Data processing method and device, equipment and storage medium

Abstract

The invention provides a data processing method and device, equipment and a storage medium, and the method comprises the steps: obtaining a retrieval request sent by a client, the retrieval request comprising at least one retrieval keyword; determining at least one retrieval tag corresponding to the at least one retrieval keyword; a target hierarchy matched with the at least one retrieval tag is inquired in a data grading model in the cache database, the data grading model is of a multi-hierarchy data structure, each hierarchy has a corresponding hierarchy tag, each hierarchy comprises a plurality of identification groups, and each identification group comprises a plurality of data identifiers; determining a target identification group matched with the retrieval keyword in the identification groups included in the at least one target hierarchy; and based on the data identifier included in the target identifier group matched with the retrieval keyword, obtaining target data in the file storage server, and sending the target data to the client. And the data scanning amount in the retrieval process is greatly reduced, so that the query efficiency and the response speed are remarkably improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a data processing method, device, equipment and storage medium. Background Art

[0002] In the current field of autonomous driving, mainstream solutions are gradually adopting end-to-end pure visual autonomous driving solutions, which collect a large amount of unstructured data through various sensors on the vehicle, such as car cameras, millimeter-wave radars, lidars, and ultrasonic radars. The cloud cleans, classifies, and removes a large amount of unstructured data to generate labeled data for large model training. In this context, the scale of unstructured data such as pictures, audio, and video has shown an exponential growth. How to efficiently filter data from various specific scenes or even combined scenes for training is a major problem currently faced.

[0003] Traditional retrieval methods usually involve manually adding label information to the data to describe unstructured data after the data is uploaded, and storing the unstructured data with label information in the database. After the user enters keywords in the client, the database is scanned in its entirety and the keywords are matched, and finally feedback is sent to the client that the keywords match the data.

[0004] The above retrieval method uses a full table scan, resulting in low retrieval efficiency and slow response speed. Summary of the invention

[0005] In order to solve the above technical problems, the present application provides a data processing method, device, equipment and storage medium to improve retrieval efficiency and response speed.

[0006] In a first aspect, the present application provides a data processing method, the method comprising:

[0007] Obtaining a search request sent by a client, where the search request includes at least one search keyword;

[0008] Determine at least one search tag corresponding to at least one search keyword;

[0009] Querying a target level matching at least one search tag in a data hierarchical model in a cache database, wherein the data hierarchical model is a multi-level data structure, each level has a corresponding level tag, each level includes a plurality of identification groups, each identification group has an associated index keyword, and each identification group includes a plurality of data identifiers;

[0010] Determining a target identification group matching the search keyword from identification groups included in at least one target level;

[0011] Based on the data identifier included in the target identification group matching the search keyword, the target data is obtained in the file storage server and sent to the client. The file storage server stores unstructured data, and each unstructured data has a unique data identifier.

[0012] In a second aspect, the present application provides a data processing device, the device comprising:

[0013] A search request acquisition module, used to acquire a search request sent by a client, wherein the search request includes at least one search keyword;

[0014] A search tag determination module, used to determine at least one search tag corresponding to at least one search keyword;

[0015] A target level determination module is used to query a target level matching at least one search tag in a data hierarchical model in a cache database, wherein the data hierarchical model is a multi-level data structure, each level has a corresponding level tag, each level includes multiple identification groups, each identification group has an associated index keyword, and each identification group includes multiple data identifiers;

[0016] A target identification group determination module, used to determine a target identification group matching the search keyword from identification groups included in at least one target level;

[0017] The target data sending module is used to obtain target data in the file storage server based on the data identifier included in the target identification group matching the search keyword, and send the target data to the client. The file storage server stores unstructured data, and each unstructured data has a unique data identifier.

[0018] In a third aspect, the present application provides a data processing device, comprising: one or more processors; a storage device for storing one or more programs; when the one or more programs are executed by one or more processors, the one or more processors implement the data processing method as described in the first aspect above.

[0019] In a fourth aspect, the present application provides a storage medium, which may be a computer-readable storage medium, storing a computer program thereon, which, when executed by a processor, implements the data processing method in the first aspect described above.

[0020] In a fifth aspect, an embodiment of the present application provides a computer program product, which includes a computer program or instructions, and when the computer program or instructions are executed by a processor, implements any data processing method as described in the first aspect above.

[0021] Compared with the prior art, the technical solution provided by the embodiments of the present application has the following advantages:

[0022] An embodiment of the present application provides a data processing method, apparatus, device and storage medium, the method comprising: obtaining a retrieval request sent by a client, the retrieval request including at least one retrieval keyword; determining at least one retrieval tag corresponding to the at least one retrieval keyword; querying a target level matching the at least one retrieval tag in a data hierarchical model in a cache database, the data hierarchical model being a multi-level data structure, each level having a corresponding level tag, each level including a plurality of identification groups, each identification group having an associated index keyword, and each identification group including a plurality of data identifiers; determining a target identification group matching the retrieval keyword in the identification groups included in at least one target level; obtaining target data in a file storage server based on the data identifier included in the target identification group matching the retrieval keyword, and sending the target data to the client, the file storage server storing unstructured data, each unstructured data having a unique data identifier.

[0023] The combination of data classification model and cache database enables fast retrieval of unstructured data. The hierarchical storage of data allows users to perform refined searches based on different search keywords. By encapsulating scene information tags into hierarchical indexes and storing them in the cache database, the amount of data scanning in the retrieval process is greatly reduced, thereby significantly improving query efficiency and response speed, meeting the needs of fast data processing in fields such as autonomous driving. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0025] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0026] Figure 1 A flowchart of a data processing method provided in an embodiment of the present application;

[0027] Figure 2 A schematic diagram of a data storage method provided in an embodiment of the present application;

[0028] Figure 3 A schematic diagram of the structure of a data processing device provided in an embodiment of the present application;

[0029] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0030] In order to more clearly understand the above-mentioned purposes, features and advantages of the present application, the scheme of the present application will be further described below. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0031] In the following description, many specific details are set forth to facilitate a full understanding of the present application, but the present application may also be implemented in other ways different from those described herein; obviously, the embodiments in the specification are only part of the embodiments of the present application, rather than all of the embodiments.

[0032] The term "including" and its variations used herein are open inclusions, i.e., "including but not limited to". The term "based on" means "based at least in part on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the following description.

[0033] It should be noted that the concepts such as "first" and "second" mentioned in this application are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0034] It should be noted that the modifications of "one" and "plurality" mentioned in the present application are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0035] The data processing method provided by the present application is described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0036] Figure 1 This is a flowchart of a data processing method in an embodiment of the present application. This embodiment can be applied to the storage and retrieval of a large amount of unstructured data. The method can be executed by a data processing device. The data processing device can be implemented in software and / or hardware. The data processing device can be configured in an electronic device, and the electronic device can include a server.

[0037] like Figure 1 As shown, the data processing method provided in the embodiment of the present application mainly includes steps S101-S104.

[0038] S101: Obtain a search request sent by a client, where the search request includes at least one search keyword.

[0039] Among them, the client can be understood as a tool or application for the user to interact with the model training server or model training platform during the vehicle autonomous driving model training process. The client has an interface for interacting with the model training server or model training platform to obtain training data, set training tasks, monitor training progress, etc. Furthermore, the client can also interact with the server that executes the data processing method of the embodiment of the present application to obtain the training data required for the training model from the server.

[0040] A search request is a search instruction consisting of a set of keywords or phrases input by a user through a client interface, which is used to instruct the server to perform search matching based on the search request and feed back the retrieved target data to the client.

[0041] The search request includes at least one search keyword, which refers to one or more words or phrases carried in the search request. In the process of training the vehicle automatic driving model, it is necessary to obtain data of various specific scenes or even combined scenes for training. The search keyword is used to indicate the specific scene or combined scene that the user wants to obtain.

[0042] For example, if a search keyword is "daytime", the server needs to retrieve the unstructured data that satisfies the search keyword "daytime". If multiple search keywords are "daytime", "sunny day", "city road", and "heavy traffic", the server needs to retrieve the unstructured data that satisfies the four search keywords "daytime", "sunny day", "city road", and "heavy traffic".

[0043] In one possible implementation, the user directly enters one or more search keywords in the search box of the client interface. The client generates a search request in response to the user input and sends it to the server. After receiving the search request sent by the client, the server parses the search request to obtain one or more search keywords included in the search request.

[0044] For example: the user directly enters four phrases "daytime", "sunny day", "city road", and "heavy traffic volume" as search keywords in the search box of the client interface. The client generates a search request in response to the four phrases entered by the user and sends it to the server. After the server receives the search request sent by the client, it parses the search request and obtains the four phrases "daytime", "sunny day", "city road", and "heavy traffic volume" included in the search request.

[0045] In one possible implementation, a user enters search content described in natural language in a search box on a client interface. The client generates a retrieval request in response to the search content described in natural language entered by the user, and sends it to a server. After receiving the retrieval request sent by the client, the server parses the retrieval request to obtain the search content described in natural language included in the retrieval request.

[0046] Specifically, inputting search content in the search box may include a variety of different input methods such as text box input, voice input, or scanning input, so as to provide diverse input methods to meet the input needs of different users.

[0047] The natural language processing (NLP) model is used to parse the search content described in natural language input by the user and extract multiple corresponding search keywords. Specifically, the natural language processing technology (such as word segmentation tool) is used to preliminarily process the search content described in natural language input by the user and segment it into keywords, phrases or word groups. For example, the keywords "sunny", "city road" and "congestion" are extracted from "the traffic volume is large on the city road on a sunny day".

[0048] Furthermore, if the search keywords entered by the user are unclear or incomplete, the input information can be further inferred and supplemented with context. If the user input is not clear enough, it can be inferred and supplemented based on historical data, geographic location, weather and other factors. For example, if the user only enters "daytime", the possible time range can be automatically inferred.

[0049] S102: Determine at least one search tag corresponding to at least one search keyword.

[0050] Each search keyword has its corresponding search tag. In the embodiment of the present application, the search tag corresponding to the search keyword can be determined by the category to which the search keyword belongs.

[0051] Furthermore, the correspondence between keywords and tags may be predefined. After the search keyword is acquired in S101, the search tag corresponding to the search keyword may be obtained by searching in the predefined correspondence between keywords and tags.

[0052] In the correspondence between predefined keywords and tags, one tag can correspond to multiple keywords. For example: when the tag is "time period", its corresponding keywords may include "daytime", "night", etc. For another example: when the tag is "weather", its corresponding keywords may include "sunny", "cloudy", "rainy", "foggy", etc. For another example: when the tag is "road type", its corresponding keywords may include "urban roads", "rural roads", "highways", "viaducts", "parking lots", etc. For another example: when the tag is "road conditions", its corresponding keywords may include "smooth", "congested", "downhill", "uphill", "curve", etc.

[0053] If the search keywords determined in S101 include "sunny day", "city road" and "congestion", then by querying the correspondence between the above keywords and tags, the corresponding search tags can be obtained as "weather", "road type" and "road condition" respectively.

[0054] It should be noted that the correspondence between the above keywords and tags can be customized according to actual conditions, and the specific correspondence is not limited in the embodiments of the present application.

[0055] Furthermore, each search keyword determines its corresponding search tag. If the corresponding search tag cannot be found in the above correspondence based on the search keyword, then the search keyword is identified based on its semantics and the tag with the closest semantic relationship to it is selected from multiple tags as the search tag corresponding to the search keyword.

[0056] S103. Query a target level that matches at least one search tag in a data hierarchy model in a cache database, wherein the data hierarchy model is a multi-level data structure, each level has a corresponding level tag, each level includes multiple identification groups, each identification group has an associated index keyword, and each identification group includes multiple data identifiers.

[0057] First, the data classification model is a data model that manages data according to certain levels or data categories. The above data classification model can be stored in the form of a hash table or a tree structure. In the embodiment of the present application, the data classification model is stored in a tree structure as an example for explanation.

[0058] The data classification model is a data structure with multiple hierarchical relationships. Each level has its corresponding level label, and the level label corresponds one-to-one with the label predefined in the above S102.

[0059] In one possible implementation, the data classification model includes at least 4 layers of data structure; the hierarchical labels of the first layer are time period information used to describe driving data, the hierarchical labels of the second layer are weather information used to describe driving scenarios, the hierarchical labels of the third layer are environmental information used to describe driving roads, and the hierarchical labels of the fourth layer are road condition information used to describe road conditions.

[0060] For example, a data classification model including a four-layer data structure is used as an example for explanation. The first layer has a label of "time period", the second layer has a label of "weather", the third layer has a label of "road type", and the fourth layer has a label of "road condition".

[0061] Furthermore, each level in the data classification model includes multiple identification groups, and each identification group has an associated index keyword. For example, the first level includes two identification groups, and the index keywords associated with these two identification groups are "daytime" and "nighttime", respectively. Then, the identification group associated with the index keyword "daytime" includes multiple data identifiers, and the scene of data display corresponding to these multiple data identifiers is "daytime". The identification group associated with the index keyword "night" includes multiple data identifiers, and the scene of data display corresponding to these multiple data identifiers is "night".

[0062] Furthermore, each upper-level identification group is associated with one or more lower-level identification groups, and the scene displayed by the data corresponding to the lower-level identification group satisfies the index keywords of itself and all upper-level identification groups associated with it.

[0063] The hierarchical label of the first level is "time period", and the first level includes 2 identification groups. The index keywords associated with these 2 identification groups are "daytime" and "nighttime", respectively. The hierarchical label of the second level is "weather", and the second level includes 4 identification groups associated with the identification group associated with "daytime" of the first level, and 4 identification groups associated with the identification group associated with "night" of the first level. The index keywords associated with these 4 identification groups are "sunny", "cloudy", "rainy", and "foggy", respectively. That is, the second level includes 8 identification groups. In the second level, the "sunny" identification group associated with the identification group associated with "daytime", the data scene corresponding to the data identifier in this identification group satisfies both "daytime" and "sunny".

[0064] The third level has a level label of "road type", and includes 5 identification groups associated with the identification group associated with "sunny day" of the second level, 5 identification groups associated with the identification group associated with "cloudy day" of the second level, 5 identification groups associated with the identification group associated with "rainy day" of the second level, and 5 identification groups associated with the identification group associated with "foggy day" of the second level. The index keywords associated with these 5 identification groups are "urban road", "rural road", "highway", "viaduct", and "parking lot".

[0065] The level label of the fourth level is "Road Conditions", and the fourth level includes 5 identification groups associated with the identification group associated with the third level "Urban Roads", 5 identification groups associated with the identification group associated with the third level "Rural Roads", 5 identification groups associated with the identification group associated with the third level "Expressways", 5 identification groups associated with the identification group associated with the third level "Viaducts", and 5 identification groups associated with the identification group associated with the third level "Parking Lots". The index keywords associated with these 5 identification groups are "smooth", "congested", "downhill", "uphill", and "curve".

[0066] After the search tag is determined in S102, it is matched with the hierarchical tag in the data hierarchical model to determine the level corresponding to the search tag, thereby narrowing the search scope of the search keyword. For example: the search tag is "weather", after matching with the hierarchical tag in the data hierarchical model, it is determined that the level corresponding to the "weather" tag is the second level, then the second level can be used as the target level. Another example: the search tag is "traffic conditions", after matching with the hierarchical tag in the data hierarchical model, it is determined that the level corresponding to the "traffic conditions" tag is the fourth level, then the fourth level can be used as the target level.

[0067] When there is one search tag, a matching target level can be determined in the data classification model. When there are multiple search tags, multiple target levels can be determined in the data classification model.

[0068] In one possible implementation, when there are multiple retrieval tags, querying the data hierarchy model in the cache database for a target level that matches at least one retrieval tag includes: querying the data hierarchy model in the cache database for multiple target levels that match multiple retrieval tags.

[0069] For example: if the search tag is "weather", after matching with the level tags in the data classification model, it is determined that the level corresponding to the "weather" tag is the second level, then the second level can be used as the target level. If the search tag is "traffic conditions", after matching with the level tags in the data classification model, it is determined that the level corresponding to the "traffic conditions" tag is the fourth level, then the fourth level can be used as the target level. Both the second level and the fourth level are target levels.

[0070] S104: Determine a target identification group that matches the search keyword from identification groups included in at least one target level.

[0071] Each identification group has its corresponding index keyword, and the search keyword is matched with the index keyword in its corresponding target level, and the identification group associated with the matched index keyword is used as the target identification group. The corresponding target level refers to the target level determined by the search tag of the search keyword.

[0072] For example: the search keyword is "sunny", and its corresponding search tag is "weather". The target level determined by the search tag "weather" in the data classification model is the second level, and the second level includes multiple index keywords. The search keyword "sunny" is matched with the multiple index keywords in the second level, and the identification group whose matched index keyword is "sunny" is used as the target identification group.

[0073] In one possible implementation, determining a target identification group that matches the search keyword from among the identification groups included in at least one of the target levels includes: based on a plurality of the search keywords, sequentially searching among the identification groups included in the plurality of target levels in the cache database in order from top to bottom to determine a target identification group that matches the search keyword.

[0074] After determining the search keywords in step S101, the search keywords are sorted in the order from top to bottom of the hierarchical division of the data hierarchical model. That is, the order of the search keywords is consistent with the hierarchical order in the data hierarchical model. For example: the hierarchical labels from top to bottom of the data hierarchical model are: time period, weather, road type, road condition, then the search keywords are also arranged in the above order. The arranged search keywords are obtained in turn and searched and matched from top to bottom in the data hierarchical model, and the last matched index group is used as the index group matching the search keyword.

[0075] In one possible implementation, a search request includes multiple search keywords, including: the search request includes a first search keyword, a second search keyword, a third search keyword and a fourth search keyword; based on the first search keyword, searching in the first-level identification group of the data hierarchy model to obtain the first-level identification group that matches the first search keyword; based on the second search keyword, searching in the second-level identification group that has a hierarchical relationship with the first-level identification group that matches the first search keyword to obtain the second-level identification group that matches the second search keyword; based on the third search keyword, searching in the third-level identification group that has a hierarchical relationship with the second-level identification group that matches the second search keyword to obtain the third-level identification group that matches the third search keyword; based on the fourth search keyword, searching in the fourth-level identification group that has a hierarchical relationship with the fourth-level identification group that matches the third search keyword to obtain the fourth-level identification group that matches the fourth search keyword; using the fourth-level identification group that matches the fourth search keyword as the identification group that matches the multiple search keywords.

[0076] The identification group that matches the search keyword means that the index keyword associated with the identification group is consistent with the search keyword.

[0077] The following description will be given by taking as an example the case where the first search keyword is “daytime”, the second search keyword is “sunny day”, the third search keyword is “city road”, and the fourth search keyword is “congestion”.

[0078] The first search keyword is "daytime", and a match is performed among the various label groups (the index keywords associated with the label groups are "daytime" and "night") in the first level (the level label is the time period) of the data classification model to obtain an identification group whose index keyword is "daytime".

[0079] Get the second-level (level label is weather) identification group associated with the identification group whose index keyword is "sunny" and match them among the identification groups (index keywords associated with the label group are "sunny", "rainy", "cloudy" and "foggy") to obtain the identification group whose index keyword is "sunny".

[0080] The identification group whose index keyword is "sunny day" is obtained by matching each identification group (the index keywords associated with the label group are "urban roads", "rural roads", "highways", "viaducts", and "parking lots") of the third level (the level label is the time period road type) associated with the identification group whose index keyword is "sunny day" to obtain the identification group whose index keyword is "urban roads".

[0081] The identification group whose search keyword is "urban road" is obtained by matching the identification groups of the fourth level (the level label is road conditions) (the index keywords associated with the label group are "smooth", "congested", "downhill", "uphill", and "curve") to obtain the identification group whose search keyword is "congested".

[0082] The identification group whose search keyword "congestion" is matched in sequence from top to bottom is used as the target identification group for matching of multiple search keywords.

[0083] S105. Based on the data identifier included in the target identification group matching the search keyword, the target data is obtained in the file storage server, and the target data is sent to the client. The file storage server stores unstructured data, and each unstructured data has a unique data identifier.

[0084] In an embodiment of the present application, the data classification model and the unstructured data are stored in different storage locations, the data classification model is stored in a cache database, and the unstructured data is stored in a file storage server.

[0085] The cache database may be a high-speed cache database for quick access and improved access efficiency.

[0086] The file storage server may be a distributed file storage server, which is used to store unstructured data and ensure high availability and scalability of data through features such as file distributed storage, load balancing, and fault recovery. Furthermore, the data identifier is assigned by the file storage server to the uploaded data and is a unique identifier for subsequent data tracking.

[0087] After determining the data identifier included in the index group matching the search keyword from the data classification model in the cache database, data corresponding to the determined data identifier is read from the file storage server based on the determined data identifier as target data.

[0088] The target data is fed back to the client as a response to the retrieval request, so that the client can display the requested target data.

[0089] Since retrieval and matching are performed in the cache database, a large amount of unstructured data is stored in the file storage server. While realizing the storage of unstructured data, the query efficiency and response speed can be significantly improved.

[0090] Through the combination of data hierarchical model, cache database and file storage server, fast retrieval of unstructured data is achieved. By encapsulating scene information tags into hierarchical indexes and storing them in the cache database, a large amount of unstructured data is stored in the file storage server and associated through data identifiers. Since only the cache database needs to be searched during the retrieval process, the query efficiency and response speed can be significantly improved.

[0091] The data classification model uses a hierarchical data storage method to store scene tags, allowing users to perform refined searches based on different search keywords. This greatly reduces the amount of data scanning during the search process, further improves query efficiency and response speed, and meets the needs of fast data processing in areas such as autonomous driving.

[0092] In the above embodiments, the process of data retrieval is mainly introduced. In the embodiments of the present application, the process of data storage is mainly introduced. Specifically, Figure 2 As shown, the data storage process provided in the embodiment of the present application mainly includes:

[0093] S201. Acquire a plurality of unstructured data to be processed.

[0094] The data to be processed refers to various types of unstructured data uploaded by users, including common file formats such as audio, video, and pictures. The above-mentioned data to be processed is a large amount of collected scene data, which is mainly used for subsequent training of vehicle autonomous driving models. Unstructured data refers to data with various file formats, not a single format.

[0095] Furthermore, in order to support different types of data, interface verification and permission control are first performed through the data access layer (API gateway). The access layer is responsible for confirming the user's identity and ensuring that the uploaded data meets the data security requirements. After determining that the data meets the data security requirements, the data to be processed is stored in the file storage server.

[0096] S202: Store each piece of data to be processed in a file storage server, and receive a data identifier of each piece of data to be processed fed back by the file storage server.

[0097] Store multiple pending data in a distributed file storage server (such as HDFS, Ceph, etc.). The file storage server performs reasonable distributed management based on the size, type and upload time of the data. Each uploaded pending data will be assigned a unique data identifier for subsequent tracking.

[0098] Furthermore, the data to be processed will be backed up to multiple physical nodes of the file storage server at the same time to ensure high data availability. The backup operation is automatically executed according to the system's preset cycle and strategy, supporting fault recovery and disaster recovery.

[0099] Distributed file storage servers are used to store unstructured data. They ensure high availability and scalability of data through features such as distributed data storage, load balancing, and fault recovery.

[0100] S203: extract multiple keywords from each piece of data to be processed and a label corresponding to each keyword.

[0101] Keywords of the data to be processed are used to describe key scene information in the data to be processed, such as: collection period, weather conditions during collection, collected road types and road conditions, etc.

[0102] Furthermore, deep learning models (such as CNN, Transformer, etc.) are used to identify the content of audio, video, and pictures. The audio files are analyzed through Mel-frequency cepstral coefficients (MFCC) and audio classification models (such as LSTM or CNN) to identify the sound environment (such as car horns, rain, etc.) and its scene information. The frame data in the video is extracted through computer vision algorithms (such as target detection, image segmentation, etc.) to identify the time (such as day, night) and weather conditions (such as sunny, rainy). The pictures use image classification and object detection models (such as YOLO, ResNet) to extract key visual information.

[0103] In a possible implementation, extracting multiple keywords from each data to be processed also includes: identifying the file type of the data to be processed; when the file type of the data to be processed is an audio file, using audio feature extraction technology and an audio classification model to analyze the data to be processed, and identifying multiple keywords in the data to be processed; when the file type of the data to be processed is a video file, using a computer vision algorithm to extract frame data in the data to be processed, and identifying multiple keywords in the data to be processed; when the file type of the data to be processed is a picture file, identifying the data to be processed through an image classification algorithm and an object detection model, and identifying multiple keywords in the data to be processed.

[0104] The file types of the data to be processed include audio files, video files, and picture files. Obtaining the file type of the data to be processed may include: reading the file extension of the data to be processed, and determining the file type of the data to be processed according to the file extension. For example: when the file extension is a video extension such as .MP4, .mov, etc., it is determined that the file type of the data to be processed is a video file. When the file extension is an audio extension such as .MP3, .mav, etc., it is determined that the file type of the data to be processed is an audio file. When the file extension is an image extension such as .jpg, .png, .gif, etc., it is determined that the file type of the data to be processed is a picture file.

[0105] When the file type of the data to be processed is an audio file, audio feature extraction technology (such as Mel-frequency cepstral coefficient MFCC) and audio classification models (such as LSTM or CNN) are used to analyze the audio file to identify the sound environment (such as car horns, rain sounds, etc.) and its corresponding scene information.

[0106] When the file type of the data to be processed is a video file, frame data in the video is extracted through computer vision algorithms (such as target detection, image segmentation, etc.) to identify dynamic scene information, including time (such as daytime, night) and weather conditions (such as sunny, rainy).

[0107] When the file type of the data to be processed is an image file, the uploaded image is processed using image classification and object detection models (such as YOLO and ResNet) to extract key visual information.

[0108] In the embodiment of the present application, image classification and object detection models (such as YOLO and ResNet) are used to process the uploaded data and extract key visual information, thereby avoiding the manual labeling process and improving the data processing speed.

[0109] Integrate audio, video and image content recognition into one system for intelligent keyword generation, and generate fine-grained tags based on specific scenarios, especially for applications in traffic scenarios, to improve the practicality and accuracy of data and support more precise applications.

[0110] In one possible implementation, multiple keywords are extracted for each data to be processed, including: for each data to be processed, obtaining the collection time of the data to be processed, and determining the first keyword of the data to be processed based on the collection time; obtaining the collection location of the data to be processed, and determining the third keyword of the data to be processed based on the collection location and pre-set map data; determining the second keyword of the data to be processed based on the collection location, the collection time, and pre-acquired weather data; and determining the fourth keyword of the data to be processed based on the collection location, the collection time, and pre-acquired road condition information.

[0111] The collection time of the data to be processed refers to the specific time point or time period when the data is collected or recorded. The collection time of the data to be processed can be obtained from the metadata of the data to be processed. The timestamp is used as the collection time of the data to be processed. The first keyword is a keyword that describes the time period information.

[0112] Furthermore, the collection time of the data to be processed includes the collection date and the precise collection time. For example, the collection date can be year, month, day, and the precise collection time can be hour, minute, second, or even more precise to milliseconds or microseconds.

[0113] Based on the collection date, the sunrise time and sunset time corresponding to the date are queried from the Internet. If the precise collection time is within the interval between the sunrise time and the sunset time, it indicates that the first keyword is daytime. If the precise collection time is not within the interval between the sunrise time and the sunset time, it indicates that the first keyword is nighttime.

[0114] The collection location of the data to be processed refers to the specific geographical location where the data to be processed is collected, which can be expressed using a geographic coordinate system (such as longitude and latitude). The collection location of the data to be processed can be obtained from the metadata of the data to be processed. The third keyword is a keyword describing the road type information.

[0115] The preset map data includes the road type of each road and the geographical coordinates of each road. Based on the collection location of the data to be processed, the geographical coordinates matching the collection location are determined in the map data, and the road type of the road at the geographical coordinates is obtained as the third keyword of the data to be processed.

[0116] Weather data includes weather conditions in multiple regions at different time periods. Based on the collection location and collection time, the weather data obtained in advance is queried, and the weather conditions matching the collection location and collection time are determined as the second keyword of the data to be processed. The second keyword is a keyword that describes weather information.

[0117] The pre-acquired road condition information includes the specific road conditions of each road in each time period. Based on the collection location and collection time, the pre-acquired road condition information is searched, and the road condition information matching the collection location and collection time is determined as the fourth keyword of the data to be processed. The fourth keyword is a keyword that describes the road condition information.

[0118] In the embodiment of the present application, multiple keywords of the data to be processed can be inferred from the metadata of the data to be processed, thus avoiding the manual labeling process of the data to be processed and avoiding the problem that manual labeling depends on the subjective judgment of the labeler, which may lead to uneven label quality and affect the quality of model training. In addition, the data processing speed can be improved.

[0119] In the embodiment of the present application, after the keywords of the data to be processed are extracted, the tags corresponding to the processed data are determined according to the keywords of the data to be processed.

[0120] Furthermore, in an embodiment of the present application, a process for customizing labels according to application scenarios in the field of autonomous driving is provided. For example, in an embodiment of the present application, four types of labels are mainly defined: time period labels, which are used to describe the time environment of driving data. Keywords in time period labels may include "daytime", "night", etc. Weather labels, which are used to describe the weather conditions of driving scenarios. Keywords in weather labels may include "sunny", "cloudy", "rainy", "foggy", etc. Road environment labels, which are used to define different road environments. Keywords in road environment labels may include "urban roads", "rural roads", "highways", "viaducts", "parking lots", etc. Road condition labels, which are used to indicate the current road status. Keywords in road condition labels may include "smooth", "congested", "downhill", "uphill", "curve", etc.

[0121] According to extracting keywords from the data to be processed, the tags corresponding to the keywords of the data to be processed are searched in the predefined correspondence between keywords and tags. The specific query method is the same as the process of querying the search tags based on the search keywords provided in the above embodiment, and the specific description can be referred to in the above embodiment.

[0122] Furthermore, the keywords and tags of the data to be processed can be displayed in the client interface, and on this basis, users can be used to edit and supplement them. A user feedback mechanism can be used to further optimize model training using the keywords and tag information provided by users, thereby improving the accuracy of subsequent tag generation.

[0123] S204: Generate a data classification model based on multiple keywords of each data to be processed and a label corresponding to each keyword.

[0124] The data classification model is a data structure with multi-level relationships. Each level has its corresponding level label. The data classification model includes at least 4 layers of data structure; the level label of the first level is the time period information used to describe the driving data, the level label of the second level is the weather information used to describe the driving scene, the level label of the third level is the environmental information used to describe the driving road, and the level label of the fourth level is the road condition information used to describe the road status.

[0125] For example, a data classification model including a four-layer data structure is used as an example for explanation. The first layer has a label of "time period", the second layer has a label of "weather", the third layer has a label of "road type", and the fourth layer has a label of "road condition".

[0126] In one possible implementation, a data hierarchical model is generated based on multiple keywords of each data to be processed and labels corresponding to each keyword, including: for each data to be processed, based on the label corresponding to the keyword of the data to be processed, querying the data hierarchical model for a hierarchical label that matches the label of the data to be processed; determining an index keyword that matches the keyword of the data to be processed from a plurality of index keywords included in the hierarchy corresponding to the hierarchical label; and encapsulating the data identifier of the data to be processed into an identification group associated with the index keyword that matches the keyword of the data to be processed.

[0127] For example, the keyword of the data to be processed is "daytime", and its corresponding label is "weather". Therefore, the identification group associated with the keyword "daytime" is determined in the first level including multiple index keywords, and the data identifier of the data to be processed is encapsulated into the identification group.

[0128] The keywords of the data to be processed also include "sunny", and a match is performed among the second-level index keywords associated with the identification group associated with the keyword "daytime" to obtain an identification group whose keyword is "sunny", and the data identifier of the data to be processed is encapsulated into this identification group.

[0129] The keywords of the data to be processed also include "city roads", which are matched among the third-level index keywords associated with the identification group associated with the keyword "sunny day" to obtain an identification group whose keyword is "city roads", and the data identifier of the data to be processed is encapsulated into the identification group.

[0130] The keywords of the data to be processed also include "congestion", which is matched with the fourth-level index keywords associated with the identification group associated with the keyword "urban road" to obtain an identification group with the keyword "congestion", and the data identifier of the data to be processed is encapsulated into the identification group.

[0131] Furthermore, encapsulating the data identifier into the identifier group also includes encapsulating specific file pointers, timestamps, label weights and other information into the identifier group.

[0132] The above processing is performed for each piece of data to be processed, so that the data identifier of each piece of data to be processed can be encapsulated into the data classification model.

[0133] Combine database technology and data structures (such as NoSQL databases and key-value storage) to store tags and metadata, associate the extracted keywords with data files through data identifiers to form complete data entries, and store the extracted keywords and tags in a hierarchical structure to facilitate subsequent retrieval and call.

[0134] S205: Store the data classification model in a cache database.

[0135] In a possible implementation, storing the data classification model in the cache database includes: storing identification groups of different levels in the data classification model in different cache intervals of the cache database.

[0136] The cache database will dynamically adjust the cache strategy to prioritize the caching of frequently accessed index keywords and level tags. The load balancing mechanism distributes cache queries to multiple nodes to ensure that the query processing capacity is sufficient to handle large-scale concurrent requests.

[0137] The data storage method provided in the embodiments of the present application has the following advantages:

[0138] (1) Through intelligent label generation and enhancement, efficient and automated label generation of the data to be processed is achieved. The content recognition algorithm is used to automatically extract keywords in the data to be processed, such as time, weather, road conditions, etc., which greatly reduces the reliance on manual labeling. At the same time, the accuracy and completeness of the labels are further improved by manually supplementing the labels to ensure the consistency and accuracy of the label quality. This combination of automation and manual work not only improves processing efficiency, but also adapts to the needs of different complex scenarios, effectively improving the availability of data.

[0139] (2) The data classification model and data classification index are introduced to make the data index more structured. The extracted keywords are encapsulated into a hierarchical data classification model and stored hierarchically in the cache database, which helps to maintain efficient access speed and retrieval performance as the data scale grows. This hierarchical storage method can support combined condition queries, such as fast screening by multiple conditions such as time period, weather and scene, thereby effectively improving the query efficiency under complex retrieval requirements.

[0140] (3) The data classification model is stored in the cache database, which ensures instant access and rapid response to data, and is particularly suitable for the application requirements of large-scale data management in the field of autonomous driving.

[0141] (4) The association of data hierarchical model storage and metadata realizes the precise mapping relationship between index information and unstructured data, providing a fast and accurate path for subsequent retrieval. By associating the hierarchical model with metadata, the corresponding unstructured data can be quickly located during retrieval, avoiding the inefficiency of full traversal search. This association method not only improves the query response speed of the system, but also enhances the flexibility of data storage management, allowing users to quickly access target data files according to specific needs, significantly improving data management and retrieval efficiency, and enhancing the scalability and flexibility of the system.

[0142] Figure 3 is a structural diagram of a data processing device in an embodiment of the present application, such as Figure 3As shown, the data processing device 30 provided in the embodiment of the present application mainly includes: a search request acquisition module 31, a search tag determination module 32, a target level determination module 33, a target identification group determination module 33 and a target data sending module 35.

[0143] Among them, the retrieval request acquisition module 31 is used to obtain the retrieval request sent by the client, and the retrieval request includes at least one retrieval keyword; the retrieval tag determination module 32 is used to determine at least one retrieval tag corresponding to the at least one retrieval keyword; the target level determination module 33 is used to query the target level that matches the at least one retrieval tag in the data hierarchy model of the cache database, the data hierarchy model is a multi-level data structure, each level has a corresponding level tag, each level includes multiple identification groups, each identification group has associated index keywords, and each identification group includes multiple data identifiers; the target identification group determination module 33 is used to determine the target identification group that matches the retrieval keyword in the identification group included in at least one target level; the target data sending module 35 is used to obtain the target data in the file storage server based on the data identifier included in the target identification group that matches the retrieval keyword, and send the target data to the client, the file storage server stores unstructured data, and each unstructured data has a unique data identifier.

[0144] In one possible implementation, when there are multiple retrieval tags, the target level determination module 33 is specifically used to query multiple target levels that match the multiple retrieval tags in the data hierarchy model in the cache database; the target identification group determination module 33 is specifically used to search in sequence among the identification groups included in the multiple target levels in the cache database in a top-down order based on multiple retrieval keywords to determine the target identification group that matches the retrieval keywords.

[0145] In one possible implementation, the data classification model includes at least three layers of data structure; the hierarchical labels of the first level are time period information used to describe driving data, the hierarchical labels of the second level are weather information used to describe driving scenarios, the hierarchical labels of the third level are environmental information used to describe driving roads, and the hierarchical labels of the fourth level are road condition information used to describe road conditions.

[0146] In a possible implementation, it also includes: a data storage module, which is used to obtain multiple unstructured data to be processed; store each data to be processed in a file storage server, and receive a data identifier of each data to be processed fed back by the file storage server; extract multiple keywords of each data to be processed and a label corresponding to each keyword; generate a data classification model based on the multiple keywords of each data to be processed and the label corresponding to each keyword; store the data classification model in a cache database.

[0147] In one possible implementation, the data storage module is specifically used to query the data classification model for a hierarchical label that matches the label of the data to be processed for each data to be processed based on the label corresponding to the keyword of the data to be processed; determine the index keyword that matches the keyword of the data to be processed from the multiple index keywords corresponding to the hierarchy label; and encapsulate the data identifier of the data to be processed into an identification group associated with the index keyword that matches the keyword of the data to be processed.

[0148] In a possible implementation, the data storage module is specifically configured to store identification groups of different levels in the data classification model in different cache intervals of the cache database.

[0149] In one possible implementation, the data storage module is specifically used to obtain, for each piece of data to be processed, a collection time of the data to be processed, and determine a first keyword of the data to be processed based on the collection time; obtain a collection location of the data to be processed, and determine a third keyword of the data to be processed based on the collection location and pre-set map data; determine a second keyword of the data to be processed based on the collection location, the collection time, and pre-acquired weather data; and determine a fourth keyword of the data to be processed based on the collection location, the collection time, and pre-acquired road condition information.

[0150] In one possible implementation, the data storage module is specifically used to identify the file type of the data to be processed; when the file type of the data to be processed is an audio file, the audio feature extraction technology and the audio classification model are used to analyze the data to be processed, and multiple keywords in the data to be processed are identified; when the file type of the data to be processed is a video file, the frame data in the data to be processed is extracted by a computer vision algorithm, and multiple keywords in the data to be processed are identified; when the file type of the data to be processed is a picture file, the data to be processed is identified by an image classification algorithm and an object detection model, and multiple keywords in the data to be processed are identified.

[0151] The data processing device provided in the embodiment of the present application can execute the data processing method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0152] Figure 4 is a schematic diagram of the structure of an electronic device provided in this embodiment. The electronic device may include a data processing device, such as Figure 4 As shown, the electronic device 400 includes a processor 410, a memory 420, an input device 430, and an output device 440; the number of the processor 410 in the electronic device can be one or more. Figure 4A processor 410 is taken as an example; the processor 410, the memory 420, the input device 430 and the output device 440 in the electronic device can be connected via a bus or other means. Figure 4 The example of connecting through bus is taken in the following.

[0153] The memory 420 is a computer-readable storage medium that can be used to store software programs, computer executable programs and modules, such as program instructions / modules corresponding to the data processing method in the embodiment of the present invention. The processor 410 executes various functional applications and data processing of the electronic device by running the software programs, instructions and modules stored in the memory 420, that is, implements the data processing method provided in the embodiment of the present invention.

[0154] The memory 420 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system and at least one application required for a function; the data storage area may store data created according to the use of the terminal, etc. In addition, the memory 420 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 420 may further include a memory remotely arranged relative to the processor 410, and these remote memories may be connected to the electronic device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0155] The input device 430 may be used to receive input digital or character information and generate key signal input related to user settings and function control of the electronic device, and may include a keyboard, a mouse, etc. The output device 440 may include a display device such as a display screen.

[0156] This embodiment also provides a storage medium containing computer executable instructions, which are used to implement the data processing method provided by the embodiment of the present invention when executed by a computer processor.

[0157] Of course, the computer executable instructions of a storage medium including computer executable instructions provided by an embodiment of the present invention are not limited to the operations of the method described above, but can also execute related operations in the data processing method provided by any embodiment of the present invention.

[0158] Through the above description of the implementation methods, the technicians in the relevant field can clearly understand that the present invention can be implemented by means of software and necessary general hardware, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods described in each embodiment of the present invention.

[0159] It is worth noting that in the embodiment of the above-mentioned data processing device, the various units and modules included are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0160] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0161] The above description is only a specific implementation of the present application, so that those skilled in the art can understand or implement the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments described herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A data processing method, characterized in that: The method comprises: Obtaining a search request sent by a client, wherein the search request includes at least one search keyword; Determine at least one search tag corresponding to the at least one search keyword; Querying a target level matching the at least one search tag in a data hierarchical model in a cache database, wherein the data hierarchical model is a multi-level data structure, each level has a corresponding level tag, each level includes a plurality of identification groups, each identification group has an associated index keyword, and each identification group includes a plurality of data identifiers; Determining a target identification group matching the search keyword from identification groups included in at least one of the target levels; Based on the data identifier included in the target identification group matching the search keyword, the target data is obtained in the file storage server and sent to the client. The file storage server stores unstructured data, each of which has a unique data identifier.

2. The method according to claim 1, characterized in that When there are multiple search tags, Querying a target level matching the at least one search tag in a data hierarchical model in the cache database includes: querying a data hierarchical model in the cache database for a plurality of target levels matching the plurality of search tags; Determining a target identification group matching the search keyword from identification groups included in at least one of the target levels includes: Based on the multiple search keywords, the identification groups included in the multiple target levels in the cache database are searched in sequence from top to bottom to determine the target identification group matching the search keywords.

3. The method according to claim 2, characterized in that The data classification model includes at least 4 layers of data structure; the hierarchical labels of the first layer are time period information used to describe driving data, the hierarchical labels of the second layer are weather information used to describe driving scenes, the hierarchical labels of the third layer are environmental information used to describe driving roads, and the hierarchical labels of the fourth layer are road condition information used to describe road conditions.

4. The method according to claim 1, characterized in that: Also includes: Obtain multiple unstructured data to be processed; Storing each of the to-be-processed data in a file storage server, and receiving a data identifier of each of the to-be-processed data fed back by the file storage server; Extracting multiple keywords of each of the data to be processed and labels corresponding to each keyword; Generate a data classification model based on multiple keywords of each of the data to be processed and labels corresponding to each keyword; The data classification model is stored in a cache database.

5. The method according to claim 4, characterized in that Based on the multiple keywords of each of the data to be processed and the labels corresponding to each keyword, a data classification model is generated, including: For each data to be processed, based on the tag corresponding to the keyword of the data to be processed, query the data classification model for a level tag that matches the tag of the data to be processed; Determining, from among a plurality of index keywords included in a level corresponding to the level tag, an index keyword that matches the keyword of the data to be processed; The data identifier of the data to be processed is encapsulated into an identification group associated with an index keyword that matches the keyword of the data to be processed.

6. The method according to any one of claims 4, characterized in that The storing the data classification model in a cache database includes: The identification groups of different levels in the data classification model are stored in different cache intervals of the cache database.

7. The method according to claim 5, characterized in that The extracting of multiple keywords from each of the to-be-processed data includes: For each piece of data to be processed, acquiring the collection time of the data to be processed, and determining a first keyword of the data to be processed based on the collection time; Acquire a collection location of the data to be processed, and determine a third keyword of the data to be processed based on the collection location and preset map data; Determining a second keyword of the data to be processed based on the collection location, the collection time, and pre-acquired weather data; A fourth keyword of the data to be processed is determined based on the collection location, the collection time, and pre-acquired road condition information.

8. The method according to claim 7, characterized in that The extracting of multiple keywords from each of the to-be-processed data further includes: Identifying the file type of the data to be processed; When the file type of the data to be processed is an audio file, the data to be processed is analyzed using audio feature extraction technology and an audio classification model to identify multiple keywords in the data to be processed; When the file type of the data to be processed is a video file, extracting frame data in the data to be processed by using a computer vision algorithm, and identifying a plurality of keywords in the data to be processed; When the file type of the data to be processed is a picture file, the data to be processed is identified by an image classification algorithm and an object detection model, and a plurality of keywords in the data to be processed are identified.

9. A data processing device, characterized in that: The device comprises: A search request acquisition module, used to acquire a search request sent by a client, wherein the search request includes at least one search keyword; A search tag determination module, used to determine at least one search tag corresponding to the at least one search keyword; a target level determination module, configured to query a target level matching the at least one search tag in a data hierarchical model in a cache database, wherein the data hierarchical model is a multi-level data structure, each level has a corresponding level tag, each level includes a plurality of identification groups, each identification group has an associated index keyword, and each identification group includes a plurality of data identifiers; A target identification group determination module, used to determine a target identification group matching the search keyword from identification groups included in at least one of the target levels; The target data sending module is used to obtain target data in a file storage server based on a data identifier included in a target identification group matching the search keyword, and send the target data to the client. The file storage server stores unstructured data, and each unstructured data has a unique data identifier.

10. An electronic device, characterized in that: The device comprises: one or more processors; A storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the data processing method according to any one of claims 1 to 8.

11. A storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the data processing method according to any one of claims 1 to 8 is implemented.

Images