A data processing method and system

Abstract

The present invention provides a data processing method and system. The method includes: collecting various types of data in a target scenario and storing the collected data in a first database; based on the data requirements of the target scenario, obtaining feature data for each piece of data of each type stored in the first database, and storing the feature data corresponding to each piece of data of each type in a second database in the form of a structured data table. The structured data table corresponding to each piece of data of each type includes the feature data corresponding to the corresponding data and identification information; receiving a data request containing identification information sent by a data user in the target scenario, obtaining the feature data from the second database based on the identification information in the data request, and sending the obtained feature data to the data user. The data processing method and system provided by the present invention can perform standardized processing on multimodal data, which enables the data user to directly request and obtain the required data without having to collect and process the data.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular, to a data processing method and system. Background Art

[0002] Currently, when data users use data, they collect and process data based on specific problems to obtain the data they need. Under the current various rich data systems, how to standardize multi-modal data so that data users can directly obtain the data they need is an urgent problem to be solved currently. Summary of the Invention

[0003] In view of this, the present invention provides a data processing method and system for standardizing multi-modal data so that data users can directly obtain the data they need. The technical solutions are as follows:

[0004] A data processing method includes:

[0005] Collect various types of data in a target scenario, and store the collected data in a first database;

[0006] Based on the data requirements of the target scenario, obtain feature data for each piece of data of each type stored in the first database to obtain the feature data corresponding to each piece of data of each type;

[0007] Store the feature data corresponding to each piece of data of each type in a second database in the form of a structured data table, where the structured data table corresponding to each piece of data of each type includes the feature data corresponding to the corresponding data and identification information;

[0008] Receive a data request containing identification information sent by a data user in the target scenario, obtain feature data from the second database based on the identification information in the data request, and send the obtained feature data to the data user.

[0009] Optionally, the various types of data include some or all of video data, audio data, text data, picture data, and user behavior data;

[0010] The obtaining feature data for each piece of data of each type stored in the first database based on the data requirements of the target scenario includes:

[0011] If the various types of data include video data, then process each piece of video data into video data in a unified video format, and obtain video feature data for each piece of processed video data;

[0012] If the multiple types of data include audio data, each piece of audio data is processed into audio data in a unified audio format, and audio feature data is obtained for each processed piece of audio data;

[0013] If the multiple types of data include text data, each piece of text data is processed into text data in a unified text format, and text feature data is obtained for each processed piece of text data;

[0014] If the multiple types of data include picture data, each piece of picture data is processed into picture data in a unified picture format, and picture feature data is obtained for each processed piece of picture data;

[0015] If the multiple types of data include user behavior data, user behavior feature data is obtained for each piece of user behavior data.

[0016] Optionally, the multiple types of data include some or all of the types of data among video data, audio data, text data, picture data, and user behavior data;

[0017] If the multiple types of data include video data, the video feature data obtained for a piece of video data includes the representation vector of the object involved in the piece of video data and the concrete information of the piece of video data;

[0018] If the multiple types of data include audio data, the audio feature data obtained for a piece of audio data includes the recognition text of the piece of audio data and the concrete information of the piece of audio data;

[0019] If the multiple types of data include text data, the text feature data obtained for a piece of text data includes the relevant information of the object involved in the piece of text data;

[0020] If the multiple types of data include picture data, the picture feature data obtained for a piece of picture data includes the information of the picture itself and the information inferred from the picture;

[0021] If the multiple types of data include user behavior data, the user behavior feature data obtained for a piece of user behavior data includes the relationship data between the user and the object and user data.

[0022] Optionally, the multiple types of data include some or all of the types of data among video data, audio data, text data, picture data, and user behavior data;

[0023] If the multiple types of data include video data, the structured data table corresponding to each piece of video data contains the object identifier of the object involved in the corresponding video data;

[0024] If the multiple types of data include audio data, the structured data table corresponding to each piece of audio data contains the object identifier of the object involved in the corresponding audio data;

[0025] If the multiple types of data include text data, the structured data table corresponding to each piece of text data contains the object identifier of the object involved in the corresponding text data;

[0026] If the multiple types of data include picture data, the structured data table corresponding to each piece of picture data contains the object identifier of the object involved in the corresponding picture data;

[0027] If the multiple types of data include user behavior data, the structured data table corresponding to each piece of user behavior data contains the global auto-increment identifier, user identifier, and related object identifier corresponding to the user behavior data.

[0028] Optionally, the multiple types of data include some or all of the data types of video data, audio data, text data, picture data, and user behavior data;

[0029] The identification information in the data request is the global auto-increment identifier range;

[0030] The obtaining of the feature data from the second database based on the identification information in the data request includes:

[0031] Based on the global auto-increment identifier range in the data request, determine the target structured data table from the structured data tables corresponding to each piece of user behavior data;

[0032] Obtain the user identifier and object identifier from the target structured data table, and use the obtained user identifier as the target user identifier and the obtained object identifier as the target object identifier;

[0033] Based on the target user identifier and the target object identifier, obtain the feature data requested by the data request from the second database.

[0034] Optionally, the multiple types of data include video data, audio data, text data, picture data, and user behavior data;

[0035] The obtaining of the feature data requested by the data request from the second database based on the target user identifier and the target object identifier includes:

[0036] Based on the target object identifier, obtain video feature data, audio feature data, text feature data, and picture feature data from the second database;

[0037] Obtain user behavior feature data from the second database based on the target user identifier.

[0038] Optionally, the obtaining video feature data, audio feature data, text feature data, and picture feature data from the second database based on the target object identifier includes:

[0039] Invoke the video feature acquisition interface to obtain video feature data from the structured data table containing the target object identifier among the structured data tables corresponding to each video data.

[0040] Invoke the audio feature acquisition interface to obtain audio feature data from the structured data table containing the target object identifier among the structured data tables corresponding to each audio data.

[0041] Invoke the text feature acquisition interface to obtain text feature data from the structured data table containing the target object identifier among the structured data tables corresponding to each text data.

[0042] Invoke the picture feature acquisition interface to obtain picture feature data from the structured data table containing the target object identifier among the structured data tables corresponding to each picture data.

[0043] The obtaining user behavior feature data from the second database based on the target user identifier includes:

[0044] Invoke the user behavior feature acquisition interface to obtain user behavior feature data from the structured data table containing the target user identifier among the structured data tables corresponding to each user behavior data.

[0045] A data processing system includes: a data collection subsystem, a data processing subsystem, and a data acquisition subsystem;

[0046] The data collection subsystem is configured to collect various types of data in a target scenario and store the collected data in the first database;

[0047] The data processing subsystem is configured to obtain feature data for each piece of data of each type stored in the first database based on the data requirements of the target scenario to obtain the feature data corresponding to each piece of data of each type; and store the feature data corresponding to each piece of data of each type in the second database in the form of a structured data table, where the structured data table corresponding to each piece of data of each type includes the feature data corresponding to the corresponding data and identification information;

[0048] The data acquisition subsystem is configured to receive a data request containing identification information sent by a data user in the target scenario, obtain feature data from the second database based on the identification information in the data request, and send the obtained feature data to the data user.

[0049] Optionally, the multiple types of data include some or all of video data, audio data, text data, picture data, and user behavior data. The data processing subsystem correspondingly includes some or all of a video data processing module, an audio data processing module, a picture data processing module, a text data processing module, and a user behavior data processing module.

[0050] The video data processing module is configured to obtain video feature data for each piece of video data. Among them, the video feature data obtained for a piece of video data includes the representation vector of the object involved in the piece of video data and the concrete information of the piece of video data.

[0051] The audio data processing module is configured to obtain audio feature data for each piece of audio data. Among them, the audio feature data obtained for a piece of audio data includes the recognized text of the piece of audio data and the concrete information of the piece of audio data.

[0052] The text data processing module is configured to obtain text feature data for each piece of text data. Among them, the text feature data obtained for a piece of text data includes the relevant information of the object involved in the text data.

[0053] The picture data processing module is configured to obtain picture feature data for each piece of picture data. Among them, the picture feature data obtained for a piece of picture data includes the information of the picture itself and the information inferred from the picture.

[0054] The user behavior data processing module is configured to obtain user behavior feature data for each piece of user behavior data. Among them, the user behavior feature data obtained for a piece of user behavior data includes the relationship data between the user and the object and user data.

[0055] Optionally, the multiple types of data include some or all of video data, audio data, text data, and picture data, as well as user behavior data.

[0056] The structured data table corresponding to each piece of user behavior data contains a global auto-incrementing identifier corresponding to the user behavior data, a user identifier, and a related object identifier; the structured data table corresponding to each piece of data of other types contains an object identifier of the object involved in the corresponding data.

[0057] When the data acquisition subsystem acquires feature data from the second database based on the identification information in the data request, it is specifically configured to:

[0058] Based on the global auto-increment identification range in the data request, determine a target structured data table from the structured data tables corresponding to each piece of user behavior data;

[0059] Obtain a user identification and an object identification from the target structured data table, where the obtained user identification is used as the target user identification, and the obtained object identification is used as the target object identification;

[0060] Based on the target user identification and the target object identification, acquire the feature data requested by the data request from the second database.

[0061] The data processing method and system provided by the present invention first collect various types of data in a target scenario, store the collected data in a first database, and then, based on the data requirements of the target scenario, acquire feature data for each piece of data of each type stored in the first database to obtain the feature data corresponding to each piece of data of each type. Finally, store the feature data corresponding to each piece of data of each type in the form of a structured data table in a second database. When receiving a data request sent by a data user in the target scenario, feature data can be acquired from the second database based on the data request, and then the acquired feature data is sent to the data user. The data processing method provided by the embodiments of the present invention can perform standardized processing on multi-modal data, which enables the data user to directly request and obtain the data it needs without having to collect and process the data. Description of the Drawings

[0062] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on the provided drawings without creative efforts.

[0063] Figure 1 It is a schematic flowchart of the data processing method provided by the embodiments of the present invention;

[0064] Figure 2 It is a schematic flowchart of acquiring feature data from the second database based on the identification information in the data request provided by the embodiments of the present invention;

[0065] Figure 3 It is a schematic structural diagram of the data processing system provided by the embodiments of the present invention;

[0066] Figure 4This is an example of the data processing system provided by the embodiments of the present invention. Detailed implementation manners

[0067] Next, in combination with the accompanying drawings in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0068] In order to enable data users to obtain data that they can directly use, the inventors of this case attempted to propose a solution that can standardize multi-modal data. For this purpose, research was carried out. Through continuous research, a data processing method and system were finally proposed. The proposed data processing method and system can standardize multi-modal data, which enables data users to directly request and obtain the data they need without having to collect and process the data. Next, the data processing method and system provided by the present invention will be introduced through the following embodiments.

[0069] The first embodiment

[0070] Please refer to Figure 1 , which shows a schematic flowchart of the data processing method provided by the embodiments of the present invention, and may include:

[0071] Step S101: Collect various types of data in the target scenario and store the collected data in the first database.

[0072] Among them, the collected data may include some or all of the following types of data: video data, audio data, text data, picture data, user behavior data. Preferably, the collected data includes video data, audio data, picture data, text data, and user behavior data.

[0073] Optionally, the target scenario may be, but is not limited to, a data recommendation scenario, such as a movie and TV recommendation scenario. If the target scenario is a movie and TV recommendation scenario, the feature data obtained for each piece of data of each type stored in the first database is feature data related to movie and TV recommendations. Optionally, the first database may be, but is not limited to, a distributed database.

[0074] Step S102: Based on the data requirements of the target scenario, obtain feature data for each piece of data of each type stored in the first database to obtain the feature data corresponding to each piece of data of each type, and store the feature data corresponding to each piece of data of each type in the second database in the form of a structured data table.

[0075] If the data stored in the first database includes video data, video feature data is obtained for each piece of video data to obtain the video feature data corresponding to each piece of video data; if the data stored in the first database includes audio data, audio feature data is obtained for each piece of audio data to obtain the audio feature data corresponding to each piece of audio data; if the data stored in the first database includes text data, text feature data is obtained for each piece of text data to obtain the text feature data corresponding to each piece of text data; if the data stored in the first database includes picture data, picture feature data is obtained for each piece of picture data to obtain the picture feature data corresponding to each piece of picture data; if the data stored in the first database includes user behavior data, user behavior feature data is obtained for each piece of user behavior data to obtain the user behavior feature data corresponding to each piece of user behavior data.

[0076] After obtaining the feature data corresponding to each piece of data of each type stored in the first database, the feature data corresponding to each piece of data of each type is stored in the second database in the form of a structured data table, that is, the second database stores the structured data table corresponding to each piece of data of each type. The structured data table corresponding to each piece of data of each type includes an identification field and a feature field. The identification corresponding to the identification field in the structured data table is the identification information corresponding to the corresponding data, and the feature corresponding to the feature field in the structured data table is the feature data corresponding to the corresponding data. Optionally, the second database can be but is not limited to a distributed relational database.

[0077] Step S103: Receive a data request containing identification information sent by a data user in the target scenario, obtain feature data from the second database based on the identification information in the data request, and send the obtained feature data to the data user.

[0078] After obtaining the feature data, the data user can use the obtained feature data based on specific requirements. Exemplarily, the target scenario is a movie and TV recommendation scenario, and the data user is a movie and TV recommendation system. After obtaining the feature data, the movie and TV recommendation system can process the feature data into the required training data and use the training data to train a movie and TV recommendation model. For example, the feature data is processed into Batch data, and then the Batch data is divided into multiple parts and distributed to multiple GPUs to implement single-machine multi-GPU computing or multi-machine multi-GPU computing.

[0079] The data processing method provided by the embodiments of the present invention first collects various types of data in a target scenario, stores the collected data in a first database, and then, based on the data requirements of the target scenario, obtains feature data for each piece of data of each type stored in the first database to obtain the feature data corresponding to each piece of data of each type. Finally, the feature data corresponding to each piece of data of each type is stored in a second database in the form of a structured data table. When receiving a data request sent by a data user in the target scenario, the feature data can be obtained from the second database based on the data request, and then the obtained feature data is sent to the data user. The data processing method provided by the embodiments of the present invention can perform standardized processing on multi-modal data, which enables the data user to directly request and obtain the data it needs without having to collect and process the data.

[0080] Second Embodiment

[0081] Taking the data collected in step S101 including video data, audio data, picture data, text data, and user behavior data as an example, this embodiment introduces the process of "step S102: Based on the data requirements of the target scenario, obtain feature data for each piece of data of each type stored in the first database to obtain the feature data corresponding to each piece of data of each type, and store the feature data corresponding to each piece of data of each type in the second database in the form of a structured data table" in the above embodiment.

[0082] (1) Acquisition and Storage of Video Feature Data

[0083] The process of obtaining the video feature data corresponding to each piece of video data stored in the first database and storing the video feature data corresponding to each piece of video data in the second database in the form of a structured data table may include: For each piece of video data stored in the first database, first process the piece of video data into video data in a unified video format, then obtain the video feature data for the processed video data to obtain the video feature data corresponding to the piece of video data, and finally store the video feature data corresponding to the piece of video data in the second database in the form of a structured data table.

[0084] Among them, the structured data table stored in the second database for a piece of video data includes an identification field and a video feature field. The identification corresponding to the identification field is the object identification of the object involved in the piece of video data (such as a film or television work in a film and television recommendation scenario), and the video feature corresponding to the video feature field is the video feature data obtained for the piece of video data.

[0085] It should be noted that there are various formats for video data, such as mp4, avi, mkv, etc. For each piece of video data stored in the first database, first, the formats of these video data are unified, that is, each piece of video data is processed into video data in a unified video format, and then the standardization of data extraction is carried out. That is, for each piece of video data after formatting, standardized video feature data is obtained respectively. Finally, the video feature data obtained for each piece of video data after formatting is stored in the second database in the form of a structured data table.

[0086] Taking the target scenario as the movie and TV recommendation scenario as an example: First, each piece of video data in the movie and TV recommendation scenario stored in the first database is processed into video data in a unified video format, and then video feature data is obtained for each piece of video data after formatting. The obtained video feature data can include, but is not limited to, the vector representation corresponding to the object involved in the video data, the concrete information of the video data, etc. Among them, the concrete information of the video data can include the information of each scene involved in the video data, such as background, theme, recognized character list, etc. After obtaining the above video feature data, the obtained video feature data can be stored in the second database in the following form of a structured data table:

[0087] Table 1 Structured data table corresponding to video data

[0088]

[0089]

[0090] (2) Acquisition and storage of audio feature data

[0091] The process of obtaining the audio feature data corresponding to each piece of audio data stored in the first database and storing the audio feature data corresponding to each piece of audio data in the second database in the form of a structured data table can include: For each piece of audio data stored in the first database, first, the piece of audio data is processed into audio data in a unified audio format, and then audio feature data is obtained for the processed audio data to obtain the audio feature data corresponding to the piece of audio data. Finally, the audio feature data corresponding to the piece of audio data is stored in the second database in the form of a structured data table.

[0092] Among them, the structured data table stored in the second database for a piece of audio data includes an identification field and an audio feature field. The identification corresponding to the identification field is the object identification of the object involved in the piece of audio data, and the audio feature corresponding to the audio feature field is the audio feature data obtained for the piece of audio.

[0093] It should be noted that there are various formats for audio data, such as mp3, wav, wax, mpeg, etc. For each piece of audio data stored in the first database, first, the formats of these audio data are unified, that is, each piece of audio data is processed into audio data in a unified audio format, and then the standardization of data extraction is carried out, that is, for each piece of audio data after formatting, standardized audio feature data is obtained respectively. Finally, the audio feature data obtained for each piece of audio data after formatting is stored in the second database in the form of a structured data table.

[0094] Taking the target scenario as the movie and TV recommendation scenario as an example: First, each piece of audio data in the movie and TV recommendation scenario stored in the first database is processed into audio data in a unified audio format, and then audio feature data is obtained for each piece of audio data after formatting. The obtained audio feature data may include the recognition text corresponding to the audio data, and may also include some concrete information of the audio data (such as the start time, end time, background theme, etc. of the audio). After obtaining the above audio feature data, the obtained audio feature data can be stored in the second database in the following form of a structured data table:

[0095] Table 2 Structured data table corresponding to audio data

[0096]

[0097] (3) Acquisition and storage of text feature data

[0098] The process of obtaining the text feature data corresponding to each piece of text data stored in the first database and storing the text feature data corresponding to each piece of text data in the second database in the form of a structured data table may include: For each piece of text data stored in the first database, first, the piece of text data is processed into text data in a unified text format, and then text feature data is obtained for the processed text data to obtain the text feature data corresponding to the piece of text data. Finally, the text feature data corresponding to the piece of text data is stored in the second database in the form of a structured data table.

[0099] Among them, the structured data table stored by the second database for a piece of text data includes an identification field and a text feature field. The identification corresponding to the identification field is the object identification of the object involved in the piece of text data, and the text feature corresponding to the text feature field is the text feature data obtained for the piece of text data.

[0100] It should be noted that there are various text formats, such as txt, parquet, zip, etc. For each piece of text data stored in the first database, first, the formats of these text data are unified, that is, each piece of text data is processed into text data with a unified text format, and then the standardization of data refinement is carried out. That is, for each piece of text data after formatting, standardized text feature data is obtained respectively. Finally, the text feature data obtained for each piece of text data after formatting is stored in the second database in the form of a structured data table.

[0101] Taking the target scenario as the movie and TV recommendation scenario as an example: First, each piece of text data in the movie and TV recommendation scenario stored in the first database is processed into text data with a unified text format. Then, for each piece of text data after formatting, text feature data is obtained. The obtained text feature data may include information related to the movie or TV work involved in the text data (such as the name of the movie or TV work, the director of the movie or TV work, the actors in the movie or TV work, the language of the movie or TV work, etc.). After obtaining the above text feature data, the obtained text feature data can be stored in the second database in the following form of a structured data table:

[0102] Table 3 Structured Data Table Corresponding to Text Data

[0103]

[0104]

[0105] (4) Acquisition and Storage of Image Feature Data

[0106] The process of obtaining the image feature data corresponding to each piece of image data stored in the first database and storing the image feature data corresponding to each piece of image data in the second database in the form of a structured data table may include: For each piece of image data stored in the first database, first, the piece of image data is processed into image data with a unified image format. Then, for the processed image data, image feature data is obtained to obtain the image feature data corresponding to the piece of image data. Finally, the image feature data corresponding to the piece of image data is stored in the second database in the form of a structured data table.

[0107] Among them, the structured data table stored by the second database for a piece of image data includes an identification field and an image feature field. The identification corresponding to the identification field is the object identification of the object involved in the piece of image data, and the image feature corresponding to the image feature field is the image feature data obtained for the piece of image data.

[0108] It should be noted that there are various image formats, such as bmp, jpg, png, etc. For each piece of image data stored in the first database, first, the format of these image data is unified, that is, each piece of image data is processed into image data with a unified image format, and then the standardization of data refinement is carried out, that is, for each piece of image data after formatting, standardized image feature data is obtained respectively. Finally, the image feature data obtained for each piece of image data after formatting is stored in the second database in the form of a structured data table.

[0109] Taking the target scenario as the movie and TV recommendation scenario as an example: First, each piece of image data in the movie and TV recommendation scenario in the first database is processed into image data with a unified image format. Then, for each piece of image data after formatting, image feature data is obtained. The obtained image feature data may include image identifiers (an object may have multiple images, so there may be multiple image identifiers), background information of the image, a list of people in the image, text in the image, information inferred from the image, etc. After obtaining the above image feature data, the obtained image feature data can be stored in the second database in the following form of a structured data table:

[0110] Table 4 Structured Data Table Corresponding to Image Data

[0111]

[0112]

[0113] (4) Acquisition and Storage of User Behavior Feature Data

[0114] The process of obtaining the user behavior feature data corresponding to each piece of user behavior data stored in the first database and storing the user behavior feature data corresponding to each piece of user behavior data in the second database in the form of a structured data table may include: For each piece of user behavior data stored in the first database, first, the user behavior feature data is obtained for this piece of user behavior data to obtain the user behavior feature data corresponding to this piece of user behavior data. Then, the user behavior feature data corresponding to this piece of user behavior data is stored in the second database in the form of a structured data table. Optionally, the user behavior feature data may include relationship data between the user and the object and may also include user data.

[0115] In the case where the user behavior characteristic data includes the relationship data between the user and the object and the user data, the structured data table stored by the second database for a piece of user behavior data may include a first structured data table and a second structured data table. Among them, the first structured data table includes an identification field and a feature field. The identification field may include a user identification field and an object identification field. The user identification corresponding to the user identification field is the user identification of the user corresponding to the corresponding user behavior data, and the object identification corresponding to the object identification field is the object identification of the object concerned by the user corresponding to the corresponding user behavior data. Among them, the second structured data table includes a user identification field and a feature field. The user identification corresponding to the user identification field in the second structured data table is the same as the user identification corresponding to the user identification field in the first structured data table, and the feature corresponding to the feature field is the user data.

[0116] Taking the target scenario as the movie and TV recommendation scenario as an example: First, each piece of user behavior data in the movie and TV recommendation scenario in the first database is processed to obtain the relationship data between the user and the object (for example, the behavior category of the user towards the object he / she is concerned about, etc.) and the user data (for example, the number of active days of the user in the past 7 days, the number of active days of the user in the past 14 days, the hour period when the user is most active, etc.). The above-mentioned feature data obtained can be stored in the second database in the form of the following structured data table:

[0117] Table 5 First Structured Data Table Corresponding to User Behavior Data

[0118]

[0119]

[0120] Table 6 Second Structured Data Table Corresponding to User Behavior Data

[0121] Field Description id Global auto-incrementing id user_id User id live_days_7 Active days in the past 7 days live_days_14 Active days in the past 14 days most_active_hour The hour period with the most frequent activity, such as 5 o'clock to 6 o'clock

[0122] Third Embodiment

[0123] This embodiment introduces the process of "obtaining feature data from the second database based on the identification information in the data request" in step S103 of the above embodiment.

[0124] As mentioned in the above embodiments, the structured data table corresponding to each video data contains the object identifier of the object involved in the corresponding video data, the structured data table corresponding to each audio data contains the object identifier of the object involved in the corresponding audio data, the structured data table corresponding to each text data contains the object identifier of the object involved in the corresponding text data, the structured data table corresponding to each picture data contains the object identifier of the object involved in the corresponding picture data, and the structured data table corresponding to each user behavior data includes the global auto-increment identifier, user identifier, and related object identifier corresponding to the user behavior data. When sending a data request, the data user in the target scenario can carry identification information in the data request to indicate the data it requests.

[0125] In a possible implementation, the data request sent by the data user may include a global auto-increment identifier range. Please refer to Figure 2 , which shows a schematic flowchart of obtaining feature data from the second database based on the identification information in the data request when the identification information included in the data request sent by the data user is a global auto-increment identifier range, and may include:

[0126] Step S201: Determine the target structured data table from the structured data tables corresponding to each piece of user behavior data based on the identification information in the data request.

[0127] Among them, the target structured data table is the structured data table in which the global auto-increment identifier contained in the structured data tables corresponding to each piece of user behavior data is within the global auto-increment identifier range in the data request, that is, the global auto-increment identifier contained in the target structured data table is within the global auto-increment identifier range in the data request.

[0128] Step S202: Obtain the user identifier and object identifier from the target structured data table. The obtained user identifier is used as the target user identifier, and the obtained object identifier is used as the target object identifier.

[0129] Step S203: Obtain the feature data requested by the data request from the second database based on the target user identifier and target object identifier.

[0130] Specifically, the process of obtaining the feature data requested by the data request from the second database based on the target user identifier and target object identifier may include: obtaining video feature data, audio feature data, text feature data, and picture feature data from the second database based on the target object identifier; obtaining user behavior feature data from the second database based on the target user identifier.

[0131] Among them, the process of obtaining video feature data, audio feature data, text feature data, and image feature data from the second database based on the target object identifier may include: calling the video feature acquisition interface to obtain video feature data from the structured data table containing the target object identifier in the structured data tables corresponding to each video data; calling the audio feature acquisition interface to obtain audio feature data from the structured data table containing the target object identifier in the structured data tables corresponding to each audio data; calling the text feature acquisition interface to obtain text feature data from the structured data table containing the target object identifier in the structured data tables corresponding to each text data; calling the image feature acquisition interface to obtain image feature data from the structured data table containing the target object identifier in the structured data tables corresponding to each image data.

[0132] Among them, the process of obtaining user behavior feature data from the second database based on the target user identifier may include: calling the user behavior feature acquisition interface to obtain user behavior feature data from the structured data table containing the target user identifier in the structured data tables corresponding to each user behavior data.

[0133] Optionally, in this embodiment, the feature acquisition interface is based on a microservice framework, with an interface design pattern in RestFul format, and all feature data is obtained through the network socket method rather than the offline method.

[0134] The format definition of the video feature acquisition interface (API) is:

[0135] {ip}:{port} / video_feature / {item_id} /

[0136] Among them, ip is the address specified during deployment, port is the port specified during deployment, item_id is the incoming target object identifier, and the return result is the video feature data in the structured data table containing item_id.

[0137] The format definition of the audio feature acquisition interface (API) is:

[0138] {ip}:{port} / audio_feature / {item_id} /

[0139] Among them, ip is the address specified during deployment, port is the port specified during deployment, item_id is the incoming target object identifier, and the return result is the audio feature data in the structured data table containing item_id.

[0140] The format definition of the image feature acquisition interface (API) is:

[0141] {ip}:{port} / photo_feature / {item_id} /

[0142] Among them, ip is the address specified during deployment, port is the port specified during deployment, item_id is the identifier of the target object passed in, and the returned result is the picture feature data in the structured data table containing item_id.

[0143] The format definition of the text feature acquisition interface (API) is:

[0144] {ip}:{port} / text_feature / {item_id} /

[0145] Among them, ip is the address specified during deployment, port is the port specified during deployment, item_id is the identifier of the target object passed in, and the returned result is the text feature data in the structured data table containing item_id.

[0146] The format definition of the user behavior feature acquisition interface (API) is:

[0147] {ip}:{port} / user_feature / {user_id}

[0148] Among them, ip is the address specified during deployment, port is the port specified during deployment, user_id is the identifier of the target user passed in, and the returned result contains the user behavior feature data in the structured data table containing user_id.

[0149] The format definition of the acquisition interface (API) for the target user identifier and the target object identifier is:

[0150] {ip}:{port} / behavior_data / {id}

[0151] Among them, ip is the address specified during deployment, port is the port specified during deployment, and id is the global auto-incrementing identifier within the range of the global auto-incrementing identifiers in the structured data table corresponding to each piece of user behavior data in the data request. The returned result includes the target user identifier and the target object identifier.

[0152] The Fourth Embodiment

[0153] The embodiment of the present invention provides a data processing system. The data processing system provided by the embodiment of the present invention will be described below. The data processing system described below can be mutually corresponding and referred to the data processing method described above.

[0154] Please refer to Figure 3, which shows the structural schematic diagram of the data processing system provided by the embodiment of the present invention, may include: a data collection subsystem 301, a data processing subsystem 302, and a data acquisition subsystem 303.

[0155] The data collection subsystem 301 is used to collect various types of data in the target scenario and store the collected data in the first database.

[0156] The data processing subsystem 302 is used to obtain feature data for each piece of data of each type stored in the first database based on the data requirements of the target scenario, so as to obtain the feature data corresponding to each piece of data of each type; and store the feature data corresponding to each piece of data of each type in the second database in the form of a structured data table, where the structured data table corresponding to each piece of data of each type includes the feature data corresponding to the corresponding data and identification information.

[0157] The data acquisition subsystem 303 is used to receive a data request containing identification information sent by a data user in the target scenario, obtain feature data from the second database based on the identification information in the data request, and send the obtained feature data to the data user.

[0158] In a possible implementation manner, the various types of data include some or all of video data, audio data, text data, picture data, and user behavior data. The data processing subsystem 302 correspondingly includes some or all of a video data processing module, an audio data processing module, a picture data processing module, a text data processing module, and a user behavior data processing module.

[0159] The video data processing module is used to obtain video feature data for each piece of video data. Among them, the video feature data obtained for a piece of video data includes the representation vector of the object involved in the piece of video data and the figurative information of the piece of video data.

[0160] The audio data processing module is used to obtain audio feature data for each piece of audio data. Among them, the audio feature data obtained for a piece of audio data includes the recognized text of the piece of audio data and the figurative information of the piece of audio data.

[0161] The text data processing module is used to obtain text feature data for each piece of text data. Among them, the text feature data obtained for a piece of text data includes the relevant information of the object involved in the piece of text data.

[0162] The picture data processing module is used to obtain picture feature data for each piece of picture data. Among them, the picture feature data obtained for a piece of picture data includes the information of the picture itself and the information inferred from the picture.

[0163] When the video data processing module obtains video feature data for each piece of video data, it is specifically configured to process each piece of video data into video data in a unified video format, and obtain video feature data for each processed piece of video data.

[0164] When the audio data processing module obtains audio feature data for each piece of audio data, it is specifically configured to process each piece of audio data into audio data in a unified audio format, and obtain audio feature data for each processed piece of audio data.

[0165] When the text data processing module obtains text feature data for each piece of text data, it is specifically configured to process each piece of text data into text data in a unified text format, and obtain text feature data for each processed piece of text data.

[0166] When the picture data processing module obtains picture feature data for each piece of picture data, it is specifically configured to process each piece of picture data into picture data in a unified picture format, and obtain picture feature data for each processed piece of picture data.

[0167] In a possible implementation, the multiple types of data include some or all of video data, audio data, text data, picture data, and user behavior data.

[0168] If the multiple types of data include video data, the video feature data obtained for a piece of video data includes the representation vector of the object involved in the piece of video data and the concrete information of the piece of video data; if the multiple types of data include audio data, the audio feature data obtained by the data processing subsystem 302 for a piece of audio data includes the recognized text of the piece of audio data and the concrete information of the piece of audio data; if the multiple types of data include text data, the text feature data obtained by the data processing subsystem 302 for a piece of text data includes the relevant information of the object involved in the piece of text data; if the multiple types of data include picture data, the picture feature data obtained by the data processing subsystem 302 for a piece of picture data includes the information of the picture itself and the information inferred from the picture; if the multiple types of data include user behavior data, the user behavior feature data obtained by the data processing subsystem 302 for a piece of user behavior data includes the relationship data between the user and the object and user data.

[0169] In a possible implementation, the multiple types of data include some or all of video data, audio data, text data, picture data, and user behavior data.

[0170] The structured data table corresponding to each piece of user behavior data contains a globally incrementing identifier corresponding to the user behavior data, a user identifier, and a related object identifier; the structured data table corresponding to each piece of data of other types contains an object identifier of the object involved in the corresponding data.

[0171] When the data acquisition subsystem 303 acquires feature data from the second database based on the identification information in the data request, it is specifically used for:

[0172] Based on the range of the globally incrementing identifier in the data request, determine the target structured data table from the structured data tables corresponding to each piece of user behavior data; acquire the user identifier and the object identifier from the target structured data table, the acquired user identifier is used as the target user identifier, and the acquired object identifier is used as the target object identifier; based on the target user identifier and the target object identifier, acquire the feature data requested by the data request from the second database

[0173] In a possible implementation manner, the multiple types of data include video data, audio data, text data, picture data, and user behavior data. When the data acquisition subsystem 303 acquires the feature data requested by the data request from the second database based on the target user identifier and the target object identifier, it is specifically used for:

[0174] Based on the target object identifier, acquire video feature data, audio feature data, text feature data, and picture feature data from the second database; based on the target user identifier, acquire user behavior feature data from the second database.

[0175] In a possible implementation manner, when the data acquisition subsystem 303 acquires video feature data, audio feature data, text feature data, and picture feature data from the second database based on the target object identifier, it is specifically used for:

[0176] Call the video feature acquisition interface to acquire video feature data from the structured data table containing the target object identifier in the structured data tables corresponding to each piece of video data; call the audio feature acquisition interface to acquire audio feature data from the structured data table containing the target object identifier in the structured data tables corresponding to each piece of audio data; call the text feature acquisition interface to acquire text feature data from the structured data table containing the target object identifier in the structured data tables corresponding to each piece of text data; call the picture feature acquisition interface to acquire picture feature data from the structured data table containing the target object identifier in the structured data tables corresponding to each piece of picture data.

[0177] When the data acquisition subsystem 303 acquires user behavior feature data from the second database based on the target user identifier, it is specifically configured to:

[0178] Call the user behavior feature acquisition interface, and acquire user behavior feature data from the structured data table containing the target user identifier among the structured data tables corresponding to each piece of user behavior data.

[0179] In a possible implementation manner, each subsystem included in the data processing system provided in this embodiment can be implemented by a server. The server here can be a single server, a server cluster composed of multiple servers, or a cloud computing server center. The server may include a processor, a memory, and a network interface, etc. Please refer to Figure 4 , which shows an example of the data processing system provided in the embodiment of the present invention. Figure 4 The shown data processing system is a data processing system applied to a data recommendation scenario. Figure 4 The data collection subsystem 401 in [] collects various types of data (i.e., multi-modal data) in the data recommendation scenario, stores the collected data in the first database (such as a distributed database), and the data processing subsystem 402 acquires feature data for each piece of data of each type stored in the first database based on the data requirements of the data recommendation scenario to obtain the feature data corresponding to each piece of data of each type, and stores the feature data corresponding to each piece of data of each type in the second database (such as a distributed relational database) in the form of a structured data table. The data acquisition subsystem 403 receives a data request sent by the data recommendation system, acquires feature data from the second database based on the data request, and feeds back the acquired feature data (multi-modal feature data) to the data recommendation system. The data recommendation system constructs training data based on the obtained feature data, trains the data recommendation model with the constructed training data, and then recommends data to the user based on the trained data recommendation model.

[0180] The data processing system provided in the embodiment of the present invention first collects various types of data in the target scenario, stores the collected data in the first database, then acquires feature data for each piece of data of each type stored in the first database based on the data requirements of the target scenario to obtain the feature data corresponding to each piece of data of each type, and finally stores the feature data corresponding to each piece of data of each type in the second database in the form of a structured data table. When receiving a data request sent by a data user in the target scenario, it can acquire feature data from the second database based on the data request, and then send the acquired feature data to the data user. The data processing system provided in the embodiment of the present invention can perform standardized processing on multi-modal data, which enables the data user to directly request and obtain the data it needs without having to collect and process the data.

[0181] Finally, it should also be noted that in this text, 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 actual relationship or order between these entities or operations. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or device comprising the said element.

[0182] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. For the same or similar parts among the various embodiments, reference may be made to each other.

[0183] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. 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 invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features disclosed herein.

Claims

1. A data processing method, characterized in that, it includes: Collect various types of data in the target scenario and store the collected data in the first database; Based on the data requirements of the target scenario, obtain feature data for each piece of data of each type stored in the first database to obtain the feature data corresponding to each piece of data of each type; Store the feature data corresponding to each piece of data of each type in the form of a structured data table in the second database, where the structured data table corresponding to each piece of data of each type includes the feature data corresponding to the corresponding data and identification information; Receive a data request containing identification information sent by a data user in the target scenario, obtain feature data from the second database based on the identification information in the data request, and send the obtained feature data to the data user.

2. The data processing method according to claim 1, characterized in that, the various types of data include some or all of video data, audio data, text data, picture data, and user behavior data; the obtaining of feature data for each piece of data of each type stored in the first database based on the data requirements of the target scenario includes: If the various types of data include video data, then process each piece of video data into video data in a unified video format, and obtain video feature data for each piece of processed video data; If the various types of data include audio data, then process each piece of audio data into audio data in a unified audio format, and obtain audio feature data for each piece of processed audio data; If the various types of data include text data, then process each piece of text data into text data in a unified text format, and obtain text feature data for each piece of processed text data; If the various types of data include picture data, then process each piece of picture data into picture data in a unified picture format, and obtain picture feature data for each piece of processed picture data; If the various types of data include user behavior data, then obtain user behavior feature data for each piece of user behavior data.

3. The data processing method according to claim 1, characterized in that, the various types of data include some or all of video data, audio data, text data, picture data, and user behavior data; If the various types of data include video data, then the video feature data obtained for a piece of video data includes the representation vector of the object involved in the piece of video data and the concrete information of the piece of video data; If the various types of data include audio data, then the audio feature data obtained for a piece of audio data includes the recognized text of the piece of audio data and the concrete information of the piece of audio data; If the various types of data include text data, then the text feature data obtained for a piece of text data includes the relevant information of the object involved in the piece of text data; If the various types of data include picture data, then the picture feature data obtained for a piece of picture data includes the information of the picture itself and the information inferred from the picture; If the multiple types of data include user behavior data, the user behavior feature data obtained for a piece of user behavior data includes the relationship data between the user and the object and user data.

4. The data processing method according to claim 1, wherein, the multiple types of data include some or all of video data, audio data, text data, picture data, and user behavior data; if the multiple types of data include video data, the object identifier of the object involved in the corresponding video data is included in the structured data table corresponding to each video data; if the multiple types of data include audio data, the object identifier of the object involved in the corresponding audio data is included in the structured data table corresponding to each audio data; if the multiple types of data include text data, the object identifier of the object involved in the corresponding text data is included in the structured data table corresponding to each text data; if the multiple types of data include picture data, the object identifier of the object involved in the corresponding picture data is included in the structured data table corresponding to each picture data; if the multiple types of data include user behavior data, the global auto-increment identifier, user identifier, and related object identifier corresponding to the user behavior data are included in the structured data table corresponding to each user behavior data.

5. The data processing method according to claim 4, wherein, the multiple types of data include some or all of video data, audio data, text data, picture data, and user behavior data; the identification information in the data request is a global auto-increment identifier range; the obtaining of the feature data from the second database based on the identification information in the data request includes: determining a target structured data table from the structured data tables corresponding to each piece of user behavior data based on the global auto-increment identifier range in the data request; obtaining the user identifier and the object identifier from the target structured data table, the obtained user identifier being used as the target user identifier, and the obtained object identifier being used as the target object identifier; obtaining the feature data requested by the data request from the second database based on the target user identifier and the target object identifier.

6. The data processing method according to claim 5, wherein, the multiple types of data include video data, audio data, text data, picture data, and user behavior data; the obtaining of the feature data requested by the data request from the second database based on the target user identifier and the target object identifier includes: obtaining video feature data, audio feature data, text feature data, and picture feature data from the second database based on the target object identifier; obtaining user behavior feature data from the second database based on the target user identifier.

7. The data processing method according to claim 6, wherein, the obtaining of video feature data, audio feature data, text feature data, and picture feature data from the second database based on the target object identifier includes: Call the video feature acquisition interface to obtain video feature data from the structured data tables corresponding to each video data and containing the structured data table of the target object identifier; Call the audio feature acquisition interface to obtain audio feature data from the structured data tables corresponding to each audio data and containing the structured data table of the target object identifier; Call the text feature acquisition interface to obtain text feature data from the structured data tables corresponding to each text data and containing the structured data table of the target object identifier; Call the image feature acquisition interface to obtain image feature data from the structured data tables corresponding to each image data and containing the structured data table of the target object identifier; Based on the target user identifier, obtaining user behavior feature data from the second database includes: Call the user behavior feature acquisition interface to obtain user behavior feature data from the structured data tables corresponding to each user behavior data and containing the structured data table of the target user identifier.

8. A data processing system, characterized in that, comprising: A data collection subsystem, a data processing subsystem, and a data acquisition subsystem; The data collection subsystem is used to collect various types of data in the target scenario and store the collected data in the first database; The data processing subsystem is used to obtain feature data for each piece of data of each type stored in the first database based on the data requirements of the target scenario, so as to obtain the feature data corresponding to each piece of data of each type; and store the feature data corresponding to each piece of data of each type in the second database in the form of a structured data table, where the structured data table corresponding to each piece of data of each type includes the feature data and identification information corresponding to the corresponding data; The data acquisition subsystem is used to receive a data request containing identification information sent by a data user in the target scenario, obtain feature data from the second database based on the identification information in the data request, and send the obtained feature data to the data user.

9. The data processing system according to claim 8, characterized in that, The various types of data include some or all of video data, audio data, text data, image data, and user behavior data, and the data processing subsystem correspondingly includes some or all of a video data processing module, an audio data processing module, an image data processing module, a text data processing module, and a user behavior data processing module; The video data processing module is used to obtain video feature data for each piece of video data. Among them, the video feature data obtained for a piece of video data includes the representation vector of the object involved in the piece of video data and the concrete information of the piece of video data; The audio data processing module is used to obtain audio feature data for each piece of audio data. Among them, the audio feature data obtained for a piece of audio data includes the recognized text of the piece of audio data and the concrete information of the piece of audio data; The text data processing module is used to obtain text feature data for each piece of text data. Among them, the text feature data obtained for a piece of text data includes the relevant information of the objects involved in the text data; The picture data processing module is used to obtain picture feature data for each piece of picture data. Among them, the picture feature data obtained for a piece of picture data includes the information of the picture itself and the information inferred from the picture; The user behavior data processing module is used to obtain user behavior feature data for each piece of user behavior data. Among them, the user behavior feature data obtained for a piece of user behavior data includes the relationship data between the user and the object and user data.

10. The data processing system according to claim 8, wherein, the multiple types of data include some or all of video data, audio data, text data, picture data, and user behavior data; the structured data table corresponding to each piece of user behavior data contains the global auto-increment identifier, user identifier, and related object identifier corresponding to the user behavior data; the structured data table corresponding to each piece of data of other types contains the object identifier of the object involved in the corresponding data; when the data acquisition subsystem acquires feature data from the second database based on the identifier information in the data request, it is specifically used for: determining the target structured data table from the structured data tables corresponding to each piece of user behavior data based on the global auto-increment identifier range in the data request; acquiring the user identifier and object identifier from the target structured data table, taking the acquired user identifier as the target user identifier, and taking the acquired object identifier as the target object identifier; acquiring the feature data requested by the data request from the second database based on the target user identifier and the target object identifier.

Images