CN-122002063-A - Server, display device and task processing method

CN122002063ACN 122002063 ACN122002063 ACN 122002063ACN-122002063-A

Abstract

The application provides a server, display equipment and a task processing method. When the server receives the task instruction, semantic analysis is performed on the task instruction to obtain semantic analysis information, and behavior block information is determined according to the semantic analysis information. The semantic analysis information and the behavior block information are input into a task processing model, a target task can be determined based on the task processing model, and then a target data packet for representing the target task is transmitted back to the display device, so that the display device invokes a corresponding system tool to perform interaction with a user based on the target task. In the semantic understanding process, the behavior block information is added as prompt and evaluation information, so that the understanding capability of the model to the user intention can be improved, and further the task processing capability of the intelligent agent is improved.

Inventors

PAN JIABIN
YUE WENHAO

Assignees

海信视像科技股份有限公司

Dates

Publication Date: 20260508
Application Date: 20241107

Claims (10)

1. A server for a server, which comprises a server and a server, characterized by comprising the following steps: a communicator configured to establish a communication connection with a display device; A processor configured to: responding to a task instruction sent by the display equipment, and executing semantic analysis on the task instruction to obtain semantic analysis information; Searching behavior block information associated with the semantic analysis result according to the semantic analysis information; Inputting the semantic analysis information and the behavior block information into a task processing model, and controlling the task processing model to determine a target task according to the semantic analysis information and the behavior block information; and transmitting the target data packet for representing the target task back to the display device so that the display device can call a system tool corresponding to the target task to execute interaction with a user based on the target task.
2. The server of claim 1, wherein the behavior block information is stored in a behavior block database, the behavior block information comprising a behavior block index, wherein the processor performs searching for behavior block information associated with the semantic analysis result based on the semantic analysis result, and is specifically configured to: calculating semantic similarity between semantic analysis information and the behavior block index; Sorting the behavior block indexes corresponding to the semantic similarity according to the numerical value of the semantic similarity to determine a target behavior block index; And determining target behavior block information based on the target behavior block index.
3. The server of claim 1, wherein the processor performs inputting the semantic analysis information and the behavior block information into a task processing model, further configured to: The method comprises the steps of sending prompt information to the task processing model, wherein the prompt information is used for prompting a system tool which can be called when the task processing model executes a target task and prompting an execution flow when the task processing model executes the target task.
4. The server of claim 1, wherein the processor controls the task processing model to determine a target task from the semantic analysis information and the behavior block information, and is specifically configured to: controlling the task processing model to determine a target subtask according to the semantic analysis information and the action to be executed in the behavior block information; controlling the task processing model to call a subtask completion tool according to the task type of the target subtask so as to perform interaction with a user based on at least one subtask completion tool; And after the execution of the last target subtask is finished, recording the interaction process of the display equipment and the user, and obtaining a user behavior record so as to generate behavior block information based on the user behavior record.
5. The server of claim 4, wherein the processor execution control the task processing model to determine a target subtask according to the action to be performed in the semantic analysis information and the behavior block information is specifically configured to: controlling the task processing model to determine a first replacement word in the semantic analysis and a second replacement word in the description information based on the semantic analysis information and the description information corresponding to the key field of the behavior block information; And controlling the task processing model to replace the second replacement word with the first replacement word so as to determine a target subtask.
6. The server of claim 4, wherein the processor generates behavior block information based on the user behavior record, and is specifically configured to: Acquiring the user behavior record; Dividing the user behavior record into at least 1 behavior combination based on a behavior structure of the behavior block information and execution time of actions to be executed, wherein the behavior combination comprises at least 1 action to be executed, the behavior structure comprises the number and the type of key fields of the behavior block information, and the key fields comprise the occurrence reason, the generation motivation and the execution feedback of the actions to be executed; generating behavior block induction description information according to key fields corresponding to actions to be executed in the behavior combination; And generating behavior block information according to the behavior combination, the key field and the block induction description information.
7. The server of claim 6, wherein the actions to be performed include a first action to be performed and a second action to be performed, wherein the processor performs the actions to be performed included in the task instruction determined from the semantic analysis information, and is specifically configured to: Acquiring a first time stamp corresponding to the first action to be executed and a second time stamp corresponding to the second action to be executed; If the time difference between the time node represented by the first time stamp and the time node represented by the second time stamp is larger than a time threshold value, determining that the first action to be executed and the second action to be executed do not belong to the same behavior combination; and if the time difference between the time node represented by the first time stamp and the time node represented by the second time stamp is smaller than or equal to a time threshold value, determining that the first action to be executed and the second action to be executed belong to the same behavior combination.
8. The server of claim 6, wherein the processor performs generating behavior block information from the behavior combinations, key fields, and the block summary description information, and is specifically configured to: Generating a behavior block index according to the block induction description information; and generating behavior block information according to the behavior combination, the key field and the behavior block index.
9. A display device, characterized by comprising: A communication device configured to establish a communication connection with a server; A controller configured to: The method comprises the steps of receiving task instructions, responding to the received task instructions, sending the task instructions to a server so that the server executes semantic analysis on the task instructions to obtain semantic analysis information, and inputting the semantic analysis information and behavior block information determined according to the semantic analysis information to a task processing model by the server so as to determine a target task based on the task processing model; receiving a target data packet returned by the server, analyzing the target data packet, and determining a target task corresponding to a task instruction; and invoking a system tool corresponding to the target task to perform interaction with a user based on the system tool.
10. A task processing method, characterized by being applied to the server according to any one of claims 1 to 8, comprising: Responding to a task instruction sent by a display device, and executing semantic analysis on the task instruction to obtain semantic analysis information; Searching behavior block information associated with the semantic analysis result according to the semantic analysis information; Inputting the semantic analysis information and the behavior block information into a task processing model, and controlling the task processing model to determine a target task according to the semantic analysis information and the behavior block information; and transmitting the target data packet for representing the target task back to the display device so that the display device can call a system tool corresponding to the target task to execute interaction with a user based on the target task.

Description

Server, display device and task processing method Technical Field The present application relates to the field of natural language processing technologies, and in particular, to a server, a display device, and a task processing method. Background In the human-computer interaction process, the display device can analyze task instructions sent by the user based on the large model by means of an agent deployed in the server to identify the real intention of the user, and target task planning and execution are carried out according to the real intention. However, the understanding capability of the intelligent agent to the task is completely based on the resolving capability of the large model in the task processing process, and the sensing capability to the corresponding scene is lost. And further, the degree of matching of the result of task processing with the actual intention of the user is reduced. In order to solve the problems, the perception capability of the large model to a specific scene can be enhanced by training and fine-tuning the large model, so that the task processing capability of an intelligent agent is improved, but the data collection cost of the mode is high, and the adjusted model has hysteresis in time and is difficult to adapt to a changed application scene. In addition, links such as ReAct prompts, SOP standard flows and the like can be cited to improve the framework of the intelligent agent so as to improve the task processing capacity of the intelligent agent. However, this approach is difficult to achieve both universality and professionality of task processing procedures. Disclosure of Invention Some embodiments of the present application provide a server, a display device, and a task processing method, so as to solve the problem of poor task processing capability of an agent. In a first aspect, some embodiments of the present application provide a server including a communicator, a processor. The communicator is configured to establish a communication connection with the display device, and the processor is configured to: and responding to the received task instruction sent by the display equipment, and executing semantic analysis on the task instruction to obtain semantic analysis information. And searching behavior block information associated with the semantic analysis result according to the semantic analysis information. Inputting the semantic analysis information and the behavior block information into a task processing model, and controlling the task processing model to determine a target task according to the semantic analysis information and the behavior block information. And transmitting the target data packet for representing the target task back to the display device so that the display device can call a system tool corresponding to the target task to execute interaction with a user based on the target task. Therefore, the server can input the semantic analysis information and the behavior block information into the task processing model together, so that the task processing model can analyze and execute task instructions by combining the use characteristics of the user contained in the behavior block information in a specific scene, and the task processing capability of the intelligent agent can be improved. In some possible embodiments, the behavior block information is stored in a behavior block database, the behavior block information comprises a behavior block index, and the processor performs searching of the behavior block information associated with the semantic analysis result according to the semantic analysis result and is specifically configured to: calculating semantic similarity between semantic analysis information and the behavior block index; Sorting the behavior block indexes corresponding to the semantic similarity according to the numerical value of the semantic similarity to determine a target behavior block index; And determining target behavior block information based on the target behavior block index. In this way, the behavior block information with the highest matching degree with the task instruction can be determined by calculating the semantic similarity between the semantic analysis information obtained by the task instruction and the behavior block index. And the perceptibility of the intelligent agent to the specific use scene of the user is improved through the searching accuracy of the behavior block information. In some possible embodiments, the processor performs inputting the semantic analysis information and the behavior block information into a task processing model, further configured to: The method comprises the steps of sending prompt information to the task processing model, wherein the prompt information is used for prompting a system tool which can be called when the task processing model executes a target task and prompting an execution flow when the task processing model executes the target task. Therefore, the processin