask; the subagent determines a task execution range based on the task content of the subtask and the task execution authority and executes the subtask in the task execution range to obtain a task execution result, wherein the task execution range of the subagent is equal to or lower than the task execution authority of the subagent, and the task execution range comprises full execution, partial execution and no execution; wherein, when executing the subtask, if it is determined that there is a missing task parameter, the subagent sends parameter information of the missing task parameter to the master agent; the master agent calls a parameter collection tool based on the parameter information and obtains the missing task parameter through the parameter collection tool and sends the task parameter to the subagent; the subagent executes the subtask based on the task parameter to obtain a task execution result; the master agent generates reply information based on the task execution range and the task execution result of the subagent and sends the reply information to the client, wherein the master agent determines the completion degree of the task execution result of the subagent for the target task based on the task execution range of the subagent, and directly outputs the task execution result of the subtask based on the completion degree of the target task, or re-integrates, supplements or calls other tools to execute the target task or supplement the execution content to obtain the reply information.

2. The information processing method according to claim 1, characterized by, The subagent determines a task execution range based on the task content of the subtask and the task execution authority and executes the subtask in the task execution range to obtain a task execution result, comprising: when the subagent determines that the task execution range is full execution based on the task content under the full execution authority, the subagent executes all contents of the subtask to obtain a task execution result; and / or, when the subagent determines that the task execution range is partial execution based on the task content under the full execution authority, the subagent executes at least part of the contents of the subtask to obtain a task execution result; and / or, when the subagent determines that the task execution range is no execution based on the task content under the full execution authority, the subagent generates a refusal to execute information as a task execution result; and / or, The sub-agent determines, based on the task content, that the task execution range is partial execution under the partial execution authority, and executes at least part of the content of the sub-task to obtain a task execution result; and / or The sub-agent determines, based on the task content, that the task execution range is no execution under the partial execution authority, and generates a refusal to execute information as a task execution result.

3. The information processing method according to claim 1 or 2, characterized by, The main agent generates reply information based on the task execution range and the task execution result of the sub-agent, including: When the task execution range of the sub-agent is full execution, the main agent generates reply information based on the task execution result of the sub-agent; and / or When the task execution range of the sub-agent is partial execution, the main agent processes the task execution result of the sub-agent to generate reply information; and / or When the task execution range of the sub-agent is partial execution, the main agent determines a task execution tool based on the task execution result of the sub-agent and the target task, and generates reply information based on the task execution result of the sub-agent and the execution result of the task execution tool; and / or When the task execution range of the sub-agent is no execution, the main agent determines a task execution tool based on the target task, and generates reply information based on the execution result of the task execution tool.

4. The information processing method according to claim 1, characterized by, The main agent sends the reply information to the client, including: The main agent calls at least one information display tool based on the reply information, and sends the reply information to the client through the information display tool, wherein the information display tool includes at least one of a text display tool, a text flow display tool, and an action point display tool.

5. The information processing method according to claim 4, characterized by, When the action point display tool is included in the called information display tool, the method further includes: In response to receiving the action point adoption information, processing the action point based on the action point adoption information, and sending a message of successful action point execution to the client.

6. The information processing method according to claim 1, characterized by, The method further includes: In response to an information update request, storing the updated information in a memory system, wherein the information covered by the information update request includes at least one of user context information, task parameters, reply information, action point adoption information, and a message of successful action point execution.

7. The information processing method according to claim 6, characterized by, The memory system includes a session data layer, the session data layer includes at least one target data layer, the target data layer includes at least one task data layer, the task data layer includes at least one model data layer and at least one behavior data layer, wherein the user context information is stored in the task data layer, the task parameters are stored in one model data layer of the task data layer, and the reply information, the action point adoption information, and the message of successful action point execution are stored in one behavior data layer of the task data layer.

8. An information processing method characterized by comprising: The method includes: In response to receiving user input information, sending the user input information to a server to enable the server to execute the information processing method of any one of claims 1 to 7 to generate reply information; receiving and displaying the reply information.

9. The information processing method according to claim 8, characterized by, The receiving and displaying the reply information comprises: displaying the reply information through at least one preset component, wherein the preset component comprises at least one of a graphic-text display component, a text flow display component and an action point display component.

10. The information processing method according to claim 9, characterized by, When the action point display component is included in the displayed preset component, the method further comprises: receiving action point adoption information through the action point display component, and sending the action point adoption information to the server; receiving and displaying a message of successful action point execution.

11. The information processing method according to claim 8, characterized by, The method further comprises: displaying a parameter collection component, wherein the parameter collection component displays parameter information of a task parameter; receiving a task parameter through the parameter collection component, and sending the task parameter to the server.

12. An information processing apparatus comprising: The device comprises: a main intelligent agent configured to acquire user context information, perform intent recognition on the user context information, and obtain a target task to be executed, wherein the target task comprises at least one subtask; the main intelligent agent is further configured to determine a sub intelligent agent for executing the subtask and a task execution authority of the sub intelligent agent based on task content of the subtask and description information of the sub intelligent agent, wherein the task execution authority comprises full execution authority and partial execution authority; the main intelligent agent is further configured to set a task identifier for the subtask, and send the task identifier, the task content and the task execution authority of the subtask to the sub intelligent agent for executing the subtask; a sub intelligent agent configured to determine a task execution range based on the task content of the subtask and the task execution authority, and execute the subtask in the task execution range to obtain a task execution result, wherein the task execution range of the sub intelligent agent is equal to or lower than the task execution authority of the sub intelligent agent, and the task execution range comprises full execution, partial execution and no execution; the sub intelligent agent is further configured to, when executing the subtask, determine whether there is a missing task parameter, and if so, send parameter information of the missing task parameter to the main intelligent agent; the main intelligent agent calls a parameter collection tool based on the parameter information, acquires the missing task parameter through the parameter collection tool, and sends the task parameter to the sub intelligent agent; the sub intelligent agent executes the subtask based on the task parameter to obtain a task execution result; the main intelligent agent is further configured to generate reply information based on the task execution range and the task execution result of the sub intelligent agent, and send the reply information to a client, wherein the main intelligent agent determines a completion degree of the target task based on the task execution range of the sub intelligent agent, determines the task execution result of the sub intelligent agent based on the completion degree of the target task, and directly outputs the task execution result of the sub intelligent agent, or re-integrates, supplements or calls other tools to execute the target task or supplement execution content to obtain the reply information.

13. An information processing apparatus comprising: The device comprises: The information sending module is configured to, in response to receiving user input information, send the user input information to a server, so that the server executes the information processing method according to any one of claims 1 to 7 to generate reply information. The information display module is configured to receive and display the reply information.

14. A storage medium having stored thereon a computer program, characterized in that The computer program, when executed by a processor, implements the method according to any one of claims 1 to 11.

15. A computer device comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, The processor, when executing the computer program, implements the method according to any one of claims 1 to 11.

Images