Vehicle information management method and device based on natural language processing, equipment and storage medium

Abstract

The invention discloses a vehicle information management method and device based on natural language processing, equipment and a storage medium, and relates to the technical field of vehicle information management.The method comprises the steps that a query condition dictionary of vehicle information and a natural language instruction input by a user are obtained; according to a preset natural language processing model, intention classification and entity recognition are conducted on the natural language instruction, a target operation type and a target entity set are obtained, and the preset natural language processing model comprises an intention classification model and an entity recognition model; performing deletion completion on the target entity set according to the query condition dictionary to obtain a vehicle information entity set; generating a structured database execution instruction according to the target operation type and the vehicle information entity set; and managing a vehicle information database according to the database execution instruction to obtain a return result corresponding to the database execution instruction, thereby improving the convenience and management efficiency of vehicle information operation.

Description

Technical Field

[0001] This application relates to the technical field of vehicle information management, and in particular to a vehicle information management method, apparatus, device and storage medium based on natural language processing. Background Technology

[0002] In the automotive R&D process, test vehicles, as key carriers for performance testing and intelligent driving function verification, are increasingly numerous and diverse, leading to ever-increasing management complexity. Currently, the management of test vehicles mainly relies on information systems to achieve vehicle scheduling, status tracking, and data traceability. However, existing systems generally adopt a structured input-based design pattern, meaning users must strictly adhere to preset fixed formats or keywords to submit operation commands in order to complete operations such as querying and updating vehicle information. These systems have significant drawbacks in practical use: First, the query method is rigid and lacks error tolerance. Users must accurately memorize complex command syntax and keywords; if keywords are missing or the VIN (Vehicle Identification Number) is incomplete, the system cannot recognize it, resulting in operation failure. Second, multi-condition combined queries are difficult. When users need to filter multiple fields simultaneously (such as department, project, user), all conditions must be entered completely, failing to flexibly respond to the user's true query intent. Furthermore, the data update process is cumbersome and error-prone. Users must manually enter the complete VIN code and precise field labels; any input deviation will lead to update failure, severely impacting management efficiency.

[0003] Therefore, improving the convenience of vehicle information data operation and management efficiency is an urgent problem that needs to be solved. Summary of the Invention

[0004] The main objective of this application is to provide a vehicle information management method, device, equipment, and storage medium based on natural language processing, aiming to solve the technical problem of how to improve the convenience of vehicle information data operation and management efficiency.

[0005] To achieve the above objectives, this application proposes a vehicle information management method based on natural language processing, the method comprising: The dictionary of query criteria for obtaining vehicle information and the natural language commands input by the user; The natural language instructions are classified and entity recognized according to a preset natural language processing model to obtain the target operation type and target entity set. The preset natural language processing model includes an intent classification model and an entity recognition model. The target entity set is filled with missing information according to the query condition dictionary to obtain the vehicle information entity set. Generate structured database execution instructions based on the target operation type and the vehicle information entity set; The vehicle information database is managed according to the database execution instructions, and the return results corresponding to the database execution instructions are obtained.

[0006] In one embodiment, the step of classifying the natural language instruction by intent and recognizing entities according to a preset natural language processing model to obtain the target operation type and the target entity set includes: The natural language instruction is classified according to the intent classification model to obtain the target operation type corresponding to the natural language instruction. The target operation type includes at least one of query, export, update, add, and delete. The contextual features of the natural language instructions are extracted based on the long short-term memory network in the entity recognition model; The context features are labeled and decoded according to a preset conditional random field to obtain an entity label sequence corresponding to the natural language instruction; The target entity set is obtained by extracting the corresponding entity type from the natural language instruction based on the entity label sequence.

[0007] In one embodiment, the step of classifying the natural language instruction according to the intent classification model to obtain the target operation type corresponding to the natural language instruction includes: The natural language instruction is segmented according to the intent classification model to obtain a feature vector composed of target word features, semantic features and instruction length features; The feature vectors are subjected to histogram binning to obtain discretized feature buckets; The splitting gain of the feature bucket is calculated based on the preset gradient sampling rate, and the target probability corresponding to each operation type is determined based on the splitting gain. The target operation type is obtained by filtering the operation types corresponding to the target probability based on a preset confidence threshold.

[0008] In one embodiment, the step of obtaining the vehicle information entity set by performing missing completion on the target entity set based on the query condition dictionary includes: Obtain missing field completion strategies, including wildcard completion strategies, semantic prediction completion strategies, and historical operation completion strategies; The target entity set is matched with the key values ​​in the query condition dictionary to obtain the matched fields and missing fields; The missing field is filled in according to the missing field completion strategy to obtain the completed field; The completed fields and the matched fields are combined to obtain a set of vehicle information entities.

[0009] In one embodiment, the step of completing the missing field according to the missing field completion strategy to obtain the completed field includes: The natural language instruction is concatenated with the target entity set to form an input sequence; The input sequence is semantically understood according to the semantic prediction and completion strategy to obtain the candidate completion value and confidence level corresponding to the missing field; When the confidence level corresponding to the candidate completion value is greater than or equal to a preset confidence threshold, the missing field is completed according to the candidate completion value to obtain the completed field; When the confidence level corresponding to the candidate completion value is less than a preset confidence threshold, the missing field is completed with wildcards according to the wildcard completion strategy to obtain the completed field.

[0010] In one embodiment, the step of completing the missing field according to the missing field completion strategy to obtain the completed field includes: According to the historical operation completion strategy, the user's historical operation records within a preset time period are retrieved from the user's historical operation record database. Based on the historical operation records, the historical usage values ​​and frequency of the user for the missing field under the same operation type are statistically analyzed. Calculate the preference degree of each historical usage value based on the historical usage frequency, and take the historical usage value with the highest preference degree as the candidate completion value; The missing field is filled in according to the candidate completion value to obtain the completed field.

[0011] In one embodiment, the step of managing the vehicle information database according to the database execution instruction and obtaining the return result corresponding to the database execution instruction includes: Execute the database execution instruction to obtain a set of vehicle information records that match the database execution instruction and the number of matching results; When the number of matching results is less than or equal to the preset single batch display threshold, the vehicle information record set will be returned as a result in the form of graphic cards; When the number of matching results exceeds the preset single batch display threshold, the vehicle information record set is written into a table file according to the preset table file generation component, and the download link address of the table file is returned as the result.

[0012] Furthermore, to achieve the above objectives, this application also proposes a vehicle information management device based on natural language processing, the device comprising: The data receiving module is used to obtain a dictionary of query conditions for vehicle information and natural language commands input by the user; The intent parsing module is used to classify the natural language instruction and recognize entities according to a preset natural language processing model to obtain the target operation type and the target entity set. The preset natural language processing model includes an intent classification model and an entity recognition model. The missing optimization module is used to fill in the missing information of the target entity set according to the query condition dictionary to obtain the vehicle information entity set. The instruction generation module is used to generate structured database execution instructions based on the target operation type and the vehicle information entity set. The information management module is used to manage the vehicle information database according to the database execution instructions and obtain the return results corresponding to the database execution instructions.

[0013] In addition, to achieve the above objectives, this application also proposes a vehicle information management device based on natural language processing, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the vehicle information management method based on natural language processing as described above.

[0014] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the vehicle information management method based on natural language processing as described above.

[0015] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the vehicle information management method based on natural language processing as described above.

[0016] This application provides a vehicle information management method based on natural language processing. The method includes: acquiring a query condition dictionary for vehicle information and natural language commands input by the user; classifying the natural language commands by intent and recognizing entities according to a preset natural language processing model, obtaining a target operation type and a target entity set, wherein the preset natural language processing model includes an intent classification model and an entity recognition model; performing missing completion on the target entity set according to the query condition dictionary to obtain a vehicle information entity set; generating structured database execution commands based on the target operation type and the vehicle information entity set; and managing the vehicle information database according to the database execution commands to obtain return results corresponding to the database execution commands. In summary, this application, by constructing a collaborative architecture of intent parsing and dynamic database query command generation, and employing an entity recognition method that integrates pre-trained models and domain knowledge, achieves automatic conversion from colloquial commands to precise database operations, realizing efficient and flexible secure management of experimental vehicle information, and improving the convenience and management efficiency of vehicle information operations. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating the first embodiment of the vehicle information management method based on natural language processing in this application; Figure 2 This is a schematic diagram of the architecture of the vehicle information management system based on natural language processing in this application; Figure 3 This is a schematic diagram of the natural language instruction processing flow in one embodiment of the vehicle information management method based on natural language processing in this application; Figure 4 This is a flowchart illustrating the second embodiment of the vehicle information management method based on natural language processing in this application. Figure 5 This is a flowchart illustrating the third embodiment of the vehicle information management method based on natural language processing in this application. Figure 6 This is a schematic diagram of the module structure of the vehicle information management device based on natural language processing according to an embodiment of this application; Figure 7 This is a schematic diagram of the device structure of the hardware operating environment involved in the vehicle information management method based on natural language processing in the embodiments of this application.

[0020] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0021] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0022] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0023] In the automotive R&D process, test vehicles, as key carriers for performance testing and intelligent driving function verification, are increasingly numerous and diverse, leading to ever-increasing management complexity. Currently, the management of test vehicles mainly relies on information systems to achieve vehicle scheduling, status tracking, and data traceability. However, existing systems generally adopt a structured input-based design pattern, meaning users must strictly follow preset fixed formats or keywords to submit operation commands in order to complete operations such as querying and updating vehicle information. These systems have significant drawbacks in practical use: First, the query method is rigid and lacks fault tolerance; users must accurately memorize complex command syntax and keywords. If keywords are missing or the VIN code is incomplete, the system cannot recognize them, leading to operation failure. Second, multi-condition combined queries are difficult; when users need to filter multiple fields simultaneously (such as department, project, user), all conditions must be entered completely, making it impossible to flexibly respond to the user's true query intent. Furthermore, the data update process is cumbersome and error-prone; users must manually enter the complete VIN code and accurate field labels, and any input deviation will lead to update failure, severely impacting management efficiency. Therefore, how to improve the convenience of data operations while ensuring data security is an urgent problem that needs to be solved.

[0024] It should be noted that the executing entity of this embodiment can be a vehicle information management system based on natural language processing, or a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device capable of implementing the aforementioned vehicle information management functions based on natural language processing. This embodiment does not specifically limit it in this way. The following uses a vehicle information management system based on natural language processing as an example to describe this embodiment and the following embodiments.

[0025] Based on this, embodiments of this application provide a vehicle information management method based on natural language processing, referring to... Figure 1 , Figure 1This is a flowchart illustrating the first embodiment of the vehicle information management method based on natural language processing of this application.

[0026] In this embodiment, the vehicle information management method based on natural language processing includes steps S10 to S50: Step S10: Obtain the dictionary of query conditions for vehicle information and the natural language commands input by the user.

[0027] It should be noted that the system initializes a preset query condition dictionary, which defines all queryable fields for test vehicle information management (such as VIN, user, department, project name, vehicle status, purpose, last maintenance time, parking location, etc.), serving as the basis for subsequent entity mapping and completion. Figure 2 As shown, users input natural language commands in any format through interactive terminals (such as the Lark client or web management terminal), such as "Export Zhang San's car from the Intelligent Driving Network Test Section" or "Search for the test vehicle with SL99588". The system receives the command text and uses it, along with a dictionary of search criteria, as input for subsequent processing. Natural language commands refer to user requests expressed in everyday conversational language without adhering to specific grammatical formats. Their core characteristic is unstructured nature, and they can contain ambiguous information (such as partial VINs or abbreviated names) and flexible sentence structures.

[0028] Step S20: Perform intent classification and entity recognition on the natural language instruction according to the preset natural language processing model to obtain the target operation type and target entity set. The preset natural language processing model includes an intent classification model and an entity recognition model.

[0029] It should be noted that, as Figure 2 As shown, in this step, the system sends the received natural language instructions to a pre-defined natural language processing model for parsing. This process consists of two sub-steps: First, the semantics of the instructions are analyzed using an intent classification model (such as the LightGBM classifier), classifying them into a predefined target operation type such as "query," "update," "export," or "delete." Second, the instructions are sequence-labeled using an entity recognition model (such as the Bi-LSTM+CRF model), extracting key information fragments related to vehicle management as a set of target entities, such as VIN fragments, personnel names, department names, field names to be modified, and their new values. It can be understood that the purpose of this step is to initially transform unstructured natural language instructions into structured operational intents and key parameters.

[0030] Additionally, it should be noted that the pre-defined natural language processing models include an intent classification model and an entity recognition model. These are machine learning models pre-trained using a labeled test vehicle domain instruction dataset, possessing high-accuracy understanding capabilities tailored to the language characteristics of this domain. During the training phase (offline) of the intent classification model, the system constructs a test vehicle instruction annotation dataset. A large number of labeled test vehicle domain natural language instruction samples are collected (e.g., "Export the F710C project vehicle of Zhang San in the Intelligent Driving Network Test Section"), and entities are labeled using the BIO annotation system. During model training and optimization, the cross-entropy loss function (negative log-likelihood loss of the CRF layer) is used as the optimization objective, employing the Adam optimizer (with an initial learning rate set to 1e-3, decaying to 0.8 of the previous round every 10 rounds). After 50 rounds of training, the entity recognition accuracy is stabilized at ≥95%. The natural language processing model can also be replaced by a BERT+Softmax model. Semantic features are extracted using a pre-trained BERT model, and entity recognition is achieved through Softmax classification, suitable for environments with more abundant computing power.

[0031] Step S30: Complete the missing data of the target entity set according to the query condition dictionary to obtain the vehicle information entity set.

[0032] It's important to note that in this step, the system maps the extracted target entity set to a query condition dictionary. Each preset field (key) in the dictionary attempts to match an entity in the target entity set. If a preset field (such as "project name") does not find a corresponding value in the target entity set, the system marks that field as "missing." For missing fields, the system performs autocomplete logic, for example, using the wildcard "%" as a default value, or combining context-based deep learning models (such as BERT) or user historical behavior data for intelligent prediction, thereby generating a complete and structured set of vehicle information entities.

[0033] Additionally, it should be noted that the query condition dictionary is a predefined key-value pair data structure. Its keys are all the core query dimension fields for test vehicle information management, and the values ​​are initially empty or default values. This structure is used to systematically organize the scattered entity information extracted from natural language. The vehicle information entity set is a structured data collection containing all the completed condition values ​​that correspond one-to-one with the fields in the query condition dictionary, used to precisely construct database queries or operation statements.

[0034] Step S40: Generate structured database execution instructions based on the target operation type and the vehicle information entity set.

[0035] It should be noted that, as Figure 2As shown, database execution instructions refer to standard Structured Query Language (SQL) statements containing conditions, such as SELECT and UPDATE. These are the final command forms used by the system to interact with the underlying vehicle information database. In this step, the system automatically constructs SQL instructions that can be directly understood and executed by the database system based on the determined target operation type (e.g., "export") and the generated complete set of vehicle information entities. For example, for an export operation, the system converts the conditions in the entity set into SQL WHERE clauses; for an update operation, it generates an UPDATE SET statement. This process is completed by the system's SQL conversion unit, ensuring seamless conversion from natural language to machine instructions.

[0036] Additionally, it should be noted that, as Figure 3 As shown, after completing instruction parsing, the system checks the parsed target operation type. Based on a set of logical rules binding user categories to their executable operation scope, add, delete, and update (or "modify") operations are identified as high-privilege operations. Once an operation type is determined to be a high-privilege operation, the system immediately triggers the permission verification process. Specifically, the system automatically extracts the unique identifier of the user initiating the instruction (e.g., the user's "Lark account ID") from the current session or user login context, and then matches the user's role information (i.e., permissions) with the user's unique identifier by calling the enterprise's existing permission management system. If the user role is "Experimental Vehicle Administrator" or another role authorized to perform high-privilege operations, the permission verification passes. The system allows the process to continue, proceeding to the subsequent SQL generation and execution steps; if the user role is "Ordinary Employee" or an unauthorized role, the permission verification fails. The system immediately interrupts the process, does not perform any database operations, and returns a clear rejection message to the user, such as "No operation permission, please contact the administrator."

[0037] Understandably, the purpose of this step is to implement security grading for data operations, ensuring that only users with the appropriate permissions can perform specific operations, thereby improving operational convenience while protecting system data security.

[0038] Step S50: Manage the vehicle information database according to the database execution instruction, and obtain the return result corresponding to the database execution instruction.

[0039] It should be noted that in this step, the system sends the generated data execution instructions to the vehicle information database for execution. After execution, the system generates corresponding return results based on the operation type and the size of the result. For example, executing a query and returning matching vehicle records. It can be understood that the purpose of this step is to execute different processing steps based on the type of query operation and the amount of result data.

[0040] In one feasible implementation, step S50 specifically includes: Step S501: Execute the database execution instruction to obtain the set of vehicle information records that match the database execution instruction and the number of matching results.

[0041] It should be noted that in this step, the system executes generated SQL commands through the database engine to retrieve or modify data from the vehicle information database. After execution, the system obtains a set of vehicle information records consisting of all vehicle records that meet the criteria, and counts the number of records in this set, i.e., the number of matching results.

[0042] Step S502: When the number of matching results is less than or equal to the preset single batch display threshold, the vehicle information record set is returned as a graphic card.

[0043] It should be noted that the system has a preset threshold for displaying a single batch (e.g., 3 entries). If the number of matching results does not exceed this threshold, the result volume is considered small and suitable for direct reading. The system will format the key fields (such as VIN, vehicle model, user, parking location, etc.) in the vehicle information record set and display them directly to the user in the form of graphic card messages on the interactive terminal (such as Lark chat window), making the interaction intuitive and efficient.

[0044] Step S503: When the number of matching results is greater than the preset single batch display threshold, the vehicle information record set is written into a table file according to the preset table file generation component, and the download link address of the table file is returned as the result.

[0045] It should be noted that if the number of matching results exceeds the preset threshold for displaying a single batch, the system considers the data volume too large, and direct display would be information overload. The system will automatically call a preset table file generation component (such as Python's openpyxl library) to write the complete set of vehicle information records into a table file (such as an Excel file) according to the preset field order. After generating the file, the system will not directly display all the data on the interface, but instead will generate a download link for that file and return it to the user. The user can click the link to download the complete list file for archiving, sharing, or further analysis.

[0046] This embodiment provides a vehicle information management method based on natural language processing. The method includes: acquiring a query condition dictionary for vehicle information and natural language commands input by the user; classifying the natural language commands by intent and recognizing entities according to a preset natural language processing model, obtaining a target operation type and a target entity set, wherein the preset natural language processing model includes an intent classification model and an entity recognition model; performing missing completion on the target entity set according to the query condition dictionary to obtain a vehicle information entity set; generating structured database execution commands based on the target operation type and the vehicle information entity set; and managing the vehicle information database according to the database execution commands to obtain return results corresponding to the database execution commands. In summary, this embodiment, by constructing a collaborative architecture for intent parsing and dynamic database query command generation, and employing an entity recognition method that integrates pre-trained models and domain knowledge, achieves automatic conversion from colloquial commands to precise database operations, realizing efficient and flexible secure management of experimental vehicle information, and improving the convenience and management efficiency of vehicle information operations.

[0047] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in Embodiment 1 above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 4 , Figure 4 This is a flowchart illustrating the second embodiment of the vehicle information management method based on natural language processing of this application. Step S20 specifically includes: Step S201: Classify the natural language instruction according to the intent classification model to obtain the target operation type corresponding to the natural language instruction. The target operation type includes at least one of query, export, update, add, and delete.

[0048] It's important to note that in this step, the system inputs the received natural language command into a pre-trained intent classification model (such as the LightGBM model). This model analyzes the command, outputting a probability distribution of whether the command belongs to a preset operation type (query, export, update, add, delete), and ultimately determines the type with the highest probability as the target operation type. For example, for the command "Export Zhang San's car," the model outputs that the "export" operation has the highest probability, thus classifying it as the target operation type.

[0049] In one feasible implementation, step S201 specifically includes: Step A10: Perform word segmentation on the natural language instruction according to the intent classification model to obtain a feature vector composed of target word features, semantic features, and instruction length features.

[0050] It should be noted that in this step, the system first preprocesses the natural language instructions, including word segmentation and feature extraction. Specifically, three types of features are extracted: (1) Target word features: identify whether there are preset operation keywords (such as "search", "guide", "modify", "delete", etc.) in the instructions, and encode them one-hot; (2) Semantic features: use other semantic understanding modules (such as Bi-LSTM network) to process the whole instruction and generate a fixed-dimensional semantic vector to represent the global semantics of the instruction; (3) Instruction length features: calculate the number of characters or words in the instruction and normalize them (normalize to the 0-1 range). Finally, these three types of features are concatenated into a comprehensive feature vector, which is used as the input of the intent classification model.

[0051] Additionally, it should be noted that target word features primarily capture explicit operation indicators in the instructions; semantic features are used to understand the deeper meaning and context of the instructions, compensating for the inadequacy of keywords alone; and instruction length features, as an auxiliary feature, help the model distinguish between short queries and complex descriptions.

[0052] Step A20: Perform histogram binning on the feature vector to obtain discretized feature buckets.

[0053] It should be noted that for the high-dimensional, continuous semantic features in the feature vector, the system employs histogram binning. This method divides the value range of each continuous feature into several intervals (i.e., buckets), and statistically analyzes the gradient information of samples falling into each bucket, thereby transforming the continuous features into a discrete histogram representation. This process significantly reduces computational load and memory consumption.

[0054] Understandably, histogram binning is an optimization technique in gradient boosting decision tree (GBDT) algorithms (such as LightGBM) used to accelerate the search for feature split points. Feature buckets refer to each discrete interval after partitioning and its corresponding statistical information (such as the sum of gradients and the number of samples).

[0055] Step A30: Calculate the splitting gain of the feature bucket based on the preset gradient sampling rate, and determine the target probability corresponding to each operation type based on the splitting gain.

[0056] It should be noted that in order to further improve the training efficiency, when the system calculates the feature split gain, it will adopt the GOSS (Gradient-based One-Side Sampling) strategy. Specifically, the system will set a preset gradient sampling rate (such as 0.2), and only partially sample the samples with smaller absolute gradient values, while retaining all the samples with large gradients (usually the samples that are difficult to classify). Based on the sampled data, the system will calculate the split gain of each feature bucket by the method of mean square error (used to measure the reduction of the model loss by splitting according to this feature), and iteratively construct multiple decision trees by gradient boosting. Finally, for the input instruction feature vector, the model will traverse all the decision trees and comprehensively calculate the target probability that the instruction belongs to each operation type (query, export, etc.).

[0057] Step A40: Screen the operation type corresponding to the target probability according to the preset confidence threshold to obtain the target operation type.

[0058] It should be noted that after the system obtains the target probability of each operation type, it will take the type with the highest probability as the preliminary determination result. To further ensure the accuracy of classification, the system sets a preset confidence threshold (such as 0.8). If the highest probability value in the target probability is greater than or equal to this threshold, the operation type corresponding to it will be directly determined as the final target operation type. If the highest probability value in the target probability is less than this threshold (indicating that the model has low confidence in classifying the current instruction and there may be ambiguity), the secondary determination logic will be triggered, for example, combining the results of entity recognition for auxiliary correction, so as to improve the robustness of intent classification.

[0059] Step S202: Extract the context features of the natural language instruction according to the long short-term memory network in the entity recognition model.

[0060] It should be noted that the system will input the tokenized natural language instruction sequence into the entity recognition model. One of the core components of this model is the long short-term memory network, specifically a bidirectional LSTM. Bi-LSTM can read the instruction sequence from both the forward and backward directions, so as to fully capture the dependencies between each word and its context, and generate a vector representation containing rich context features. For example, for the word "SL99588", its vector representation will fuse the semantic information of the previous word "query" and the subsequent word "test vehicle".

[0061] Step S203: Decode the labels of the context features according to the preset conditional random field to obtain the entity label sequence corresponding to the natural language instruction.

[0062] It should be noted that after obtaining the context features of each word, the system will input them into a preset conditional random field layer (CRF layer). The CRF layer can learn the transition rules between different entity tags (for example, the B-VIN tag is usually followed by I-VIN, rather than directly followed by B-department), so as to model the global dependencies of the tag sequence. The CRF layer will receive the features from the previous layer, decode the globally optimal entity tag sequence through the Viterbi algorithm, and filter out invalid tag combinations (such as sequences where there is no "I-field value" after "B-field name").

[0063] It can be understood that the preset conditional random field is a discriminative probability model, commonly used in sequence labeling tasks, which improves the overall consistency of labeling by considering the relationship between adjacent tags. The entity tag sequence refers to the sequence formed by assigning a tag (such as "B-VIN", "I-VIN", "O", etc.) to each word in the input instruction, which indicates the entity attribution of each word.

[0064] Step S204: Extract the corresponding entity types from the natural language instruction according to the entity tag sequence, and obtain the target entity set.

[0065] It should be noted that in this step, the system will parse the decoded entity tag sequence. Merge the continuous word segments with the same entity type tags (such as B-VIN and subsequent I-VIN) to restore the complete entity value. For example, if the tag sequence [O, B-VIN, I-VIN, O] corresponds to the word sequence [Check, SL995, 88, test vehicle], then the entity "VIN=SL99588" can be extracted. All the extracted entities (such as VIN, person name, department name, field name and its value, etc.) constitute the target entity set.

[0066] In this embodiment, through the collaborative processing of intent classification and entity recognition, the unstructured natural language instruction input by the user is automatically and accurately parsed into the target operation type and the structured entity set, solving the problems of the traditional system that needs to strictly follow a fixed instruction format and has low fault tolerance, realizing the intelligent understanding of colloquial and fuzzy instructions, and improving the convenience and success rate of operations such as querying and updating test vehicle information.

[0067] Based on the first and second embodiments of the present application, in the third embodiment of the present application, the same or similar content as in the above-mentioned first and second embodiments can be referred to the above introduction and will not be repeated hereinafter. On this basis, please refer to Figure 5 , Figure 5 is the flowchart of the third embodiment of the vehicle information management method based on natural language processing of the present application. The step S30 specifically includes: Step S301: Obtain missing field completion strategies, including wildcard completion strategies, semantic prediction completion strategies, and historical operation completion strategies.

[0068] It should be noted that in this step, the system will pre-configure and invoke at least three completion strategies to address field missing issues in different scenarios. The wildcard completion strategy serves as a basic backup solution, using preset values ​​to fill in fields that cannot be extracted from the user's command. The semantic prediction completion strategy is used to understand the context of the current command to intelligently predict the values ​​that need to be filled. The historical operation completion strategy provides personalized completion based on the user's past behavior habits.

[0069] Step S302: Match the target entity set with the key values ​​in the query condition dictionary to obtain the matched fields and missing fields.

[0070] It should be noted that a matched field refers to a pre-defined query field whose specific value has been successfully identified and extracted from the natural language command. A missing field refers to a field that exists in the pre-defined query condition dictionary but whose specific value has not been identified from the current natural language command. In this step, the system compares the target entity set extracted in the previous step (e.g., {User: Zhang San, Department: Intelligent Driving Network Testing Department}) with the keys of the predefined query condition dictionary (pre-defined field list, e.g., [VIN, User, Department, Project Name, ...]). If a field has a corresponding key value in the target entity set, it is marked as a matched field; otherwise, it is marked as a missing field. For example, if the comparison finds that the "Project Name" field is not in the target entity set, then "Project Name" is a missing field.

[0071] Step S303: Complete the missing field according to the missing field completion strategy to obtain the completed field.

[0072] It's important to note that in this step, the system selects an appropriate strategy to generate or determine a value for each missing field based on preset priority rules (such as prioritizing historical operation completion, followed by semantic prediction completion, and finally wildcard completion) and the specific scenario. This transforms the missing fields into completed fields. Understandably, the purpose of this step is to address the issue of incomplete query conditions, enabling the system to automatically generate a complete set of query conditions even if the user's command is incomplete.

[0073] In one feasible implementation, step S303 specifically includes: Step B10: Concatenate the natural language instruction with the target entity set to form an input sequence.

[0074] It should be noted that the input sequence is a text sequence constructed to adapt to the input format of a pre-trained language model (such as BERT), and usually contains special delimiters (such as [CLS], [SEP]) to distinguish different parts. Specifically, to achieve context-based semantic prediction, the model's input needs to be constructed. Specifically, the user's original command text is normalized and concatenated with the extracted target entity set (in the format of "field name=value"). For example, the input sequence can be constructed as: "[CLS]Export Intelligent Driving and Connected Vehicle Testing Department Zhang San's Car [SEP]Department = Intelligent Driving and Connected Vehicle Testing Department, User = Zhang San [SEP]".

[0075] Step B20: Perform semantic understanding on the input sequence according to the semantic prediction completion strategy to obtain the candidate completion value and confidence level corresponding to the missing field.

[0076] It's important to note that in this step, the system invokes a natural language understanding model (such as BERT) fine-tuned based on a semantic prediction completion strategy. The input sequence is fed into the model, which predicts the most likely value for a specified missing field (such as the project name) based on the semantics of the entire sentence and the known entity context. The model outputs multiple candidate completion values ​​for that field (such as "F710C") and their corresponding confidence scores (such as 0.85). In essence, this step leverages a deep learning model to understand the user's underlying intent and intelligently infer missing information.

[0077] Step B30: When the confidence level corresponding to the candidate completion value is greater than or equal to a preset confidence threshold, the missing field is completed according to the candidate completion value to obtain the completed field.

[0078] It should be noted that the system sets a preset confidence threshold (e.g., 0.8). If the confidence of the optimal candidate completion value output by the model reaches or exceeds this threshold, the prediction result is considered highly reliable. The system then directly uses this predicted value (e.g., "F710C") as the completion value for the missing field to complete the completion. Understandably, the purpose of this step is to ensure that intelligent completion is only used when the prediction is highly reliable, thus avoiding the introduction of errors.

[0079] Step B40: When the confidence level corresponding to the candidate completion value is less than the preset confidence threshold, the missing field is completed into a wildcard according to the wildcard completion strategy to obtain the completed field.

[0080] It's important to note that if the confidence levels of all candidate completion values ​​are below the preset confidence threshold, it indicates that the current instruction context is insufficient to make a reliable prediction. In this case, the system employs a fallback wildcard completion strategy, setting the completion value of the missing field to a wildcard character (such as "%") used in the database query. For example, the missing "Project Name" field would be completed as "LIKE '%'". This step ensures that the system can generate valid query conditions under any circumstances, preventing the process from being interrupted due to prediction uncertainty.

[0081] In one feasible implementation, step S303 further includes: Step C10: Obtain the user's historical operation records within a preset time period from the user's historical operation record database according to the historical operation completion strategy.

[0082] It should be noted that the historical operation completion strategy refers to a strategy that infers a user's operating habits and preferences by analyzing their past operation records, thereby completing missing fields. Specifically, for users with accumulated operation records, the system will prioritize trying the historical operation completion strategy. The system will query the log database for all vehicle management-related operation records within a preset time period (such as the past 3 months) based on the user's identity identifier that initiated the current command.

[0083] Step C20: Based on the historical operation records, calculate the user's historical usage value and historical usage frequency for the missing field under the same operation type.

[0084] It's important to note that in this step, the system filters and analyzes the retrieved historical operation records. First, it filters out records with the same operation type as the current command (e.g., "Export"). Then, for the currently missing field (e.g., "Project Name"), it counts the specific historical usage values ​​(e.g., "F710C", "F720") and their frequency of occurrence within these similar historical records, i.e., historical usage frequency. This step essentially quantifies the user's preference for different values ​​under specific conditions.

[0085] Step C30: Calculate the preference degree of each historical usage value based on the historical usage frequency, and take the historical usage value with the highest preference degree as the candidate completion value.

[0086] It's important to note that preference score is a numerical value calculated by an algorithm, used to comprehensively measure the strength of a user's preference for a particular field's value; a higher value indicates a stronger preference. In this step, the system quantifies the user's preference for each historical usage value using a weighted formula based on historical usage frequency and group behavior (such as the usage rate of users in a department). For example, Preference Score = (Individual Usage Frequency / Total Operations) * 0.7 + (Department Usage Rate) * 0.3. After calculation, the system selects the historical usage value with the highest preference score as the candidate value for this completion. In essence, this step aims to combine individual and group behavior to select the completion value that best aligns with the user's habits.

[0087] Step C40: Complete the missing field according to the candidate completion value to obtain the completed field.

[0088] It's important to note that in this step, the system directly uses the candidate completion value with the highest calculated preference (such as "F710C") as the completion value for the missing field, thus completing the completion. Compared to semantic prediction, this completion value is derived from the user's actual historical behavior and usually has higher contextual relevance. Understandably, the purpose of this step is to improve the accuracy of missing field completion based on user habits.

[0089] Step S304: Combine the completed fields and the matched fields to obtain a vehicle information entity set.

[0090] It should be noted that after the completion process, all missing fields have been converted into completed fields. The system merges these completed fields (e.g., {Project Name: F710C, Vehicle Status: %}) with the obtained matched fields (e.g., {User: Zhang San, Department: Intelligent Driving and Connected Vehicle Testing Department}) to form a complete set of vehicle information entities covering all preset query dimensions. In essence, the vehicle information entity set is a structured dictionary or object, where the keys are the preset vehicle information field names, and the values ​​are specific condition values ​​or wildcards extracted from user commands or obtained through strategy completion. It fully defines all the filtering conditions required for this database operation.

[0091] In this embodiment, by matching the entities extracted from user instructions with preset fields, and comprehensively using three completion strategies—wildcards, semantic prediction, and user historical behavior—to intelligently fill in missing fields, the problem of traditional test vehicle management systems being unable to perform operations or having inaccurate queries due to incomplete query conditions is solved. This realizes the intelligence and flexibility of multi-condition combined queries, thereby improving the query success rate and ease of operation.

[0092] This application also provides a vehicle information management device based on natural language processing; please refer to... Figure 6 The vehicle information management device based on natural language processing includes: The data receiving module 10 is used to obtain a dictionary of query conditions for vehicle information and natural language commands input by the user; The intent parsing module 20 is used to classify the natural language instructions and recognize entities according to a preset natural language processing model to obtain the target operation type and the target entity set. The preset natural language processing model includes an intent classification model and an entity recognition model. Missing Optimization Module 30 is used to fill in the missing information of the target entity set according to the query condition dictionary to obtain the vehicle information entity set; Instruction generation module 40 is used to generate structured database execution instructions based on the target operation type and the vehicle information entity set; The information management module 50 is used to manage the vehicle information database according to the database execution instructions and obtain the return results corresponding to the database execution instructions.

[0093] The vehicle information management device based on natural language processing provided in this application, employing the vehicle information management method based on natural language processing in the above embodiments, can solve the technical problem of how to improve the convenience of vehicle information data operation and management efficiency. Compared with the prior art, the beneficial effects of the vehicle information management device based on natural language processing provided in this application are the same as the beneficial effects of the vehicle information management method based on natural language processing provided in the above embodiments, and other technical features in the vehicle information management device based on natural language processing are the same as the features disclosed in the methods of the above embodiments, and will not be repeated here.

[0094] In one embodiment, the intent parsing module 20 is further configured to classify the natural language instruction according to an intent classification model to obtain a target operation type corresponding to the natural language instruction, wherein the target operation type includes at least one of query, export, update, add, and delete; extract context features of the natural language instruction according to a long short-term memory network in an entity recognition model; perform label decoding on the context features according to a preset conditional random field to obtain an entity label sequence corresponding to the natural language instruction; and extract the corresponding entity type from the natural language instruction according to the entity label sequence to obtain a target entity set.

[0095] In one embodiment, the intent parsing module 20 is further configured to perform word segmentation on the natural language instruction according to the intent classification model to obtain a feature vector composed of target word features, semantic features, and instruction length features; perform histogram binning on the feature vector to obtain discretized feature buckets; calculate the split gain of the feature buckets according to a preset gradient sampling rate, and determine the target probability corresponding to each operation type according to the split gain; and filter the operation types corresponding to the target probabilities according to a preset confidence threshold to obtain the target operation type.

[0096] In one embodiment, the missing field optimization module 30 is further configured to obtain a missing field completion strategy, which includes a wildcard completion strategy, a semantic prediction completion strategy, and a historical operation completion strategy; match the target entity set with the key values ​​in the query condition dictionary to obtain matched fields and missing fields; complete the missing fields according to the missing field completion strategy to obtain completed fields; and combine the completed fields and the matched fields to obtain a vehicle information entity set.

[0097] In one embodiment, the missing data optimization module 30 is further configured to concatenate the natural language instruction with the target entity set to form an input sequence; perform semantic understanding on the input sequence according to the semantic prediction completion strategy to obtain candidate completion values ​​and confidence levels corresponding to the missing fields; when the confidence level corresponding to the candidate completion value is greater than or equal to a preset confidence threshold, complete the missing fields according to the candidate completion values ​​to obtain completed fields; when the confidence level corresponding to the candidate completion value is less than the preset confidence threshold, complete the missing fields with wildcards according to the wildcard completion strategy to obtain completed fields.

[0098] In one embodiment, the missing field optimization module 30 is further configured to: obtain the user's historical operation records within a preset time period from the user's historical operation record database according to the historical operation completion strategy; statistically analyze the user's historical usage values ​​and historical usage frequencies for the missing field under the same operation type based on the historical operation records; calculate the preference degree of each historical usage value based on the historical usage frequency, and take the historical usage value with the highest preference degree as a candidate completion value; and complete the missing field based on the candidate completion value to obtain the completed field.

[0099] In one embodiment, the information management module 50 is further configured to execute the database execution instruction to obtain a set of vehicle information records matching the database execution instruction and a number of matching results; when the number of matching results is less than or equal to a preset single-batch display threshold, the set of vehicle information records is returned as a graphic card; when the number of matching results is greater than the preset single-batch display threshold, the set of vehicle information records is written into a table file according to a preset table file generation component, and the download link address of the table file is returned as a result.

[0100] This application provides a vehicle information management device based on natural language processing. The vehicle information management device based on natural language processing includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the vehicle information management method based on natural language processing in the above embodiment 1.

[0101] The following is for reference. Figure 7 This document illustrates a structural schematic diagram of a vehicle information management device based on natural language processing suitable for implementing embodiments of this application. The vehicle information management device based on natural language processing in these embodiments may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 7 The vehicle information management device based on natural language processing shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0102] like Figure 7As shown, a vehicle information management device based on natural language processing may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to programs stored in ROM (Read Only Memory) 1002 or programs loaded from storage device 1003 into RAM (Random Access Memory) 1004. RAM 1004 also stores various programs and data required for the operation of the vehicle information management device based on natural language processing. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via bus 1005. Input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, LCDs (Liquid Crystal Displays), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the natural language processing-based vehicle information management device to exchange data wirelessly or via wired communication with other devices. Although a natural language processing-based vehicle information management device with various systems is shown in the figure, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems may be implemented alternatively.

[0103] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0104] The vehicle information management device based on natural language processing provided in this application, employing the vehicle information management method based on natural language processing in the above embodiments, can solve the technical problem of how to improve the convenience of vehicle information data operation and management efficiency. Compared with the prior art, the beneficial effects of the vehicle information management device based on natural language processing provided in this application are the same as the beneficial effects of the vehicle information management method based on natural language processing provided in the above embodiments, and other technical features in this vehicle information management device based on natural language processing are the same as those disclosed in the method of the previous embodiment, and will not be repeated here.

[0105] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0106] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0107] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the vehicle information management method based on natural language processing in the above embodiments.

[0108] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, RAM (Random Access Memory), ROM (Read Only Memory), EPROM (Erasable Programmable Read Only Memory or Flash Memory), optical fibers, CD-ROM (CD-Read Only Memory), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0109] The aforementioned computer-readable storage medium may be included in a vehicle information management device based on natural language processing; or it may exist independently and not be installed in a vehicle information management device based on natural language processing.

[0110] The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by a vehicle information management device based on natural language processing, the device performs the following actions: acquires a dictionary of query conditions for vehicle information and natural language commands input by the user; performs intent classification and entity recognition on the natural language commands according to a preset natural language processing model to obtain a target operation type and a target entity set, wherein the preset natural language processing model includes an intent classification model and an entity recognition model; performs missing completion on the target entity set according to the query condition dictionary to obtain a vehicle information entity set; generates a structured database execution command according to the target operation type and the vehicle information entity set; and manages the vehicle information database according to the database execution command to obtain a return result corresponding to the database execution command.

[0111] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including LAN (Local Area Network) or WAN (Wide Area Network)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0112] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0113] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0114] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., computer programs) for executing the above-described vehicle information management method based on natural language processing. This addresses the technical problem of improving the convenience and efficiency of vehicle information data operations and management. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the vehicle information management method based on natural language processing provided in the above embodiments, and will not be elaborated upon here.

[0115] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the vehicle information management method based on natural language processing as described above.

[0116] The computer program product provided in this application can solve the technical problem of how to improve the convenience of vehicle information data operation and management efficiency. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the vehicle information management method based on natural language processing provided in the above embodiments, and will not be repeated here.

[0117] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A vehicle information management method based on natural language processing, characterized in that, The method includes: The dictionary of query criteria for obtaining vehicle information and the natural language commands input by the user; The natural language instructions are classified and entity recognized according to a preset natural language processing model to obtain the target operation type and target entity set. The preset natural language processing model includes an intent classification model and an entity recognition model. The target entity set is filled with missing information according to the query condition dictionary to obtain the vehicle information entity set. Generate structured database execution instructions based on the target operation type and the vehicle information entity set; The vehicle information database is managed according to the database execution instructions, and the return results corresponding to the database execution instructions are obtained.

2. The method as described in claim 1, characterized in that, The step of classifying the natural language instruction according to a preset natural language processing model to obtain the target operation type and the target entity set includes: The natural language instruction is classified according to the intent classification model to obtain the target operation type corresponding to the natural language instruction. The target operation type includes at least one of query, export, update, add, and delete. The contextual features of the natural language instructions are extracted based on the long short-term memory network in the entity recognition model; The context features are labeled and decoded according to a preset conditional random field to obtain an entity label sequence corresponding to the natural language instruction; The target entity set is obtained by extracting the corresponding entity type from the natural language instruction based on the entity label sequence.

3. The method as described in claim 2, characterized in that, The step of classifying the natural language instruction according to the intent classification model to obtain the target operation type corresponding to the natural language instruction includes: The natural language instruction is segmented according to the intent classification model to obtain a feature vector composed of target word features, semantic features and instruction length features; The feature vectors are subjected to histogram binning to obtain discretized feature buckets; The splitting gain of the feature bucket is calculated based on the preset gradient sampling rate, and the target probability corresponding to each operation type is determined based on the splitting gain. The target operation type is obtained by filtering the operation types corresponding to the target probability based on a preset confidence threshold.

4. The method as described in claim 1, characterized in that, The step of obtaining the vehicle information entity set by performing missing data completion on the target entity set based on the query condition dictionary includes: Obtain missing field completion strategies, including wildcard completion strategies, semantic prediction completion strategies, and historical operation completion strategies; The target entity set is matched with the key values ​​in the query condition dictionary to obtain the matched fields and missing fields; The missing field is filled in according to the missing field completion strategy to obtain the completed field; The completed fields and the matched fields are combined to obtain a set of vehicle information entities.

5. The method as described in claim 4, characterized in that, The step of completing the missing field according to the missing field completion strategy to obtain the completed field includes: The natural language instruction is concatenated with the target entity set to form an input sequence; The input sequence is semantically understood according to the semantic prediction and completion strategy to obtain the candidate completion value and confidence level corresponding to the missing field; When the confidence level corresponding to the candidate completion value is greater than or equal to a preset confidence threshold, the missing field is completed according to the candidate completion value to obtain the completed field; When the confidence level corresponding to the candidate completion value is less than a preset confidence threshold, the missing field is completed with wildcards according to the wildcard completion strategy to obtain the completed field.

6. The method as described in claim 4, characterized in that, The step of completing the missing field according to the missing field completion strategy to obtain the completed field includes: According to the historical operation completion strategy, the user's historical operation records within a preset time period are retrieved from the user's historical operation record database. Based on the historical operation records, the historical usage values ​​and frequency of the user for the missing field under the same operation type are statistically analyzed. Calculate the preference degree of each historical usage value based on the historical usage frequency, and take the historical usage value with the highest preference degree as the candidate completion value; The missing field is filled in according to the candidate completion value to obtain the completed field.

7. The method as described in claim 1, characterized in that, The step of managing the vehicle information database according to the database execution instruction and obtaining the return result corresponding to the database execution instruction includes: Execute the database execution instruction to obtain a set of vehicle information records that match the database execution instruction and the number of matching results; When the number of matching results is less than or equal to the preset single batch display threshold, the vehicle information record set will be returned as a result in the form of graphic cards; When the number of matching results exceeds the preset single batch display threshold, the vehicle information record set is written into a table file according to the preset table file generation component, and the download link address of the table file is returned as the result.

8. A vehicle information management device based on natural language processing, characterized in that, The device includes: The data receiving module is used to obtain a dictionary of query conditions for vehicle information and natural language commands input by the user; The intent parsing module is used to classify the natural language instruction and recognize entities according to a preset natural language processing model to obtain the target operation type and the target entity set. The preset natural language processing model includes an intent classification model and an entity recognition model. The missing optimization module is used to fill in the missing information of the target entity set according to the query condition dictionary to obtain the vehicle information entity set. The instruction generation module is used to generate structured database execution instructions based on the target operation type and the vehicle information entity set. The information management module is used to manage the vehicle information database according to the database execution instructions and obtain the return results corresponding to the database execution instructions.

9. A vehicle information management device based on natural language processing, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the vehicle information management method based on natural language processing as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the vehicle information management method based on natural language processing as described in any one of claims 1 to 7.

Images