CN-121979478-A - Dialogue task management method, dialogue task management system, dialogue task management vehicle, dialogue task management storage medium and dialogue task management program product

CN121979478ACN 121979478 ACN121979478 ACN 121979478ACN-121979478-A

Abstract

The application provides a dialogue task management method, a dialogue task management system, a dialogue task management vehicle, a dialogue task storage medium and a dialogue task program product, and relates to the technical field of vehicles. According to the dialogue task management method, a multi-dialogue task interrupt recovery mechanism based on a task stack is adopted, when a user initiates a new second dialogue task in the process of executing the first dialogue task by the system, the first dialogue task is interrupted, task information of the first dialogue task is stored in the task stack, the second dialogue task is activated, further when the second dialogue task is executed, or a task recovery instruction input by the user is detected, the task information is popped up from the task stack, the execution of the first dialogue task is recovered, seamless connection recovery of the interrupted dialogue task and dynamic switching among the multi-dialogue tasks are realized, the distraction in the driving process is reduced, the voice interaction efficiency and the driving safety are remarkably improved, and the safety and the interaction continuity requirements in a driving scene are met.

Inventors

LI XIAOHUA
YAO LEI
LIU JUNRONG

Assignees

华勤技术股份有限公司

Dates

Publication Date: 20260505
Application Date: 20260407

Claims (10)

1. A method for managing conversational tasks, comprising: acquiring a current dialogue instruction input by a user and task information of a first dialogue task, wherein the first dialogue task is a dialogue task being executed by a system; determining whether to interrupt the first dialogue task according to the current dialogue instruction and the task information; If the first dialogue task is determined to be interrupted, storing the task information into a task stack, and activating a second dialogue task corresponding to the current dialogue instruction, wherein the task stack adopts a last-in-first-out structure, and the background execution function of the interrupted first dialogue task continuously operates; And responding to the completion of the execution of the second dialogue task or detecting a task recovery instruction input by a user, popping up the task information from the task stack, and recovering the execution of the first dialogue task according to the task information.
2. The method according to claim 1, wherein determining whether to interrupt the first session based on the current session instruction and the task information comprises: Determining a multi-semantic dimension level according to the current dialogue instruction and the task information; Based on a preset multi-semantic dimension decision matrix, identifying a target topic boundary type corresponding to the current dialogue instruction according to the multi-semantic dimension level; And determining whether to interrupt the first dialogue task according to the target topic boundary type.
3. The dialog task management method of claim 2, wherein the task information includes a history dialog instruction input by a previous round of user, an N-round history dialog text corresponding to the first dialog task, and a first task system prompt word, the multi-semantic dimension level includes a semantic jump level, a reference dependency level, and a task relevance level, and the determining the multi-semantic dimension level according to the current dialog instruction and the task information includes: After the current dialogue instruction, the historical dialogue instruction and the first task system prompt word are spliced, inputting a lightweight semantic model to perform text pair correlation calculation, obtaining text pair correlation scores output by the lightweight semantic model, and determining the semantic jump degree grade according to the text pair correlation scores; After the current dialogue instruction, the N rounds of historical dialogue texts and the first task system prompt word are spliced, inputting a lightweight large language model for index dependency detection, obtaining index dependency information output by the lightweight large language model, and determining the index dependency level according to the index dependency information; and after the current dialogue instruction and the first task system prompt word are spliced, inputting a lightweight semantic model for task relevance evaluation, obtaining task relevance scores output by the lightweight semantic model, and determining the task relevance grade according to the task relevance scores.
4. The dialog task management method of claim 2, wherein the target topic boundary type includes intra-task continuation, intra-task switch, off-task interrupt, new task on, and the determining whether to interrupt the first dialog task according to the target topic boundary type includes: determining whether the target topic boundary type is an off-task interrupt; and if the target topic boundary type is the off-task interrupt, determining to interrupt the first dialogue task.
5. The conversational task management method of claim 4, further comprising: And if the target topic boundary type is that the new task is started, suspending the first dialogue task, and clearing the history dialogue text corresponding to the first dialogue task and all the completed dialogue tasks stored in the task stack.
6. The dialog task management method of any of claims 1 to 5, further comprising, after the activating the second dialog task corresponding to the current dialog instruction: inputting the current dialogue instruction and a second task system prompt word corresponding to the second dialogue task into a large language model to generate a natural language response; and executing the background execution function of the second dialogue task according to the natural language response.
7. A vehicle system comprising a conversational task management module configured to perform the conversational task management method of any one of claims 1-6.
8. The vehicle system of claim 7, wherein the conversational task management module includes a task state management layer to manage conversational tasks stored in a hierarchical context storage structure.
9. A vehicle comprising a vehicle body and the vehicle system according to claim 7 or 8.
10. A computer readable storage medium having stored therein computer executable instructions which when executed by a processor are for implementing the dialog task management method of any of claims 1 to 6.

Description

Dialogue task management method, dialogue task management system, dialogue task management vehicle, dialogue task management storage medium and dialogue task management program product Technical Field The present application relates to the field of vehicle technologies, and in particular, to a method, a system, a vehicle, a storage medium, and a program product for managing a session task. Background Along with the rapid development of intelligent network automobiles, the demand of users for intelligent cabin voice interaction is remarkably enhanced, and the users need to complete various dialogue task interactions in driving scenes and realize flexible switching among a plurality of dialogue tasks. In the related art, a multi-domain dialog system based on user intention classification is generally used for dialog task management. Specifically, the multi-domain dialogue system presets fixed dialogue domains such as navigation, car control, audio and video, chat and the like, carries out intention classification and entity extraction on a voice instruction input by a user through a natural language understanding model, and then executes dialogue response according to a preset dialogue flow by a dialogue manager. The state machine and state transition are predefined in each dialogue field, so that multi-round dialogue task management is realized. However, the inventor researches and discovers that in the manner of performing dialogue task management by adopting the multi-field dialogue system based on user intention classification, in some scenes, the voice interaction efficiency is low, the potential safety hazard exists, and the requirements of safety and interaction continuity in driving scenes are difficult to meet. Disclosure of Invention The application provides a dialogue task management method, a dialogue task management system, a dialogue task management vehicle, a dialogue task management storage medium and a dialogue task management program product, which are used for solving the problems that in the related art, a multi-field dialogue system based on user intention classification is adopted to conduct dialogue task management, in some scenes, the efficiency of voice interaction is low, potential safety hazards exist, and the requirements of safety and interaction continuity in driving scenes are difficult to meet. The application provides a dialogue task management method, which comprises the steps of obtaining a current dialogue instruction input by a user and task information of a first dialogue task, wherein the first dialogue task is a dialogue task which is being executed by a system, determining whether to interrupt the first dialogue task according to the current dialogue instruction and the task information, storing the task information into a task stack and activating a second dialogue task corresponding to the current dialogue instruction if the first dialogue task is determined to be interrupted, enabling a background execution function of the interrupted first dialogue task to continuously run by the task stack in a last-in-first-out structure, responding to the completion of the execution of the second dialogue task or detecting a task recovery instruction input by the user, popping the task information from the task stack, and recovering the execution of the first dialogue task according to the task information. In one possible implementation, determining whether to interrupt the first dialog task according to the current dialog instruction and task information includes determining a multi-semantic dimension level according to the current dialog instruction and task information, identifying a target topic boundary type corresponding to the current dialog instruction according to the multi-semantic dimension level based on a preset multi-semantic dimension decision matrix, and determining whether to interrupt the first dialog task according to the target topic boundary type. In one possible implementation, the task information comprises a historical dialogue instruction input by a previous round of users, N rounds of historical dialogue texts corresponding to a first dialogue task and a first task system prompt word, the multi-semantic dimension level comprises a semantic jump level, an indication dependency level and a task correlation level, the multi-semantic dimension level is determined according to the current dialogue instruction and the task information, the multi-semantic dimension level comprises the steps of splicing the current dialogue instruction, the historical dialogue instruction and the first task system prompt word, inputting a lightweight semantic model for performing text-to-correlation calculation to obtain a text-to-correlation score output by the lightweight semantic model, determining a semantic jump level according to the text-to-correlation score, inputting a lightweight large language model for indication dependency detection to obtain indication dependency