Vehicle intelligent question-answering method and device, electronic equipment, medium and program product
By acquiring vehicle fault information and complaint information, analyzing maintenance scenarios, and generating responses using a pre-set script library and vehicle question-and-answer language model, the problem of stiff responses in vehicle intelligent question-and-answer services has been solved, thus improving the user experience.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- LAUNCH TECH CO LTD
- Filing Date
- 2026-01-23
- Publication Date
- 2026-05-05
AI Technical Summary
Existing vehicle intelligent question-answering services rely on a general large language model, which provides rigid responses that fail to consider user emotions and negatively impact user experience.
By acquiring fault information and complaint information of the target vehicle, analyzing the repair scenario, generating fault repair responses using script templates in the preset script library and vehicle question-and-answer language model, and combining emotional expression prompts, a professional and vivid response is generated.
It improves the user experience of intelligent vehicle Q&A services by accurately analyzing vehicle maintenance scenarios and user emotions to generate professional and vivid responses, reducing the mechanical and rigid nature of the responses.
Smart Images
Figure CN121979992A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the field of vehicle maintenance technology, and in particular relates to vehicle intelligent question-and-answer methods, devices, electronic devices, media and program products. Background Technology
[0002] With the development of artificial intelligence technology, intelligent interactive assistance has become a common work mode in vehicle maintenance. Currently, commonly used intelligent vehicle interaction solutions rely on general-purpose language models to provide professional responses. However, from a service perspective, these models provide mechanical responses that are often rigid and fail to consider user emotions, thus negatively impacting the user experience of intelligent vehicle Q&A services. Therefore, how to maintain the professionalism of intelligent vehicle Q&A services while reducing the rigidity of responses and improving the user experience has become a pressing technical problem. Summary of the Invention
[0003] This application provides vehicle intelligent question-and-answer methods, devices, electronic devices, media, and program products, which can solve the problem of how to reduce the mechanical and rigid nature of the answers while maintaining the professionalism of vehicle intelligent question-and-answer services and improving the user experience of vehicle intelligent question-and-answer services.
[0004] In a first aspect, embodiments of this application provide a vehicle intelligent question-answering method, including:
[0005] Obtain the target maintenance content for the target vehicle; Extract target fault information and target complaint information of the target vehicle from the target maintenance content; Based on the vehicle malfunctions indicated by the target fault information and the maintenance anomalies indicated by the target complaint information, analyze the target maintenance scenarios of the target vehicles. The target maintenance scenarios represent the types of faults, resource requirements, or maintenance anomalies that need to be addressed during the maintenance of the target vehicles. The target dialogue template that matches the target maintenance scenario is determined from the preset dialogue script library. The preset dialogue script library includes multiple dialogue templates corresponding to different maintenance scenarios. The dialogue templates include emotional expression prompts that instruct the vehicle question-and-answer language model to generate fault maintenance responses. Input the target maintenance content and target dialogue template into the vehicle question-and-answer language model to obtain the target fault maintenance response for the target maintenance content.
[0006] In some embodiments, determining a target dialogue template that matches the target maintenance scenario from a preset dialogue script library includes: Determine the similarity between the maintenance scenarios corresponding to multiple script templates in the preset script library and the target maintenance scenario to obtain the similarity between the multiple script templates; The script template with the highest similarity is selected as the target script template.
[0007] In some embodiments, the method further includes: Obtain historical maintenance corpus data, which includes at least one historical dialogue during the vehicle malfunction maintenance process; From the historical dialogue content of the historical maintenance corpus data, extract the reference maintenance scenario and reference fault maintenance response corresponding to each historical dialogue content. For each reference maintenance scenario, extract the corresponding script template from the reference fault maintenance response for the reference maintenance scenario. Each reference maintenance scenario and its corresponding script template are associated and stored to obtain a preset script library.
[0008] In some embodiments, the method further includes: Obtain model training data, which includes sample fault repair responses corresponding to multiple sample repair contents, and sample script templates corresponding to the sample fault repair responses. During the nth model iteration training process, the inspection content of each sample and the corresponding sample dialogue template are input into the large language model. The large language model is used to infer the predicted fault repair response for each sample inspection content based on the sample inspection content and the corresponding sample dialogue template, where n is a positive integer. Based on the difference between the sample fault repair response and the predicted fault repair response for each sample repair content, the performance index corresponding to the large language model is determined. The performance index is used to indicate the ability of the large language model to generate fault repair responses that conform to the speech template. If the performance metrics do not meet the preset performance threshold, proceed to the (n+1)th model iteration training process. If the performance indicators meet the preset performance threshold, or the number of iterations of the large language model reaches the preset number, the trained large language model is determined to be the vehicle question-answering language model.
[0009] In some embodiments, the method further includes: The preset script library is updated according to a preset cycle; The vehicle question-and-answer language model is incrementally updated based on the updated content in the preset dialogue script library.
[0010] Secondly, embodiments of this application provide a vehicle intelligent question-answering device, comprising: The acquisition module is used to acquire the target maintenance content of the target vehicle; The extraction module is used to extract target fault information and target complaint information of target vehicles from the target maintenance content; The analysis module is used to analyze the target vehicle's target maintenance scenario based on the vehicle faults indicated by the target fault information and the maintenance anomalies indicated by the target complaint information. The target maintenance scenario represents the types of faults, resource requirements, or maintenance anomalies that need to be addressed during the maintenance of the target vehicle. The determination module is used to determine the target dialogue template that matches the target maintenance scenario from the preset dialogue script library. The preset dialogue script library includes multiple dialogue templates corresponding to different maintenance scenarios. The dialogue templates include emotional expression prompts that instruct the vehicle question-and-answer language model to generate fault maintenance responses. The question-and-answer module is used to input the target maintenance content and target dialogue template into the vehicle question-and-answer language model to obtain the target fault maintenance response for the target maintenance content.
[0011] Thirdly, embodiments of this application provide an electronic device including a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it causes the electronic device to perform the method described in any of the embodiments of the first aspect.
[0012] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in any of the embodiments of the first aspect.
[0013] Fifthly, embodiments of this application provide a computer program product, including a computer program, which, when run, causes the method described in any embodiment of the first aspect to be executed.
[0014] The advantages of the embodiments in this application compared with related technologies are: When performing intelligent question-and-answer sessions on vehicles, the acquired target repair information includes the target vehicle's target fault information and target complaint information. These two types of information cover both technical faults and user-perceived complaints. Based on the vehicle faults indicated by the target fault information and the repair anomalies indicated by the target complaint information, the target repair scenario for the target vehicle can be accurately analyzed from both technical and user-perceived perspectives. A target dialogue template matching the fault type, repair resource requirement type, or abnormal repair information represented by this target repair scenario is determined from a pre-set dialogue library. This library includes emotional expression prompts for generating fault repair responses, ensuring that the determined target dialogue template accurately meets the user's repair needs while also considering emotional expression. After inputting the target dialogue template and target repair content into the vehicle question-and-answer language model, the powerful reasoning and language integration capabilities of the language model can be used to generate professional and vivid target fault repair responses. This provides users with professional intelligent question-and-answer services while considering their emotions, reducing the mechanical and rigid nature of the responses and improving the user experience of the vehicle intelligent question-and-answer service. Attached Figure Description
[0015] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0016] Figure 1 This is a schematic diagram of a vehicle intelligent question-answering scenario provided in an embodiment of this application.
[0017] Figure 2 This is a flowchart illustrating a vehicle intelligent question-answering method provided in an embodiment of this application.
[0018] Figure 3 This is a flowchart illustrating another intelligent question-answering method for vehicles provided in this application embodiment.
[0019] Figure 4 This is a flowchart illustrating another intelligent question-answering method for vehicles provided in this application embodiment.
[0020] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.
[0021] Figure 6 This is a schematic diagram of the structure of a vehicle intelligent question-and-answer device provided in an embodiment of this application. Detailed Implementation
[0022] The technical solutions in the embodiments of this application will be described below with reference to the accompanying drawings.
[0023] Figure 1 This is a schematic diagram of a scenario for a vehicle intelligent question-answering method, such as... Figure 1 The scenario 100 shown includes a target vehicle 110, electronic equipment 120, and a repairman 130. The repairman 130 can be a professional vehicle malfunction repairman or a user of the target vehicle 110. The electronic equipment 120 can be a diagnostic device specifically designed for vehicle malfunction repair, or other devices capable of communicating with the vehicle, such as a mobile phone, tablet, in-vehicle terminal, laptop, ultra-mobile personal computer (UMPC), netbook, or personal digital assistant. This embodiment does not impose limitations; for convenience, a diagnostic device is used as an example. The diagnostic device may have a built-in Vehicle Communication Interface (VCI), which serves as a communication bridge connecting the diagnostic device and the target vehicle 110. The diagnostic device can also connect to an external VCI device to communicate with the target vehicle 110.
[0024] During vehicle malfunction repair, the repairman 130 can input dialogue content into the electronic device 120 via language or text. The electronic device 120 can use the dialogue content input by the repairman 130 as the target detection content for the target vehicle 110, and extract the target malfunction information and target complaint information for the target vehicle 110. Based on the vehicle malfunction indicated by the extracted target malfunction information and the repair anomaly indicated by the target complaint information, the electronic device 120 can analyze the repair scenario information of the target vehicle. The repair scenario represents the malfunction type, repair resource requirement type, and abnormal repair information of the target vehicle 110. It also determines a target dialogue template matching the target repair scenario from a preset dialogue library, which includes multiple dialogue templates corresponding to different repair scenarios. The electronic device 120 inputs the determined target dialogue template and target repair content into a trained vehicle question-and-answer language model. This model generates a solution for the vehicle malfunction of the target vehicle 110 based on the target repair content, optimizes the solution based on the target dialogue template, and outputs the target malfunction repair response. In this way, through the parallel collaboration of the dialogue module and the large language model, vivid troubleshooting responses are generated, improving the user experience.
[0025] Figure 2 This is a flowchart illustrating a vehicle intelligent question-answering method provided in an embodiment of this application, as shown below. Figure 2 As shown, the method includes the following steps.
[0026] S101, Obtain the target maintenance content of the target vehicle.
[0027] The vehicle intelligent question-answering method in this embodiment is applied to an electronic device, which may be... Figure 1 Electronic device 120 in the scene shown.
[0028] The target vehicle can be a specific vehicle that needs to be inspected.
[0029] When a user wants to learn about vehicle troubleshooting methods or vehicle troubleshooting functions of an electronic device, or other vehicle-related troubleshooting information, the user can input the troubleshooting question into the electronic device using at least one of the following methods: language, text, image, or video. Upon receiving the troubleshooting question in language, text, image, and / or video mode, the electronic device can convert the received question into text content and use this text content as the target troubleshooting content. In other words, when a user inputs a troubleshooting question as text, the electronic device can directly use that question as the target troubleshooting content.
[0030] S102, extract target fault information and target complaint information of the target vehicle from the target maintenance content.
[0031] The target fault information is data used to represent the direct or indirect fault status of the vehicle, and may include at least one of the following: fault codes, abnormal sensor data (e.g., abnormal voltage of the oxygen sensor or abnormal engine speed) or failure component records (e.g., excessive wear threshold of the transmission clutch plate).
[0032] Target complaint information is used to indicate abnormal repair records related to the vehicle repair process, which may include at least one of the following: abnormal repair experience of vehicle faults (e.g., repeated repairs, inability to diagnose faults, diagnostics being blocked or functions not supported) or obstacles to repair operations (e.g., "screws stripped and cannot be removed" or "parts model not compatible with the vehicle").
[0033] In one implementation, the electronic device can input the target maintenance content into a trained Natural Language Processing (NLP) model. This NLP model can extract target fault information and target complaint information from the target maintenance content. The NLP model can be a Convolutional Neural Network (CNN), a Long Short-Term Memory (LSTM) network, or a Bidirectional Encoder Representations from Transformers (BERT) network, etc. The NLP model can accurately recognize and understand human language. Using the trained NLP model, accurate target fault information and target complaint information can be obtained, providing a data foundation for subsequently generating vivid target fault maintenance responses.
[0034] In one implementation, the electronic device can also input the target repair content into a vehicle question-and-answer language model. This model analyzes the fault and complaint information within the target repair content and returns the fault and complaint information identified by the dialogue module to the electronic device. The electronic device can then use this returned fault and complaint information as the target fault and complaint information, respectively. It can be understood that the vehicle question-and-answer language model can be deployed in the electronic device; it is a large language model. Utilizing the language understanding and reasoning capabilities of the vehicle question-and-answer language model, accurate target fault and complaint information can be obtained, providing a data foundation for subsequently generating vivid responses to the target fault repair.
[0035] For example, if the target repair information obtained by the electronic device is "During diagnosis, the engine detected a P0301 fault. I'm unsure whether it's a spark plug problem or a coil problem," the electronic device inputs this target repair information into the vehicle question-and-answer language model. The vehicle question-and-answer language model analyzes the target repair information and determines that "the engine detected a P0301 fault" indicates that the target vehicle has a misfire in cylinder 1 of the engine. The "unsure whether it's a spark plug problem or a coil problem" indicates that the vehicle fault has an undiagnosable complaint. The target fault information that the electronic device can obtain can be "cylinder 1 misfire," and the target complaint information can be "undiagnosable."
[0036] In one implementation, the electronic device can also respond to a user's input operation on fault information in the target maintenance content, using the input fault information as the target fault information, and respond to a user's input operation on complaint information in the target maintenance content, using the input complaint information as the target complaint information.
[0037] S103, based on the vehicle malfunction indicated by the target fault information and / or the maintenance abnormality indicated by the target complaint information, analyze the target maintenance scenario of the target vehicle.
[0038] Among them, the target maintenance scenario refers to the types of faults, maintenance resource requirements, or maintenance anomalies that need to be addressed during the maintenance of the target vehicle.
[0039] Vehicle malfunctions are used to indicate the specific abnormal state of the target vehicle that needs to be repaired, such as "engine misfire", "brake pedal travel is too long" or "air conditioning failure".
[0040] The "Abnormal Inspection Status" section is used to indicate unexpected problems in the inspection process of the target vehicle reflected in the target complaint information, such as "incomplete inspection", "repeated repairs", "unable to diagnose", "diagnostic tool dead", or "incompatible parts".
[0041] By combining the vehicle malfunction information indicated by the target fault information and the maintenance anomaly indicated by the target complaint information, the electronic device can analyze the target vehicle's target maintenance scenario and define the specific maintenance situation of the target vehicle. For example, the maintenance situation may include information such as the maintenance object of the target vehicle (e.g., engine), the maintenance focus (e.g., investigating the cause of the fire), maintenance needs, or potential related issues (e.g., whether the maintenance anomaly masks the real fault). In the subsequent determination of the target response template, the target response template can be accurately matched to the specific maintenance situation corresponding to the target maintenance scenario, which can reduce the situation of stiff responses caused by using a general template.
[0042] In one implementation, the electronic device can input target fault information and target complaint information into a vehicle question-and-answer language model. This model can invoke a dialogue analysis tool to determine the fault type of the target vehicle based on the vehicle fault indicated by the target fault information, and to determine the maintenance anomaly type of the target vehicle based on the maintenance anomaly indicated by the target complaint information. Furthermore, based on the determined fault type and maintenance anomaly type, the model can infer the current maintenance scenario of the target vehicle, thus obtaining the target maintenance scenario. The dialogue analysis tool may include an artificial intelligence (AI) model trained on a sample training set, specifically designed to infer the vehicle's maintenance scenario based on the vehicle fault indicated by the fault information and the maintenance anomaly indicated by the complaint information. The sample training set includes multiple training sample pairs, each containing sample maintenance content and its corresponding sample fault type and sample maintenance anomaly type, along with labels for the corresponding maintenance scenarios.
[0043] In one implementation, such as Figure 3 As shown, based on the vehicle malfunction indicated by the target malfunction information and the maintenance anomaly indicated by the target complaint information, the maintenance scenario information of the target vehicle is analyzed, including the following S201 to S204.
[0044] S201, Based on the first correspondence and the vehicle fault indicated by the target fault information, determine the target fault type of the target vehicle.
[0045] The first correspondence includes multiple fault types and at least one vehicle fault corresponding to each of the multiple fault types.
[0046] Electronic devices can analyze the specific vehicle fault based on the fault status data in the target fault information, and find the corresponding fault type from the first correspondence relationship to obtain the target fault type. For example, if the target fault information includes a specific value of the intake manifold pressure (which is less than the pressure threshold) or the fault code P0107, the electronic device determines that the vehicle fault is an intake manifold leak. In the first correspondence relationship, this fault corresponds to the intake manifold leak fault type, and the electronic device can determine that the target fault type is an intake manifold leak.
[0047] S202, Based on the second correspondence and the abnormal maintenance situation indicated by the target complaint information, determine the target abnormal maintenance type of the target vehicle.
[0048] The second correspondence includes multiple maintenance anomaly types and at least one maintenance anomaly situation corresponding to each of the multiple maintenance anomaly types.
[0049] Electronic devices can analyze the abnormal repair records in the target complaint information to identify the specific repair anomalies of the target vehicle, and then find the corresponding abnormal repair type from the second correspondence relationship to obtain the target abnormal repair type. For example, if the target complaint information includes information indicating that a vehicle has undergone multiple repairs, the electronic device can determine that the detected anomaly is that the same fault exists after multiple repairs. In the second correspondence relationship, this repair anomaly corresponds to the abnormal repair type "repeated repairs," and the electronic device can determine that the target abnormal repair type is repeated repairs.
[0050] S203: Based on the target fault type, the target maintenance anomaly type, and the preset maintenance resource library, obtain the maintenance resources required when maintaining the vehicle fault of the target vehicle, and obtain the target resource requirement type.
[0051] The preset maintenance resource library includes at least one fault type and / or at least one maintenance anomaly type corresponding to various maintenance resources.
[0052] The electronic device can determine at least one maintenance resource corresponding to the target fault type from the preset maintenance resource library, and preset at least one maintenance resource corresponding to the target maintenance anomaly type in the maintenance resource library. The maintenance resources corresponding to the two are combined as the maintenance resources required when repairing the vehicle fault of the target vehicle, thus obtaining the target resource requirement type.
[0053] S204. Based on the target fault type, target maintenance anomaly type, and target resource requirement type, infer the target maintenance scenario in which the target vehicle is located.
[0054] The electronic device can input the determined target fault type, target maintenance anomaly type, and target resource requirement type into the vehicle question-and-answer language model. This model can then call a dialogue analysis tool, which can infer the target maintenance scenario through a configured AI model. It can be understood that steps S201 to S204 can also be executed by the electronic device through the dialogue analysis tool.
[0055] In this implementation, based on the first and second correspondences, objective technical fault information and user subjective complaint information are accurately mapped to target fault types at the vehicle fault technology level and target maintenance anomaly types at the scenario-based label level. This achieves the transformation from vague user inquiries to precise information identification, reducing information comprehension bias in intelligent Q&A services during vehicle maintenance. Furthermore, by pre-setting a maintenance resource library, the target fault types and target maintenance anomaly types are combined to clarify the required testing resources for the target vehicle's fault, i.e., the target resource demand type. Integrating the target resource demand type, target fault type, and target maintenance anomaly type, the target maintenance scenario is deduced, forming a complete "fault-scenario-resource" contextual profile. This ensures the professionalism of the analysis during intelligent Q&A and enhances the relevance and personalization of subsequent script matching and response generation by identifying the maintenance scenario, providing users with more vivid and engaging responses.
[0056] S104. Determine the target dialogue template that matches the target maintenance scenario from the preset dialogue script library.
[0057] The preset script library includes multiple script templates corresponding to different maintenance scenarios.
[0058] Each script template in the pre-set script library includes an emotional prompt that instructs the vehicle question-and-answer language model to generate a fault repair response.
[0059] Emotional expression cues are used to instruct the fault repair responses generated by the vehicle question-and-answer language model to be emotional and responsive to customer anxiety, rather than simple technical replies, thus reducing the rigidity of fault repair responses.
[0060] This emotional expression prompt indicates that the script template for generating vivid responses in the intelligent Q&A service for vehicle repair must possess a neutral emotional quality. Besides emotionality, the script template should also meet the requirements of standardization and task closure. Standardization indicates that the fault repair responses generated by the vehicle Q&A language model do not include any promises of absolute repair, and clarifies the reference value of subsequent fault repair responses output by the vehicle Q&A language model. Task closure indicates that the fault repair responses generated by the vehicle Q&A language model should clearly inform the customer of the resolution process, rather than simply providing a vague timeline. The requirements of standardization and task closure can be applied to the preset reference response generation prompts and task closure prompts, respectively.
[0061] In one implementation, the electronic device can determine the similarity between the maintenance scenario and the target maintenance scenario corresponding to multiple script templates in the preset script library, and obtain the similarity between the multiple script templates; the script template with the highest similarity is determined as the target script template.
[0062] The similarity can be Euclidean distance or cosine similarity, etc., and this application embodiment does not impose any limitations. It can be understood that the inspection scenario of the script template and the target inspection scenario can be vectors.
[0063] The electronic device can calculate the similarity between the inspection scenario of each script template in the preset script library and the target inspection scenario, and select the script template with the highest similarity among the determined similarity scores as the target script template. It can be understood that when determining the highest similarity score, the electronic device may encounter multiple equal similarity scores, and these scores may be higher than others. In this case, the electronic device can randomly select a script template with the highest similarity score from among the multiple equal similarity scores as the target script template.
[0064] In this implementation, the script template with the highest similarity is determined as the target script template. This ensures that the target script template is highly compatible with the target repair scenario, which in turn guides the subsequent vehicle question-and-answer language model to generate more targeted, accurate, and emotionally resonant fault repair responses. This not only focuses on the specific fault problem of the customer's vehicle, but also alleviates the customer's emotions through appropriate scripts, ultimately enhancing the user's trust and satisfaction with the vehicle intelligent question-and-answer service, and achieving a closed-loop precise service of "scenario-script-response".
[0065] S105, input the target inspection content and target dialogue template into the vehicle question and answer language model to obtain the target fault inspection response for the target inspection content.
[0066] As a well-trained large language model, the vehicle question-and-answer language model can generate and output professional and vivid responses that are relevant to the target repair content, guided by the target dialogue template and its emotional expression prompts. In this way, electronic devices can obtain the target fault repair response and output the target fault repair response in the form of voice or text.
[0067] For example, electronic devices can input target dialogue templates into a vehicle question-and-answer language model through a dialogue module tool. While generating professional response content that matches the target repair content, the vehicle question-and-answer language model can optimize the generated professional response content based on emotion expression prompts, reference response generation prompts, and task closure prompts, so that the final response content meets the requirements of dialogue standardization, emotionality, and task closure, thus obtaining the target fault repair response.
[0068] For example, a technician (an example of a user or repairman) asks: "During my diagnosis, the engine tested positive for fault P0301. I'm unsure if it's a spark plug problem or a coil problem." (An example of the target detection content). A voice AI or service AI (in this example, an example of an electronic device integrating a vehicle question-and-answer language model and script modules) can output through steps S101 to S105: "I understand your concern. This situation is indeed prone to problems. I've integrated repair data from over 80,000 similar models using my AI system. In your car, 80% of P0301 issues are spark plug problems, and 16% are coil problems. I suggest you test the spark plugs first, which usually takes 20 minutes. If you're not testing the coil, this will save you time and improve your efficiency." (An example of a response to the target fault).
[0069] Thus, when technicians use voice AI, the AI organizes and outputs content using a large language model (an example of a vehicle question-and-answer language model) and a script module (an example of a target script template obtained in the above embodiment). The voice AI, through the script (an example of a target script template), reminds technicians of the repair sequence, greatly improving the user experience. When service personnel use service AI, the script template can also generate corresponding response references for customer-provided questions, thereby improving the service quality of service personnel.
[0070] In this embodiment, when performing intelligent question-and-answer on a vehicle, the acquired target repair content includes target fault information and target complaint information for the target vehicle. These two types of information cover both technical faults and user-perceived complaints. Based on the vehicle faults indicated by the target fault information and the repair anomalies indicated by the target complaint information, the target repair scenario for the target vehicle can be accurately analyzed from both technical and user-perceived perspectives. A target dialogue template matching the fault type, repair resource requirement type, or abnormal repair information represented by the target repair scenario is determined from a preset dialogue library. This preset dialogue library includes emotional expression prompts for generating fault repair responses, ensuring that the determined target dialogue template accurately meets the user's repair needs while also considering emotional expression. After inputting the target dialogue template and target repair content into the vehicle question-and-answer language model, the powerful reasoning and language integration capabilities of the language model can be relied upon to generate professional and vivid target fault repair responses. This provides users with professional intelligent question-and-answer services while considering their emotions, reducing the mechanical and rigid nature of the responses and improving the user experience of the vehicle intelligent question-and-answer service.
[0071] Before S101, the script module tool, preset script library, and vehicle question-and-answer language model need to be configured. This embodiment takes the pre-configuration of the script module tool, preset script library, and vehicle question-and-answer language model in the electronic device as an example. In one implementation, it can be achieved through... Figure 4 The steps shown configure the preset script library, such as... Figure 4 The method shown also includes the following steps.
[0072] S301, Obtain historical maintenance corpus data.
[0073] The historical maintenance corpus data includes at least one historical dialogue from the vehicle malfunction repair process.
[0074] Historical dialogue content can be collected from various data channels and converted into text-based conversations related to vehicle malfunction repair. Examples include historical chat logs related to vehicle malfunction repair in social media, audio and text recordings of vehicle malfunction repair or diagnosis from previous versions of voice AI, repair cases from websites or professional databases, and repair request records from professional vehicle repair systems.
[0075] Electronic devices can collect historical dialogue content from at least one vehicle malfunction repair process from multiple data channels to obtain historical repair corpus data. Alternatively, electronic devices can obtain historical repair corpus data by responding to user input of historical dialogue content.
[0076] S302, extract the reference maintenance scenario and reference fault maintenance response corresponding to each historical dialogue content from the historical maintenance corpus data.
[0077] Electronic devices can respond to user annotation operations on maintenance scenarios and fault repair responses for historical dialogue content, and obtain reference maintenance scenarios and reference fault repair responses for each historical dialogue content.
[0078] S303: For each reference maintenance scenario, extract the corresponding script template from the reference fault maintenance response for the reference maintenance scenario.
[0079] The electronic device can group the reference repair scenarios and reference fault repair responses corresponding to each vehicle fault according to the principle of the same reference repair scenario, and output each group and the reference fault repair response for each group. The electronic device can respond to the user's operation of extracting the script from the output reference fault repair responses to obtain the script template under the same reference repair scenario, thereby completing the step of extracting the script template corresponding to the reference repair scenario from the reference fault repair responses corresponding to the reference repair scenario.
[0080] S304, associate and store each reference maintenance scenario and the corresponding script template to obtain a preset script library.
[0081] The electronic device stores each reference maintenance scenario and the corresponding script template. AI linkage can be established for each script case, i.e. script template, highlighting the combination of emotion and professional technology in intelligent Q&A for vehicle maintenance.
[0082] In one implementation, the electronic device can also respond to user annotation operations on fault types, maintenance anomaly types, maintenance resource requirement types, maintenance scenarios, and fault maintenance responses for historical dialogue content, obtaining the fault type, maintenance anomaly type, maintenance resource requirement type, reference maintenance scenario, and reference fault maintenance response corresponding to each historical dialogue content. The electronic device can also extract historical maintenance content from each historical dialogue content, and based on the fault type, maintenance anomaly type, maintenance resource requirement type, and maintenance scenario corresponding to that historical dialogue content, annotate the historical maintenance content corresponding to that historical dialogue content. This annotated historical maintenance content is used as sample maintenance content to obtain a sample training set for training an AI model of the vehicle's maintenance scenarios. The electronic device can train the AI model based on this sample training set and configure a preset dialogue script library into the trained AI model, thus obtaining a dialogue script model tool.
[0083] It should be noted that when labeling fault types, repair anomaly types, repair resource requirement types, and repair scenarios in historical dialogue content, vehicle fault types can be categorized based on potential faults in various functional systems such as the vehicle's powertrain or electrical systems, or the nature of the vehicle fault itself. For example, fault types can include electronic control system faults, mechanical faults, engine faults, or transmission faults. Repair anomaly types can be categorized based on repair anomalies encountered during vehicle repair. For example, repair anomaly types can include repeated repairs, inability to diagnose, diagnostic failures, or lack of functional support. Resource requirement types can be categorized based on repair needs during the vehicle repair process. For example, resource requirement types can include APP function requirements, diagnostic software requirements, or diagnostic equipment requirements. Repair scenarios can be identified by associating fault types or resource requirement types with customer complaint scenarios. For example, considering the characteristics of the automotive diagnostic industry, if a particular brand of vehicle's 3144 fault is consistently undiagnosed, this can be considered a repair scenario.
[0084] In the above technical solution, reference maintenance scenarios and reference fault maintenance responses are extracted from historical maintenance corpus data. Then, the script templates for the same reference maintenance scenario are extracted from the reference fault maintenance responses and stored in association. This ensures that the script templates are derived from experience summaries of successful actual interactions, rather than subjective designs. It retains the accurate expression for specific scenarios (such as fault type, abnormal maintenance type, or resource requirement type) (such as empathetic language or step guidance), enhances the relevance and reliability of the script templates relative to fault maintenance responses, and provides efficiency for subsequent matching of target script templates through the direct association of "scenario-template".
[0085] In one implementation, the electronic device can also acquire model training data, which includes sample fault repair responses corresponding to multiple sample repair contents and sample dialogue templates corresponding to the sample fault repair responses. During the nth model iteration training process, each sample repair content and its corresponding sample dialogue template are input into the large language model. The large language model is used to infer the predicted fault repair response for each sample repair content based on the sample repair content and its corresponding sample dialogue template, where n is a positive integer. Based on the difference between the sample fault repair response and the predicted fault repair response for each sample repair content, the performance index corresponding to the large language model is determined. The performance index is used to indicate the ability of the large language model to generate fault repair responses that conform to the dialogue template. If the performance index does not meet the preset performance threshold, the model iteration training process proceeds to the (n+1)th iteration. If the performance index meets the preset performance threshold, or the number of iterations of the large language model reaches the preset number, the trained large language model is determined to be a vehicle question-answering language model.
[0086] The sample fault repair responses are not reference responses from historical dialogues corresponding to the sample repair content. Instead, they are vivid fault repair responses generated by the electronic device in response to the user's reorganization of the response based on an extracted dialogue template. The electronic device trains the large language model using sample fault repair responses corresponding to multiple sample repair contents and sample dialogue templates corresponding to these responses. This allows the predicted fault repair responses generated by the large language model to increasingly resemble the vivid sample fault repair responses corresponding to the dialogue templates. This enables fine-tuning of the large language model and allows the final vehicle question-and-answer language model to generate vivid responses under the guidance of dialogue templates, improving the user experience of intelligent question-and-answer for vehicle repair.
[0087] In one implementation, the electronic device can also update a preset dialogue script library according to a preset period; and incrementally update the vehicle question-and-answer language model based on the updated content in the preset dialogue script library. The preset period can be one week or two weeks, etc., which is not specifically limited in this embodiment. Alternatively, the electronic device can update the preset dialogue script library in response to the user's update operation. In this way, when new vehicle faults and repair technologies occur, the preset dialogue script library can be updated in a timely manner, and the model can be incrementally updated, thereby improving the intelligence of the model.
[0088] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 5 As shown, the electronic device 120 of this embodiment includes: at least one processor 60 ( Figure 5(Only one is shown) a processor, a memory 61, and a computer program 62 stored in the memory 61 and executable on the at least one processor 60, wherein the processor 60, when executing the computer program 62, implements any of the above-described components. Steps in the method embodiments.
[0089] The electronic device 120 can be a desktop computer, laptop, handheld computer, cloud server, or other computing device. This electronic device may include, but is not limited to, a processor 60 and a memory 61. Those skilled in the art will understand that... Figure 5 This is merely an example of electronic device 120 and does not constitute a limitation on electronic device 120. It may include more or fewer components than shown, or combine certain components, or different components, such as input / output devices, network access devices, etc.
[0090] The processor 60 may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.
[0091] In some embodiments, the memory 61 may be an internal storage unit of the electronic device 120, such as a hard disk or memory of the electronic device 120. In other embodiments, the memory 61 may be an external storage device of the electronic device 120, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the electronic device 120. Furthermore, the memory 61 may include both internal and external storage units of the electronic device 120. The memory 61 is used to store the operating system, applications, bootloader, data, and other programs, such as the program code of the computer program. The memory 61 can also be used to temporarily store data that has been output or will be output.
[0092] Corresponding to the vehicle intelligent question-answering method described in the above embodiments, Figure 6 The diagram shows a structural block diagram of a vehicle intelligent question-answering device provided in an embodiment of this application. For ease of explanation, only the parts related to the embodiment of this application are shown.
[0093] Reference Figure 6 The device includes: The acquisition module 610 is used to acquire the target maintenance content of the target vehicle. The extraction module 620 is used to extract target fault information and target complaint information of the target vehicle from the target maintenance content. Analysis module 630 is used to analyze the target maintenance scenario of the target vehicle based on the vehicle fault indicated by the target fault information and the maintenance anomaly indicated by the target complaint information. The target maintenance scenario represents the fault type, resource requirement type or maintenance anomaly type that needs to be paid attention to during the maintenance of the target vehicle. The determination module 640 is used to determine the target dialogue template that matches the target maintenance scenario from the preset dialogue script library. The preset dialogue script library includes multiple dialogue templates corresponding to different maintenance scenarios. The dialogue templates include emotional expression prompts that instruct the vehicle question-and-answer language model to generate fault maintenance responses. The question-and-answer module 650 is used to input the target maintenance content and target dialogue template into the vehicle question-and-answer language model to obtain the target fault maintenance response for the target maintenance content.
[0094] In some embodiments, the analysis module is further configured to: determine the target fault type of the target vehicle based on the first correspondence relationship and the vehicle fault indicated by the target fault information, wherein the first correspondence relationship includes multiple fault types and at least one vehicle fault corresponding to each of the multiple fault types; determine the target maintenance anomaly type of the target vehicle based on the second correspondence relationship and the maintenance anomaly indicated by the target complaint information, wherein the second correspondence relationship includes multiple maintenance anomaly types and at least one maintenance anomaly corresponding to each of the multiple maintenance anomaly types; obtain the maintenance resources required for maintaining the vehicle fault of the target vehicle based on the target fault type, the target maintenance anomaly type, and a preset maintenance resource library, thereby obtaining the target resource requirement type; and infer the target maintenance scenario in which the target vehicle is located based on the target fault type, the target maintenance anomaly type, and the target resource requirement type.
[0095] In some embodiments, the determining module is further configured to determine the similarity between the maintenance scenarios corresponding to the multiple dialogue templates in the preset dialogue library and the target maintenance scenario, thereby obtaining the similarity between the multiple dialogue templates; and to determine the dialogue template with the highest corresponding similarity as the target dialogue template.
[0096] In some embodiments, the device further includes a processing module for acquiring historical maintenance corpus data, which includes at least one historical dialogue content during vehicle fault repair; extracting a reference maintenance scenario and a reference fault repair response corresponding to each historical dialogue content from the historical dialogue content of the historical maintenance corpus data; extracting a corresponding script template from the reference fault repair response corresponding to each reference maintenance scenario for each reference maintenance scenario; and storing each reference maintenance scenario and the corresponding script template in association to obtain a preset script library.
[0097] In some embodiments, the processing module is further configured to acquire model training data, which includes sample fault repair responses corresponding to multiple sample repair contents and sample dialogue templates corresponding to the sample fault repair responses; during the nth model iteration training process, each sample repair content and the sample dialogue template corresponding to each sample repair content are input into the large language model, which is used to infer the predicted fault repair response for each sample repair content based on the sample repair content and the sample dialogue template corresponding to the sample repair content, where n is a positive integer; based on the difference between the sample fault repair response and the predicted fault repair response for each sample repair content, the performance index corresponding to the large language model is determined, which is used to indicate the ability of the large language model to generate fault repair responses that conform to the dialogue template; if the performance index does not meet the preset performance threshold, the n+1th model iteration training process is entered; if the performance index meets the preset performance threshold, or the number of iterations of the large language model reaches the preset number, the trained large language model is determined to be a vehicle question-answering language model.
[0098] In some embodiments, the processing module is further configured to update a preset script library according to a preset period; and incrementally update the vehicle question-and-answer language model according to the updated content in the preset script library.
[0099] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.
[0100] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.
[0101] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps described in the various method embodiments above.
[0102] This application provides a computer program product that, when run on a mobile terminal, enables the mobile terminal to implement the steps described in the above-described method embodiments.
[0103] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to a photographing device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.
[0104] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.
[0105] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.
[0106] Furthermore, in the description of this application and the appended claims, the terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.
[0107] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.
[0108] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0109] In the embodiments provided in this application, it should be understood that the disclosed apparatus, computer equipment, and methods can be implemented in other ways. For example, the apparatus and computer equipment embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.
[0110] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.
Claims
1. A vehicle intelligent question-answering method, characterized in that, include: Obtain the target maintenance content for the target vehicle; Extract the target fault information and target complaint information of the target vehicle from the target maintenance content; Based on the vehicle malfunction indicated by the target fault information and the maintenance anomaly indicated by the target complaint information, the target maintenance scenario of the target vehicle is analyzed. The target maintenance scenario represents the type of fault, resource requirement, or maintenance anomaly that needs to be addressed during the maintenance of the target vehicle. A target dialogue template matching the target maintenance scenario is determined from a preset dialogue script library. The preset dialogue script library includes multiple dialogue templates corresponding to different maintenance scenarios. The dialogue templates include emotional expression prompts that instruct the vehicle question-and-answer language model to generate fault maintenance responses. The target repair content and the target dialogue template are input into the vehicle question-and-answer language model to obtain the target fault repair response for the target repair content.
2. The method as described in claim 1, characterized in that, The step of analyzing the maintenance scenario information of the target vehicle based on the vehicle malfunction indicated by the target malfunction information and the maintenance anomaly indicated by the target complaint information includes: Based on the first correspondence and the vehicle fault indicated by the target fault information, the target fault type of the target vehicle is determined. The first correspondence includes multiple fault types and at least one vehicle fault corresponding to each of the multiple fault types. Based on the second correspondence and the maintenance anomaly indicated by the target complaint information, the target maintenance anomaly type of the target vehicle is determined. The second correspondence includes multiple maintenance anomaly types and at least one maintenance anomaly situation corresponding to each of the multiple maintenance anomaly types. Based on the target fault type, the target maintenance anomaly type, and the preset maintenance resource library, obtain the maintenance resources required when repairing the vehicle fault of the target vehicle, and obtain the target resource requirement type. Based on the target fault type, the target maintenance anomaly type, and the target resource requirement type, the target maintenance scenario in which the target vehicle is located is inferred.
3. The method as described in claim 2, characterized in that, The step of determining the target dialogue template that matches the target maintenance scenario from the preset dialogue script library includes: Determine the similarity between the maintenance scenarios corresponding to multiple dialogue templates in the preset dialogue script library and the target maintenance scenario to obtain the similarity between the multiple dialogue templates; The script template with the highest similarity is selected as the target script template.
4. The method according to any one of claims 1 to 3, characterized in that, The method further includes: Acquire historical maintenance corpus data, which includes at least one historical dialogue during the vehicle fault maintenance process; From the historical dialogue content of the historical maintenance corpus data, extract the reference maintenance scenario and reference fault maintenance response corresponding to each historical dialogue content. For each of the reference maintenance scenarios, extract the corresponding script template from the reference fault maintenance responses for the reference maintenance scenario. The preset script library is obtained by associating and storing each reference maintenance scenario and the corresponding script template for each reference maintenance scenario.
5. The method as described in claim 4, characterized in that, The method further includes: Obtain model training data, which includes sample fault repair responses corresponding to multiple sample repair contents, and sample script templates corresponding to the sample fault repair responses. During the nth model iteration training process, each sample maintenance content and the corresponding sample dialogue template are input into the large language model. The large language model is used to infer the predicted fault maintenance response for each sample maintenance content based on the sample maintenance content and the corresponding sample dialogue template, where n is a positive integer. Based on the difference between the sample fault repair response and the predicted fault repair response for each sample repair content, a performance index corresponding to the large language model is determined. The performance index is used to indicate the ability of the large language model to generate fault repair responses that conform to the speech template. If the performance index does not meet the preset performance threshold, proceed to the (n+1)th model iteration training process; If the performance indicators meet the preset performance threshold, or if the number of iterations of the large language model reaches the preset number, the trained large language model is determined to be the vehicle question-answering language model.
6. The method as described in claim 5, characterized in that, The method further includes: The preset script library is updated according to a preset cycle; The vehicle question-and-answer language model is incrementally updated based on the updated content in the preset dialogue library.
7. A vehicle intelligent question-and-answer device, characterized in that, include: The acquisition module is used to acquire the target maintenance content of the target vehicle; The extraction module is used to extract target fault information and target complaint information of the target vehicle from the target maintenance content; The analysis module is used to analyze the target maintenance scenario of the target vehicle based on the vehicle fault indicated by the target fault information and the maintenance anomaly indicated by the target complaint information. The target maintenance scenario represents the fault type, resource requirement type or maintenance anomaly type that needs to be paid attention to during the maintenance of the target vehicle. The determination module is used to determine the target dialogue template that matches the target maintenance scenario from the preset dialogue script library. The preset dialogue script library includes multiple dialogue templates corresponding to different maintenance scenarios. The dialogue templates include emotional expression prompts that instruct the vehicle question-and-answer language model to generate fault maintenance responses. The question-and-answer module is used to input the target maintenance content and the target dialogue template into the vehicle question-and-answer language model to obtain the target fault maintenance response for the target maintenance content.
8. An electronic device, characterized in that, The device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, characterized in that, when the processor executes the computer program, the electronic device performs the method as described in any one of claims 1-6.
9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 6.
10. A computer program product, characterized in that, Includes a computer program, which, when run, causes the method as described in any one of claims 1 to 6 to be performed.