Man-machine interaction system and man-machine interaction method

Abstract

The invention discloses a man-machine interaction system and a man-machine interaction method.The man-machine interaction intelligent agent is used for receiving dialogue information input by a user and determining a user intention based on one or more rounds of first dialogue information input by the user; the intelligent control parameter agent is used for converting a user intention into one or more execution parameters; the process control agent is used for controlling the step execution agent to select one or more steps for sequential execution based on one or more execution parameters so as to complete a first process, and the first process is used for outputting a processing result about a user intention; the man-machine interaction intelligent body is further used for outputting a system response to the user, and the system response is determined based on the processing result; wherein the intelligent control parameter intelligent agent and / or the step execution intelligent agent are / is further used for interacting with the man-machine interaction intelligent agent for one time or multiple times in the operation process so as to obtain second dialogue information input by the user, and the second dialogue information represents information needed by the corresponding intelligent agent in the operation process.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and particularly relates to a human-computer interaction system and a human-computer interaction method. BACKGROUND

[0002] At present, in the execution process of a human-computer interaction business process, user supplementary information or user confirmation information often needs to be obtained through multiple user interactions. However, in the related art, a dialog task framework only supports user interaction before the execution of a business process, or the scheduling and arrangement of a business process and the calling of a tool need to be performed again after each user interaction, which affects the processing efficiency. SUMMARY

[0003] To solve the problems in the related art, the present application provides a human-computer interaction system and a human-computer interaction method.

[0004] The technical scheme of the present application is implemented as follows:

[0005] The present application provides a human-computer interaction system, which comprises a human-computer interaction intelligent agent, a smart control parameter intelligent agent, a process control intelligent agent, and a step execution intelligent agent, wherein

[0006] The human-computer interaction intelligent agent is configured to receive dialog information input by a user, and determine a user intention based on one or more rounds of first dialog information input by the user.

[0007] The smart control parameter intelligent agent is configured to convert the user intention into one or more execution parameters.

[0008] The process control intelligent agent is configured to control the step execution intelligent agent to select one or more steps to be executed in sequence based on the one or more execution parameters, so as to complete a first process, wherein the first process is configured to output a processing result related to the user intention.

[0009] The human-computer interaction intelligent agent is further configured to output a system response to the user, wherein the system response is determined based on the processing result.

[0010] The smart control parameter intelligent agent and / or the step execution intelligent agent are further configured to interact with the human-computer interaction intelligent agent one or more times during the running process, so as to obtain second dialog information input by the user, wherein the second dialog information represents information required by the corresponding intelligent agent during the running process.

[0011] In the above scheme, the human-computer interaction intelligent agent is configured to:

[0012] In a case where the user inputted conversation information belongs to a brand new conversation, the user inputted conversation information is determined as first conversation information, and the user intention is determined based on one or more rounds of the user inputted first conversation information; and / or,

[0013] In a case where the user inputted conversation information belongs to a reply to a conversation request initiated by a first intelligent agent, the user inputted conversation information is determined as second conversation information, and the second conversation information is fused with corresponding snapshot information and then sent to the first intelligent agent, so that the first intelligent agent continues to run based on the second information; wherein the first intelligent agent includes the intelligent control parameter agent or the step execution agent, and the snapshot information represents relevant processing information of the first intelligent agent about the user intention before initiating the conversation request.

[0014] In the above scheme, the human-computer interaction intelligent agent is used for:

[0015] receiving a first message sent by a first intelligent agent, the first message representing a conversation request, and the first intelligent agent including the intelligent control parameter agent or the step execution agent;

[0016] outputting the conversation request and obtaining second conversation information inputted by a user based on the conversation request;

[0017] fusing the second conversation information with corresponding snapshot information and then sending to the first intelligent agent, so that the first intelligent agent continues to run based on the second information, wherein the snapshot information represents relevant processing information of the first intelligent agent about the user intention before initiating the conversation request.

[0018] In the above scheme, the human-computer interaction intelligent agent is used for:

[0019] after receiving the first message, saving the snapshot information and source information corresponding to the conversation request, and the source information is used to indicate that the conversation request is initiated by the first intelligent agent;

[0020] after receiving the second conversation request, generating a second message, and the second message carries the second conversation information and corresponding snapshot information;

[0021] based on the source information, sending the second message to the first intelligent agent.

[0022] In the above scheme, the human-computer interaction intelligent agent is used for:

[0023] receiving a second message sent by the flow control intelligent agent, and the second message carries the processing result;

[0024] generating the system response based on the processing result.

[0025] outputting the system response to the user.

[0026] In the above solution, the intelligent control parameter agent is further configured to:

[0027] identify an intention scenario to which the user intention belongs; and / or,

[0028] perform parameter verification on the one or more execution parameters; and / or,

[0029] In a case where the human-computer interaction agent is unable to determine the user intention, determine whether to complete the first process by the process control agent.

[0030] In the above solution, the step execution agent is configured to:

[0031] select a model and / or a tool matched with the step to be executed based on the step to be executed;

[0032] invoke the model and / or the tool to execute the step to be executed.

[0033] Embodiments of the present application also provide a human-computer interaction method applied to a human-computer interaction system, the human-computer interaction system comprising a human-computer interaction agent, an intelligent control parameter agent, a process control agent and a step execution agent; the method comprising:

[0034] the human-computer interaction agent receiving user inputted dialogue information and determining a user intention based on one or more rounds of first dialogue information inputted by the user;

[0035] the intelligent control parameter agent converting the user intention into one or more execution parameters;

[0036] the process control agent controlling the step execution agent to select one or more steps to be executed in sequence based on the one or more execution parameters, so as to complete a first process, the first process being configured to output a processing result about the user intention;

[0037] the human-computer interaction agent outputting a system response to the user, the system response being determined based on the processing result;

[0038] wherein the intelligent control parameter agent and / or the step execution agent interacts with the human-computer interaction agent one or more times in a running process to obtain second dialogue information inputted by the user, the second dialogue information representing information required by the corresponding agent in the running process.

[0039] In the above solution, the method further comprises:

[0040] When the dialogue information input by the user is a completely new dialogue, the human-computer interaction agent determines the dialogue information input by the user as the first dialogue information, and determines the user's intention based on one or more rounds of the first dialogue information input by the user; and / or,

[0041] When the dialogue information input by the user is a response to a dialogue request initiated by the first intelligent agent, the human-computer interaction intelligent agent determines that the dialogue information input by the user is the second dialogue information, and sends the second dialogue information and the corresponding snapshot information to the first intelligent agent so that the first intelligent agent can continue to operate based on the second information; wherein, the first intelligent agent includes the intelligent control parameter intelligent agent or the step execution intelligent agent, and the snapshot information represents the relevant processing information of the first intelligent agent regarding the user's intent before initiating the dialogue request.

[0042] The method in the above scheme further includes:

[0043] The human-computer interaction intelligent agent receives a first message sent by a first intelligent agent, the first message being a dialogue request, and the first intelligent agent including the intelligent control parameter intelligent agent or the step execution intelligent agent;

[0044] The human-computer interaction agent outputs the dialogue request and obtains the second dialogue information input by the user based on the dialogue request;

[0045] The human-computer interaction agent merges the second dialogue information with the corresponding snapshot information and sends it to the first agent, so that the first agent can continue to operate based on the second information. The snapshot information represents the relevant processing information of the first agent regarding the user's intent before initiating the dialogue request.

[0046] The method in the above scheme further includes:

[0047] After receiving the first message, the human-computer interaction agent saves the snapshot information and source information corresponding to the dialogue request. The source information is used to indicate that the dialogue request was initiated by the first agent.

[0048] Correspondingly, the human-computer interaction agent fuses the second dialogue information with the corresponding snapshot information and sends it to the first agent, including:

[0049] Upon receiving the second dialogue request, the human-computer interaction agent generates a second message, which carries the second dialogue information and the corresponding snapshot information.

[0050] Based on the source information, the human-computer interaction agent sends the second message to the first agent.

[0051] In the above scheme, the human-computer interaction intelligent agent outputs a system response to the user, including:

[0052] The human-computer interaction agent receives a second message sent by the process control agent, the second message carrying the processing result;

[0053] The human-computer interaction agent generates the system response based on the processing result;

[0054] The human-computer interaction agent outputs the system response to the user.

[0055] The method in the above scheme further includes:

[0056] The intelligent control parameter agent identifies the intent scenario to which the user's intent belongs; and / or,

[0057] The intelligent control parameter agent performs parameter verification on the one or more execution parameters; and / or,

[0058] If the human-computer interaction agent cannot determine the user's intention, the intelligent control parameter agent determines whether the process control agent should complete the first process.

[0059] The method in the above scheme further includes:

[0060] The step-executing agent selects a model and / or tool that matches the steps to be executed, based on the steps to be executed.

[0061] The step described above executes the agent by invoking the model and / or tool to perform the required steps.

[0062] In the human-computer interaction system and method provided in this application, the human-computer interaction intelligent agent receives dialogue information input by the user and determines the user's intent based on one or more rounds of first dialogue information input by the user; the intelligent control parameter intelligent agent converts the user's intent into one or more execution parameters; the process control intelligent agent, based on one or more execution parameters, controls the step execution intelligent agent to select one or more steps to execute sequentially, so as to output the processing result regarding the user's intent; then, the human-computer interaction intelligent agent outputs the system response determined based on the processing result to the user. In the above scheme, multiple intelligent agents in the human-computer interaction system cooperate with each other, enabling flexible arrangement of processes. Furthermore, during the operation of the human-computer interaction system, when the user needs to supplement or confirm relevant information, the intelligent control parameter intelligent agent and / or the step execution intelligent agent also support one or more interactions with the human-computer interaction intelligent agent during the operation of the human-computer interaction system, thereby obtaining the user supplementary information or user confirmation information required for process execution without re-arranging or re-executing the process, improving the execution efficiency of the process. Attached Figure Description

[0063] Figure 1 This is a schematic diagram of the interaction between four intelligent agents in the human-computer interaction system of this application embodiment;

[0064] Figure 2 This is a schematic diagram of the human-computer interaction intelligent agent structure according to an embodiment of this application;

[0065] Figure 3 This is a schematic diagram of the operation flow of the human-computer interaction intelligent agent in the embodiments of this application;

[0066] Figure 4 This is a schematic diagram of the message monitoring process of the human-computer interaction intelligent agent in an embodiment of this application;

[0067] Figure 5 This is a schematic diagram of the intelligent control parameter agent structure in an embodiment of this application;

[0068] Figure 6 This is a schematic diagram of the intelligent agent structure executing the steps in the embodiments of this application;

[0069] Figure 7 This is a schematic diagram illustrating the implementation process of the human-computer interaction method in an embodiment of this application;

[0070] Figure 8 This is a schematic diagram illustrating the overall implementation process of the human-computer interaction method in an embodiment of this application. Detailed Implementation

[0071] In related technologies, dialogue-based task frameworks for realizing human-computer interaction are mainly divided into two categories:

[0072] The first type is a basic sequential process processing framework, such as the Langchain framework. After receiving a user request, this type of framework calls a large model to classify the user request intent, maps the corresponding process according to the classification result, confirms the necessary information required for the corresponding process through multiple interactions with the user, and then calls a fixed process to complete the processing in one go and returns a system response to the user.

[0073] The first type of framework can only enable user interaction before the execution of the dialogue task process. It cannot interact with the user during the execution of the process. Therefore, in complex scenarios, if information confirmation needs to be completed through user interaction in different execution stages of the process, the first type of framework cannot achieve this function. Furthermore, if the process changes later, the relevant code needs to be modified, which is not flexible enough.

[0074] The second type of framework is implemented using process orchestration agents. This approach calls a large model to identify the intent of user requests and completes the necessary information through user interaction. Then, agents based on the large model are used to schedule and orchestrate processes and invoke tools. Users can interact with the user through a specific agent among multiple agents.

[0075] The second type of framework can only interact with the user in a specific agent, and the entire process must be re-executed after each interaction with the user. For scenarios where repetitive processing is not required, this adds unnecessary steps and reduces processing efficiency.

[0076] Based on this, in this embodiment of the application, the human-computer interaction system includes: a human-computer interaction intelligent agent, an intelligent control parameter agent, a process control intelligent agent, and a step execution intelligent agent; wherein, the human-computer interaction intelligent agent receives dialogue information input by the user and determines the user's intent based on one or more rounds of first dialogue information input by the user; the intelligent control parameter agent converts the user's intent into one or more execution parameters; the process control intelligent agent, based on one or more execution parameters, controls the step execution intelligent agent to select one or more steps to execute sequentially, so as to output the processing result regarding the user's intent, and then, the human-computer interaction intelligent agent outputs the system response determined based on the processing result to the user. In the above scheme, multiple intelligent agents in the human-computer interaction system cooperate with each other, which can realize the flexible arrangement of related processes. Furthermore, during the operation of the human-computer interaction system, when the user needs to supplement or confirm relevant information, the intelligent control parameter agent and / or the step execution intelligent agent also support one or more interactions with the human-computer interaction intelligent agent during the operation of the human-computer interaction system, thereby obtaining the user supplementary information or user confirmation information required for process execution without re-arranging or re-executing the process, thus improving the execution efficiency of the process.

[0077] The present application will now be described in further detail with reference to the accompanying drawings and embodiments.

[0078] In the embodiments of this application, the conversational task framework of the human-computer interaction system is implemented based on the following four agents: human-computer interaction agent (Interact), intelligent control parameter agent (IntelliMatch), process control agent (Maintainer), and step execution agent (Director).

[0079] Among them, the human-computer interaction intelligent agent is the intelligent agent that enables the human-computer interaction system to interact with the user. It is responsible for receiving the dialogue information input by the user and outputting the system response corresponding to the dialogue information, that is, providing feedback to the user on the system response corresponding to the dialogue information.

[0080] The intelligent parameter agent can track the context of user-input dialogue information, determine the user's intent in the dialogue information, and identify and correct the execution parameters corresponding to the user's intent. Execution parameters ensure that the process of handling dialogue tasks is executed based on the user's intent and may include parameter values ​​and application programming interfaces (APIs) required during the process. These functions of the intelligent parameter agent ensure that the human-computer interaction system possesses deep understanding and responsiveness. Furthermore, when the intelligent parameter agent needs the user to supplement or modify parameters during operation, it can obtain the supplemented or modified parameters based on the interaction between the human-computer interaction agent and the user.

[0081] The process control agent selects one or more steps to be executed sequentially based on user intent in order to control the execution of the process.

[0082] The step execution agent selects the steps based on the process control agent, calls relevant tools and / or models to complete the execution of the steps, and when the step execution agent needs the user to supplement or modify parameters during the operation, the step execution agent can obtain the supplemented or modified parameters based on the interaction between the human-computer interaction agent and the user.

[0083] The four intelligent agents described above operate independently but collaborate during the execution of the business process. The following combines... Figure 1 The diagram illustrates the interaction of four agents in the human-computer interaction system, where each arrow jump represents an action execution.

[0084] Reference Figure 1In a dialogue task, the human-computer interaction agent receives a question input by the user and may engage in one or more rounds of dialogue with the user based on this question to obtain more dialogue information. Based on this, the complete question from the user is obtained, thus determining the user's intent in the dialogue task, which is then sent to the intelligent control parameter agent. The intelligent control parameter agent converts the user intent into one or more execution parameters. During the conversion of execution parameters, if additional information or confirmation from the user is required, the intelligent control parameter agent requests parameter completion or modification through communication with the human-computer interaction agent. The human-computer interaction agent obtains the additional information or confirmation from the user through interaction and returns the parameter modification confirmation result to the intelligent control parameter agent. Based on one or more execution parameters, the flow control agent selects one or more step execution agents to run sequentially, sending the relevant step descriptions to the corresponding step execution agents. After completing the relevant steps, the corresponding step execution agents return the next step number to the flow control agent, enabling the flow control agent to continue instructing the next step execution agent to run. During the execution of each step by the execution agent, if user confirmation is required, the execution agent communicates with the human-computer interaction agent. The human-computer interaction agent obtains the user's confirmation through interaction and sends it back to the execution agent. After all steps are completed, the flow control agent terminates the process and sends the processing result regarding the user's intent to the human-computer interaction agent, which then outputs a system response determined based on the processing result.

[0085] It can be seen that during the execution of the process, the intelligent control parameter agent, the process control agent, and the step execution agent all interact with the user through the human-computer interaction agent. The four agents cooperate with each other and can flexibly arrange the process according to the actual human-computer interaction scenario and needs.

[0086] Based on the above description, the human-computer interaction system provided in this application embodiment is as follows:

[0087] Human-computer interaction intelligent agents are used to receive dialogue information input by users and determine user intentions based on one or more rounds of first dialogue information input by users;

[0088] Intelligent control parameter agent, used to convert user intent into one or more execution parameters;

[0089] A process control agent is used to control the step execution agent to select one or more steps to execute sequentially based on one or more execution parameters to complete a first process, which is used to output the processing result about the user's intent;

[0090] The human-computer interaction intelligent agent is also used to output system responses to users, which are determined based on the processing results;

[0091] Among them, the intelligent control parameter agent and / or the step execution agent are also used to interact with the human-computer interaction agent one or more times during operation to obtain the second dialogue information input by the user. The second dialogue information represents the information required by the corresponding agent during operation.

[0092] In the aforementioned interaction process, the human-computer interaction agent can be understood as an agent that combines interactive information processing and monitoring functions. Firstly, regarding information processing, after receiving new dialogue information input by the user, the human-computer interaction agent transmits this information to the agent that triggered the dialogue round. For example, if the dialogue round is initiated by the human-computer interaction agent to obtain the user's complete question, then this dialogue information is used by the human-computer interaction agent to generate the user's complete question. Alternatively, if the dialogue round is initiated by the intelligent control parameter agent, then the human-computer interaction agent sends the dialogue information to the intelligent control parameter agent so that the intelligent control parameter agent can determine the corresponding execution parameters. Or, if the dialogue round is initiated by the step execution agent, then the human-computer interaction agent sends the dialogue information to the step execution agent so that the step execution agent can complete the execution of the corresponding step.

[0093] Based on this, in one embodiment, the human-computer interaction intelligent agent is used for:

[0094] If the user-input dialogue information pertains to a completely new dialogue, the user-input dialogue information is determined as the first dialogue information, and the user's intent is determined based on one or more rounds of first dialogue information input by the user; and / or,

[0095] If the dialogue information entered by the user is a response to a dialogue request initiated by the first intelligent agent, the dialogue information entered by the user is determined to be the second dialogue information. The second dialogue information is then fused with the corresponding snapshot information and sent to the first intelligent agent so that the first intelligent agent can continue to operate based on the second information. The first intelligent agent includes an intelligent control parameter agent or a step execution agent. The snapshot information represents the relevant processing information of the first intelligent agent regarding the user's intent before initiating the dialogue request.

[0096] Here, the first round or multiple rounds of dialogue information are used by the human-computer interaction agent to determine the user's complete question; this can also be understood as the dialogue information from the first interaction in a dialogue task. The second round of dialogue information is used by the intelligent control parameter agent to determine the business process execution parameters corresponding to the dialogue task, or by the step execution agent to execute the corresponding steps.

[0097] Here, when the intelligent control parameter agent or the step execution agent needs the user to supplement information and / or confirm information during operation, the current process will be interrupted. The human-computer interaction agent will send the second dialogue information input by the user and the snapshot information of the current dialogue task to the intelligent control parameter agent or the step execution agent. This enables the corresponding agent to quickly resume the previously interrupted process based on the snapshot information and continue to run the relevant process based on the second dialogue information.

[0098] In addition, for the monitoring function, the human-computer interaction intelligent agent can monitor the messages sent by other intelligent agents and confirm whether the message is used for secondary interaction with the user or to indicate the end of the process.

[0099] Based on this, in one embodiment, when a human-computer interaction agent monitors messages used for secondary interaction with a user, the human-computer interaction agent is used to:

[0100] Receive a first message sent by a first intelligent agent, the first message being a dialogue request, the first intelligent agent including an intelligent control parameter agent or a step execution agent;

[0101] Output the dialogue request and obtain the second dialogue information input by the user based on the dialogue request;

[0102] The second dialogue information is fused with the corresponding snapshot information and sent to the first intelligent agent so that the first intelligent agent can continue to operate based on the second information. The snapshot information represents the relevant processing information of the first intelligent agent about the user's intent before initiating the dialogue request.

[0103] In one embodiment, the human-computer interaction agent is used for:

[0104] After receiving the first message, the snapshot information and source information corresponding to the dialogue request are saved. The source information is used to indicate that the dialogue request was initiated by the first intelligent agent.

[0105] Upon receiving the second dialogue request, a second message is generated, which carries the second dialogue information and the corresponding snapshot information.

[0106] Based on the source information, the second message is sent to the first intelligent agent.

[0107] Here, when the intelligent control parameter agent or the step execution agent requires supplementary information and / or user confirmation during operation, the human-computer interaction agent, upon receiving the dialogue request from the intelligent control parameter agent or the step execution agent, saves corresponding snapshot information. This allows the intelligent control parameter agent or the step execution agent to quickly resume the previously interrupted process based on the snapshot information after obtaining the user input corresponding to the dialogue request. Furthermore, after receiving the dialogue request from the intelligent control parameter agent or the step execution agent, the human-computer interaction agent also records relevant information about the agent that initiated the dialogue request. Thus, after obtaining the user input corresponding to the dialogue request, the human-computer interaction agent can determine the agent that initiated the dialogue request based on the recorded relevant information, and then send the second dialogue information input by the user along with the snapshot information of the current dialogue task to the agent that initiated the dialogue request.

[0108] In one embodiment, when a human-computer interaction agent detects a message indicating the end of a process, the human-computer interaction agent is used to:

[0109] Receive a second message sent by the process control agent, which carries the processing result;

[0110] Generate a system response based on the processing results;

[0111] Output system response.

[0112] Based on the above-mentioned embodiments of human-computer interaction intelligent agents, in practical applications, such as Figure 2 As shown, the human-computer interaction intelligent agent may include a two-way communication module and a scene-driven module.

[0113] The two-way communication module includes two parts: request and / or response processing and interactive confirmation. It is used to receive dialogue information input by the user, determine whether the dialogue information input by the user belongs to the first interaction of the dialogue task or a supplementary dialogue, and return a system response.

[0114] The scenario-driven module comprises two parts: repetitive interaction processing and semantic translation. Based on the dialogue information obtained from the bidirectional communication module, the scenario-driven module triggers subsequent action strategies. Specifically, the repetitive interaction processing part integrates secondary interaction information and relevant historical dialogue information, dynamically matching subsequent processing flows for non-first-time dialogues; the semantic translation part comprehensively judges the true user intent based on known contextual information, removes redundant information, and uses it as input for post-processing logic.

[0115] In addition, the main functional logic of human-computer interaction intelligent agents is divided into two parts: user input processing and intelligent agent interaction information processing.

[0116] Among them, such as Figure 3As shown, after the human-computer interaction agent obtains the dialogue information input by the user, the user input processing part identifies the agent that gave the processing result when the system response was output last time through the corresponding dialogue history, and determines whether the dialogue information input this time is the user's reply to the previous dialogue by combining the previous system response.

[0117] If the dialogue information input this time is the user's response to the previous dialogue, then the snapshot information of the previous dialogue is restored. The dialogue information input this time and the historical snapshot information are integrated and sent to the agent that gave the processing result when the system responded last time. The agent can restore the previously interrupted process based on the snapshot information and continue to execute the process based on the dialogue information input this time.

[0118] If the dialogue information input this time is not the user's response to the previous dialogue, it will be processed according to the process of a new dialogue task. The model will be invoked to identify the user's intent and sent to the intelligent control parameter agent to continue to execute the relevant process.

[0119] Based on the above processing logic, the shortest execution flow can be executed when performing dialogue tasks, thereby improving processing efficiency.

[0120] In practical applications, the human-computer interaction intelligent agent continuously monitors messages sent by other intelligent agents. These messages are divided into two main categories: user interaction messages and process completion messages.

[0121] like Figure 4 As shown, if the human-computer interaction agent receives a user interaction message, it obtains the snapshot information of the current process state extracted by the agent that sent the user interaction message, saves the snapshot information and source information, and constructs and sends the interaction information with the user according to the interface agreed with the user.

[0122] If the human-computer interaction agent receives a process completion message, it identifies the process processing result. If the processing result is abnormal, it supplements the corresponding abnormal message, and then constructs and sends the process processing completion information to each user.

[0123] The above processing logic allows for flexible interaction between agents at any stage of the business process, based on the actual business needs of the dialogue task, thereby improving the efficiency of the process.

[0124] In one embodiment, the intelligent control parameter agent is further used for:

[0125] Identify the intent context to which the user's intent belongs; and / or,

[0126] Perform parameter validation on one or more execution parameters; and / or,

[0127] When the human-computer interaction agent cannot determine the user's intent, determine whether the process control agent should complete the first process.

[0128] Based on the above-mentioned embodiments of intelligent control parameter agents, in practical applications, such as Figure 5 As shown, the intelligent control parameter agent may include an intent scene recognition module, an intent parameter recognition module, a parameter verification and validation module, and an intent transfer module.

[0129] The intelligent control parameter agent identifies the intent scene to which the user's intent belongs based on the intent scene recognition module. This module matches the user's intent to the corresponding scene based on the context information of the tracked dialogue. The intelligent control parameter agent then converts the user's intent into specific executable parameters based on the intent parameter recognition module. Through a parameter verification and validation module, it performs multi-dimensional and multi-condition parameter verification based on predefined parameter configuration rules. When the human-computer interaction agent cannot determine the user's intent, the intelligent control parameter agent uses an intent transfer module to determine whether the process control agent should complete the first process. Here, the intent transfer module plays a post-processing role within the intelligent control parameter agent. When the human-computer interaction system cannot fully understand the user information, the intent transfer module makes a decision, determining whether the process needs to be transferred to human intervention in the current scenario.

[0130] Furthermore, the intelligent control parameter agent also employs a follow-up questioning strategy, enabling multiple interactions between the human-computer interaction system and the user. The dialogue with the user is then relayed back through the intelligent human-computer interaction agent. When the identified scenario or parameters are abnormal, or when interaction with the user is needed to confirm parameter modifications, the intelligent human-computer interaction agent can engage in multiple interactions to supplement or confirm relevant information.

[0131] In one embodiment, the step-execution agent is used to:

[0132] Based on the steps that need to be performed, select a model and / or tool that matches the steps that need to be performed;

[0133] Invoke the selected model and / or tool to perform the required steps.

[0134] Based on the above steps, relevant embodiments of the intelligent agent are executed. In practical applications, such as... Figure 6As shown, the process control agent selects different process control prompts based on the user's intent scenario and parameters, and identifies the detailed description of each step. The step-jumping agent, based on the selected step, further identifies the relevant information of the models and / or tools required for step execution, completes the step execution, and returns the result of each step to the process control agent, which then selects the next step. During step selection and execution, if secondary parameter selection and confirmation with the user are required, the process control agent can communicate with the human-computer interaction agent, which will then save relevant state information and confirm the interaction with the user.

[0135] Based on the human-computer interaction system provided in the above embodiments, this application also provides a human-computer interaction method, applied to the aforementioned human-computer interaction system. (Refer to...) Figure 7 The method includes:

[0136] Step 701: The human-computer interaction intelligent agent receives dialogue information input by the user and determines the user's intent based on one or more rounds of first dialogue information input by the user.

[0137] Step 702: The intelligent control parameter agent converts the user's intent into one or more execution parameters.

[0138] Step 703: Based on one or more execution parameters, the process control agent controls the step execution agent to select one or more steps to execute sequentially to complete the first process, which is used to output the processing result about the user intent.

[0139] Step 704: The human-computer interaction intelligent agent outputs a system response to the user, which is determined based on the processing result.

[0140] Among them, the intelligent control parameter agent and / or the step execution agent interact with the human-computer interaction agent one or more times during operation to obtain the second dialogue information input by the user. The second dialogue information represents the information required by the corresponding agent during operation.

[0141] In one embodiment, the method further includes:

[0142] When the dialogue information input by the user is a completely new dialogue, the human-computer interaction agent determines the user's input as the first dialogue information, and determines the user's intent based on one or more rounds of the first dialogue information; and / or,

[0143] When the dialogue information entered by the user is a response to a dialogue request initiated by the first intelligent agent, the human-computer interaction intelligent agent determines that the dialogue information entered by the user is the second dialogue information, and sends the second dialogue information, together with the corresponding snapshot information, to the first intelligent agent so that the first intelligent agent can continue to operate based on the second information; wherein, the first intelligent agent includes an intelligent control parameter agent or a step execution agent, and the snapshot information represents the relevant processing information of the first intelligent agent regarding the user's intent before initiating the dialogue request.

[0144] In one embodiment, the method further includes:

[0145] The human-computer interaction intelligent agent receives a first message sent by a first intelligent agent, which is characterized as a dialogue request. The first intelligent agent includes an intelligent control parameter agent or a step execution agent.

[0146] The human-computer interaction intelligent agent outputs a dialogue request and obtains second dialogue information input by the user based on the dialogue request;

[0147] The human-computer interaction agent merges the second dialogue information with the corresponding snapshot information and sends it to the first agent so that the first agent can continue to operate based on the second information. The snapshot information represents the relevant processing information of the first agent regarding the user's intent before initiating the dialogue request.

[0148] In one embodiment, the method further includes:

[0149] After receiving the first message, the human-computer interaction agent saves the snapshot information and source information corresponding to the dialogue request. The source information is used to indicate that the dialogue request was initiated by the first agent.

[0150] Correspondingly, the human-computer interaction agent fuses the second dialogue information with the corresponding snapshot information and sends it to the first agent, including:

[0151] Upon receiving the second dialogue request, the human-computer interaction agent generates a second message, which carries the second dialogue information and the corresponding snapshot information.

[0152] Based on the source information, the human-computer interaction agent sends the second message to the first agent.

[0153] In one embodiment, the human-computer interaction agent outputs a system response to the user, including:

[0154] The human-computer interaction agent receives a second message sent by the process control agent, the second message carrying the processing result;

[0155] The human-computer interaction intelligent agent generates system responses based on the processing results;

[0156] The human-computer interaction intelligent agent outputs system responses to the user.

[0157] In one embodiment, the method further includes:

[0158] The intelligent control parameter agent identifies the intent scenario to which the user's intent belongs; and / or,

[0159] The intelligent control parameter agent performs parameter validation on one or more execution parameters; and / or,

[0160] If the human-computer interaction agent cannot determine the user's intent, the intelligent control parameter agent determines whether the process control agent should complete the first process.

[0161] In one embodiment, the method further includes:

[0162] The step-execution agent selects a model and / or tool that matches the steps to be executed.

[0163] The action agent invokes the selected model and / or tool to perform the required steps.

[0164] Figure 7 A schematic diagram illustrating the overall implementation flow of the human-computer interaction method applied to a human-computer interaction system according to an embodiment of this application is provided.

[0165] In the above scheme, multiple intelligent agents in the human-computer interaction system collaborate with each other, enabling flexible orchestration of relevant business processes. Furthermore, during the operation of the human-computer interaction system, when user supplementation or confirmation of relevant information is required, the intelligent control parameter agent and / or the step execution agent can also interact with the human-computer interaction agent one or more times during the operation of the system. This allows them to obtain the user-supplemented or confirmed information required for process execution without needing to rearrange or re-execute the business processes, thereby improving the efficiency of business process execution.

[0166] This application's embodiments, based on the ReAct framework, decompose the human-computer interaction system into human-computer interaction agents, intelligent control parameter agents, process control agents, and step execution agents. Each agent has a single, singular responsibility, allowing the dialogue task processing flow to be independent of the system framework itself, thus enhancing the flexibility and scalability of human-computer interaction and making it well-suited for complex process orchestration scenarios. Furthermore, in practical applications, fixed input and output message formats can be set for each agent, providing feasibility for unit testing of each function of the human-computer interaction system.

[0167] Furthermore, it can be seen that in the human-computer interaction system provided in this application embodiment, the intelligent control parameter agent, the process control agent, and the step execution agent all interact with the user through the human-computer interaction agent, so that the human-computer interaction agent acts as the hub for internal and external interaction of the human-computer interaction system, and transmits information through such a common channel, separating the internal and external logical framework of the human-computer interaction system and reducing the coupling between the agents.

[0168] Furthermore, message interaction between different intelligent agents and the thinking and action modes in the ReAct mode enhance the automation of human-computer interaction systems in terms of automatic process orchestration and transitions.

[0169] It should be noted that in the embodiments of this application, "first," "second," etc., are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0170] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0171] Furthermore, the technical solutions described in the embodiments of this application can be combined arbitrarily without conflict.

[0172] The above description is merely a preferred embodiment of this application and is not intended to limit the scope of protection of this application.

Claims

1. A human-machine interaction system, characterized by, The human-computer interaction system comprises a human-computer interaction intelligent agent, a control parameter intelligent agent, a process control intelligent agent and a step execution intelligent agent, wherein The human-computer interaction intelligent agent is configured to receive user inputted dialogue information, and determine a user intention based on one or more rounds of first dialogue information inputted by the user; The control parameter intelligent agent is configured to convert the user intention into one or more execution parameters; The process control intelligent agent is configured to control the step execution intelligent agent to select one or more steps to execute in sequence based on the one or more execution parameters, so as to complete a first process, wherein the first process is configured to output a processing result related to the user intention; The human-computer interaction intelligent agent is further configured to output a system response to the user, wherein the system response is determined based on the processing result; The control parameter intelligent agent and / or the step execution intelligent agent are further configured to interact with the human-computer interaction intelligent agent one or more times during operation to obtain second dialogue information inputted by the user, wherein the second dialogue information represents information required by the corresponding intelligent agent during operation.

2. The human-machine interaction system of claim 1, wherein, The human-computer interaction intelligent agent is configured to: In a case where the user inputted dialogue information belongs to a brand new dialogue, determine that the user inputted dialogue information is first dialogue information, and determine the user intention based on one or more rounds of first dialogue information inputted by the user; And / or, In a case where the user inputted dialogue information belongs to a reply to a dialogue request initiated by a first intelligent agent, determine that the user inputted dialogue information is second dialogue information, and send the second dialogue information to the first intelligent agent after fusing the second dialogue information with corresponding snapshot information, so as to enable the first intelligent agent to continue operation based on the second information, wherein the first intelligent agent comprises the control parameter intelligent agent or the step execution intelligent agent, and the snapshot information represents relevant processing information of the first intelligent agent related to the user intention before the dialogue request is initiated.

3. The human-machine interaction system of claim 1, wherein, The human-computer interaction intelligent agent is configured to: Receive a first message sent by a first intelligent agent, wherein the first message represents a dialogue request, and the first intelligent agent comprises the control parameter intelligent agent or the step execution intelligent agent; Output the dialogue request and obtain second dialogue information inputted by the user based on the dialogue request; Send the second dialogue information to the first intelligent agent after fusing the second dialogue information with corresponding snapshot information, so as to enable the first intelligent agent to continue operation based on the second information, wherein the snapshot information represents relevant processing information of the first intelligent agent related to the user intention before the dialogue request is initiated.

4. The human-machine interaction system of claim 3, wherein, The human-computer interaction intelligent agent is configured to: After receiving the first message, save the snapshot information and source information corresponding to the dialogue request, wherein the source information is configured to indicate that the dialogue request is initiated by the first intelligent agent; After receiving the second dialogue request, generate a second message, wherein the second message carries the second dialogue information and corresponding snapshot information; Based on the source information, send the second message to the first intelligent agent.

5. The human-machine interaction system of claim 1, wherein, The human-computer interaction intelligent agent is configured to: receiving a second message sent by the flow control agent, the second message carrying the processing result; generating the system response based on the processing result; outputting the system response to the user.

6. The human-machine interaction system of claim 1, wherein, The intelligent control parameter agent is further configured to: identify an intention scenario to which the user intention belongs; and / or, perform parameter verification on the one or more execution parameters; and / or, in a case where the human-computer interaction agent fails to determine the user intention, determine whether to complete the first flow by the flow control agent.

7. The human-machine interaction system of claim 1, wherein, The step execution agent is configured to: select a model and / or a tool matched with the step to be executed based on the step to be executed; and invoke the model and / or the tool to execute the step to be executed.

8. A human-machine interaction method, characterized by, The method is applied to a human-computer interaction system, and the human-computer interaction system comprises a human-computer interaction agent, an intelligent control parameter agent, a flow control agent and a step execution agent. The method comprises: The human-computer interaction agent receives dialogue information input by a user, and determines a user intention based on one or more rounds of first dialogue information input by the user. The intelligent control parameter agent converts the user intention into one or more execution parameters. The flow control agent controls the step execution agent to select one or more steps to be executed in sequence based on the one or more execution parameters, so as to complete a first flow for outputting a processing result about the user intention. The human-computer interaction agent outputs a system response to the user, and the system response is determined based on the processing result. In the running process, the intelligent control parameter agent and / or the step execution agent interact with the human-computer interaction agent one or more times to obtain second dialogue information input by the user, and the second dialogue information represents information required in the running process of the corresponding agent.

9. The human-machine interaction method of claim 8, wherein, The method further comprises: In a case where dialogue information input by the user is a brand-new dialogue, the human-computer interaction agent determines that the dialogue information input by the user is first dialogue information, and determines the user intention based on one or more rounds of first dialogue information input by the user; and / or, In a case where dialogue information input by the user is a reply to a dialogue request initiated by a first agent, the human-computer interaction agent determines that the dialogue information input by the user is second dialogue information, and sends the second dialogue information fused with corresponding snapshot information to the first agent, so that the first agent continues to run based on the second information; wherein the first agent comprises the intelligent control parameter agent or the step execution agent, and the snapshot information represents relevant processing information about the user intention before the first agent initiates the dialogue request.

10. The human-machine interaction method of claim 8, wherein, The method further comprises: The human-computer interaction agent receives a first message sent by a first agent, and the first message represents a dialogue request; the first agent comprises the intelligent control parameter agent or the step execution agent. The human-computer interaction agent outputs the dialogue request and obtains second dialogue information input by the user based on the dialogue request. The human-computer interaction agent fuses the second dialogue information with corresponding snapshot information and sends the fused information to the first agent, so that the first agent continues to run based on the second information, wherein the snapshot information represents relevant processing information of the first agent about the user intent before initiating the dialogue request.

11. The human-machine interaction method of claim 10, wherein, The method further comprises: After receiving the first message, the human-computer interaction agent saves the snapshot information and source information corresponding to the dialogue request, and the source information is used to indicate that the dialogue request is initiated by the first agent. Correspondingly, the human-computer interaction agent fuses the second dialogue information with corresponding snapshot information and sends the fused information to the first agent, comprising: After receiving the second dialogue request, the human-computer interaction agent generates a second message, and the second message carries the second dialogue information and corresponding snapshot information. Based on the source information, the human-computer interaction agent sends the second message to the first agent.

12. The human-machine interaction method of claim 8, wherein, The human-computer interaction agent outputs a system response to the user, comprising: The human-computer interaction agent receives a second message sent by the flow control agent, and the second message carries the processing result. The human-computer interaction agent generates the system response based on the processing result. The human-computer interaction agent outputs the system response to the user.

13. The human-machine interaction method of claim 8, wherein, The method further comprises: The intelligent control parameter agent identifies the intent scenario to which the user intent belongs; and / or, The intelligent control parameter agent performs parameter verification on the one or more execution parameters; and / or, In the case that the human-computer interaction agent cannot determine the user intent, the intelligent control parameter agent determines whether to complete the first flow by the flow control agent.

14. The human-machine interaction method of claim 8, wherein, The method further comprises: The step execution agent selects a model and / or a tool matched with the step to be executed based on the step to be executed; The step execution agent calls the model and / or the tool to execute the step to be executed.

Images