CN-121979992-A - Vehicle intelligent question-answering method, device, electronic equipment, medium and program product

CN121979992ACN 121979992 ACN121979992 ACN 121979992ACN-121979992-A

Abstract

The application is applicable to the technical field of vehicle overhaul and provides a vehicle intelligent question-answering method, a device, electronic equipment, a medium and a program product, wherein the method comprises the steps of obtaining target overhaul content of a target vehicle; extracting target fault information and target complaint information of a target vehicle from target overhaul contents, analyzing a target overhaul scene of the target vehicle according to vehicle faults indicated by the target fault information and overhaul abnormal conditions indicated by the target complaint information, determining a target speech template matched with the target overhaul scene from a preset speech library, generating emotion expression prompts of fault overhaul replies by a speech model of a vehicle, and inputting the target overhaul contents and the target speech template into the vehicle speech model to obtain target fault overhaul replies of the target overhaul contents. The scheme can reduce the mechanical hardness degree of reply contents while considering the expertise of the intelligent question-answering service of the vehicle.

Inventors

LIU XIN
LAI ZHEN
BAO ZHENWEN

Assignees

深圳市元征科技股份有限公司

Dates

Publication Date: 20260505
Application Date: 20260123

Claims (10)

1. An intelligent question-answering method for a vehicle is characterized by comprising the following steps: acquiring target overhaul content of a target vehicle; Extracting target fault information and target complaint information of the target vehicle from the target overhaul content; analyzing a target overhaul scene of the target vehicle according to the vehicle fault indicated by the target fault information and the overhaul abnormal condition indicated by the target complaint information, wherein the target overhaul scene represents a fault type, a resource demand type or an overhaul abnormal type which needs to be concerned in the overhaul process of the target vehicle; Determining a target conversation template matched with the target overhaul scene from a preset conversation library, wherein the preset conversation library comprises overhaul scenes respectively corresponding to a plurality of conversation templates, and the conversation templates comprise emotion expression prompts for indicating a vehicle question-answer language model to generate fault overhaul answers; and inputting the target overhaul content and the target speech template into a vehicle question-answer language model to obtain a target troubleshooting answer of the target overhaul content.
2. The method of claim 1, wherein analyzing the inspection scene information of the target vehicle based on the vehicle fault indicated by the target fault information and the inspection abnormality indicated by the target complaint information comprises: determining a target fault type of the target vehicle according to a first corresponding relation and the vehicle fault indicated by the target fault information, wherein the first corresponding relation comprises a plurality of fault types and at least one vehicle fault corresponding to the plurality of fault types respectively; determining a target maintenance abnormality type of the target vehicle according to a second corresponding relation and the maintenance abnormality indicated by the target complaint information, wherein the second corresponding relation comprises a plurality of maintenance abnormality types and at least one maintenance abnormality corresponding to the plurality of maintenance abnormality types respectively; Obtaining maintenance resources required for maintaining the vehicle faults of the target vehicle according to the target fault type, the target maintenance abnormal type and a preset maintenance resource library, and obtaining a target resource requirement type; and reasoning a target overhaul scene where the target vehicle is located according to the target fault type, the target overhaul abnormality type and the target resource demand type.
3. The method of claim 2, wherein the determining, from a pre-determined phone library, a target phone template that matches the target overhaul scene, comprises: Determining the similarity between the overhaul scene corresponding to each of a plurality of conversation templates in the preset conversation library and the target overhaul scene to obtain the similarity corresponding to the plurality of conversation templates; and determining the conversation template with the highest corresponding similarity as the target conversation template.
4. A method according to any one of claims 1 to 3, wherein the method further comprises: Acquiring historical overhaul corpus data, wherein the historical overhaul corpus data comprises historical dialogue content in at least one vehicle fault overhaul process; respectively extracting a reference overhaul scene and a reference fault overhaul reply corresponding to each historical dialogue content from the historical dialogue content of the historical overhaul corpus data; extracting a conversation template corresponding to each reference overhaul scene from a reference overhaul reply corresponding to the reference overhaul scene; And storing each reference overhaul scene and a conversation template corresponding to the reference overhaul scene in an associated mode to obtain the preset conversation library.
5. The method of claim 4, wherein the method further comprises: Obtaining model training data, wherein the model training data comprises sample troubleshooting replies corresponding to a plurality of sample overhauling contents respectively and sample speaking templates corresponding to the sample troubleshooting replies; In the nth model iterative training process, inputting each sample overhaul content and a sample telephone template corresponding to each sample overhaul content into a large language model, wherein the large language model is used for reasoning a predicted fault overhaul reply of the sample overhaul content according to the sample overhaul content and the sample telephone template corresponding to the sample overhaul content aiming at each sample overhaul content, and n is a positive integer; Determining a performance index corresponding to a large language model according to the difference between the sample troubleshooting reply and the predicted troubleshooting reply of each sample troubleshooting content, wherein the performance index is used for indicating the capacity of the large language model to generate the troubleshooting reply conforming to a speaking template; Under the condition that the performance index does not accord with a preset performance threshold, entering an n+1st model iterative training process; and under the condition that the performance index accords with a preset performance threshold value or the iterative training frequency of the large language model reaches the preset frequency, determining the trained large language model as the vehicle question-answering language model.
6. The method of claim 5, wherein the method further comprises: updating the preset speech library according to a preset period; and incrementally updating the vehicle question-answering language model according to the updated contents in the preset speech library.
7. An intelligent vehicle question-answering device, comprising: the acquisition module is used for acquiring the target overhaul content of the target vehicle; the extraction module is used for extracting target fault information and target complaint information of the target vehicle from the target overhaul content; The analysis module is used for analyzing a target overhaul scene of the target vehicle according to the vehicle fault indicated by the target fault information and the overhaul abnormal condition indicated by the target complaint information, wherein the target overhaul scene represents a fault type, a resource demand type or an overhaul abnormal type which needs to be concerned in the overhaul process of the target vehicle; The determining module is used for determining a target conversation template matched with the target overhaul scene from a preset conversation library, wherein the preset conversation library comprises overhaul scenes respectively corresponding to a plurality of conversation templates, and the conversation templates comprise emotion expression prompts for indicating a vehicle question-answering language model to generate fault overhaul answers; And the question-answering module is used for inputting the target overhaul content and the target speech template into a vehicle question-answering language model to obtain a target trouble shooting answer of the target overhaul content.
8. An electronic device comprising a processor, a memory and a computer program stored in the memory and executable on the processor, wherein execution of the computer program by the processor causes the electronic device to implement the method of any one of claims 1-6.
9. A computer readable storage medium storing 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 6.
10. A computer program product comprising a computer program which, when run, causes the method of any one of claims 1 to 6 to be performed.

Description

Vehicle intelligent question-answering method, device, electronic equipment, medium and program product Technical Field The application belongs to the technical field of vehicle overhaul, and particularly relates to an intelligent vehicle question-answering method, device, electronic equipment, medium and program product. Background With the development of artificial intelligence technology, maintenance work is assisted by intelligent interaction in the field of vehicle maintenance and repair, and has become a common working mode. The conventional intelligent vehicle interaction scheme is to rely on a general large language model to provide professional answer content, but from the service perspective, the large language model provides mechanical answer content which is hard and cannot take the emotion of a user into consideration, so that the use experience of the intelligent vehicle answer service is affected. Therefore, how to reduce the mechanical hardness of the answer content and improve the use experience of the intelligent vehicle answering service of the user while considering the professionality of the intelligent vehicle answering service has become a technical problem to be solved at present. Disclosure of Invention The embodiment of the application provides a vehicle intelligent question-answering method, a device, electronic equipment, a medium and a program product, which can solve the problems of reducing the mechanical hardness degree of answer content and improving the use experience of the user vehicle intelligent question-answering service while considering the speciality of the vehicle intelligent question-answering service. In a first aspect, an embodiment of the present application provides a vehicle intelligent question-answering method, including: acquiring target overhaul content of a target vehicle; Extracting target fault information and target complaint information of a target vehicle from the target overhaul content; Analyzing a target overhaul scene of the target vehicle according to the vehicle faults indicated by the target fault information and the overhaul abnormal conditions indicated by the target complaint information, wherein the target overhaul scene represents a fault type, a resource demand type or an overhaul abnormal type which needs to be concerned in the overhaul process of the target vehicle; Determining a target conversation template matched with a target overhaul scene from a preset conversation library, wherein the preset conversation library comprises overhaul scenes respectively corresponding to a plurality of conversation templates, and the conversation templates comprise emotion expression prompts for indicating a vehicle question-answer language model to generate fault overhaul replies; and inputting the target overhaul content and the target speaking template into a vehicle question-answer language model to obtain a target troubleshooting answer of the target overhaul content. In some embodiments, determining a target session template from a pre-set session library that matches the target overhaul scene includes: determining the similarity between the overhaul scene and the target overhaul scene corresponding to a plurality of conversation templates in a preset conversation library respectively, and obtaining the similarity corresponding to the conversation templates; And determining the speech template with the highest corresponding similarity as the target speech template. In some embodiments, the method further comprises: Acquiring historical overhaul corpus data, wherein the historical overhaul corpus data comprises historical dialogue content in at least one vehicle fault overhaul process; respectively extracting a reference overhaul scene and a reference fault overhaul reply corresponding to each historical dialogue content from the historical dialogue content of the historical overhaul corpus data; Extracting a conversation template corresponding to each reference overhaul scene from the reference overhaul replies corresponding to the reference overhaul scenes; And storing each reference overhaul scene and a conversation template corresponding to the reference overhaul scene in an associated mode to obtain a preset conversation library. In some embodiments, the method further comprises: Obtaining model training data, wherein the model training data comprises sample troubleshooting replies corresponding to a plurality of sample overhauling contents respectively and sample speaking templates corresponding to the sample troubleshooting replies; in the nth model iterative training process, each sample overhaul content and a sample conversation template corresponding to each sample overhaul content are input into a large language model, the large language model is used for reasoning a prediction fault overhaul reply of the sample overhaul content according to the sample overhaul content and the sample conversation template corresponding to the sample