Data processing method, apparatus and device

Abstract

Embodiments of the present specification provide a data processing method, device and equipment, wherein the method comprises: obtaining picture credential data corresponding to triggering execution of a target service by a target user; obtaining a pre-trained service processing model corresponding to the target service; determining, through a second module of the pre-trained service processing model, sub-picture credential data of different regions containing text information in the picture credential data, and performing text feature extraction processing on the sub-picture credential data respectively to obtain text feature information corresponding to the picture credential data; performing, through a first module of the pre-trained service processing model, credential analysis processing on the text feature information to obtain a credential analysis result for the picture credential data; and determining, based on the credential analysis result, a service processing result corresponding to the target user triggering execution of the target service.

Description

Technical Field

[0001] This document relates to the field of data processing technology, and in particular to a data processing method, apparatus and equipment. Background Technology

[0002] With the rapid development of the internet industry, network operators are providing users with an increasing variety and number of services. As a result, how to parse the service credentials of users when using these services in order to better provide them with services (such as fast and accurate identity authentication based on the parsing results of service credentials to protect users' privacy data from being leaked) has become a focus of attention for network operators.

[0003] Business vouchers can be parsed using preset voucher parsing rules corresponding to the business to be processed. For example, the voucher can be parsed to determine whether it contains risky keywords that violate regulations. However, as the types of business become increasingly diverse, parsing business vouchers using preset voucher parsing rules can lead to low accuracy and efficiency in voucher parsing. Therefore, a solution is needed that can improve the accuracy and efficiency of business voucher parsing to ensure accurate business processing. Summary of the Invention

[0004] The purpose of the embodiments in this specification is to provide a solution that can improve the accuracy and efficiency of business credential parsing, so as to accurately perform business processing.

[0005] To achieve the above technical solution, the embodiments in this specification are implemented as follows:

[0006] In a first aspect, an embodiment of this specification provides a data processing method, comprising: acquiring image credential data corresponding to a target service triggered by a target user; acquiring a pre-trained service processing model corresponding to the target service, wherein the service processing model is obtained by training a service processing model constructed by a first module and a pre-trained second module using first image credential data corresponding to the target service, the second module being obtained by training a module constructed by a preset deep learning algorithm based on sub-image credential data of different regions in the second image credential data and text description data corresponding to the second image credential data, the first module being used to perform credential parsing processing on the output result of the second module; determining sub-image credential data of different regions containing text information in the image credential data through the second module of the pre-trained service processing model, and performing text feature extraction processing on the sub-image credential data respectively to obtain text feature information corresponding to the image credential data; performing credential parsing processing on the text feature information through the first module of the pre-trained service processing model to obtain a credential parsing result for the image credential data; and determining the service processing result corresponding to the target service triggered by the target user based on the credential parsing result.

[0007] Secondly, embodiments of this specification provide a data processing apparatus, the apparatus comprising: a first acquisition module, configured to acquire image credential data corresponding to a target service triggered by a target user; and a model acquisition module, configured to acquire a pre-trained business processing model corresponding to the target service, wherein the business processing model is obtained by training a business processing model constructed from a first module and a pre-trained second module using the first image credential data corresponding to the target service, the second module being obtained by training a module constructed from a preset deep learning algorithm based on sub-image credential data of different regions in the second image credential data and text description data corresponding to the second image credential data; the first module is used for... The output of the second module is subjected to credential parsing processing; the first processing module is used to determine the sub-image credential data containing text information in different regions of the image credential data through the second module of the pre-trained business processing model, and to perform text feature extraction processing on the sub-image credential data respectively to obtain the text feature information corresponding to the image credential data; the second processing module is used to perform credential parsing processing on the text feature information through the first module of the pre-trained business processing model to obtain the credential parsing result for the image credential data; the result determination module is used to determine the business processing result corresponding to triggering the execution of the target business for the target user based on the credential parsing result.

[0008] Thirdly, embodiments of this specification provide a data processing device, the data processing device comprising: a processor; and a memory arranged to store computer-executable instructions, wherein the executable instructions, when executed, cause the processor to: acquire image credential data corresponding to a target service triggered by a target user; acquire a pre-trained service processing model corresponding to the target service, the service processing model being obtained by training a service processing model constructed from a first module and a pre-trained second module using first image credential data corresponding to the target service, the second module being based on sub-image credential data of different regions in the second image credential data, and text description data corresponding to the second image credential data, for... The system is trained using modules constructed from a pre-defined deep learning algorithm. The first module performs credential parsing processing on the output of the second module. The second module of the pre-trained business processing model determines sub-image credential data containing text information in different regions of the image credential data, and performs text feature extraction processing on each sub-image credential data to obtain the text feature information corresponding to the image credential data. The first module of the pre-trained business processing model performs credential parsing processing on the text feature information to obtain the credential parsing result for the image credential data. Based on the credential parsing result, the system determines the business processing result corresponding to triggering the execution of the target business for the target user.

[0009] Fourthly, embodiments of this specification provide a storage medium for storing computer-executable instructions. When executed, these instructions implement the following process: acquiring image credential data corresponding to a target service triggered by a target user; acquiring a pre-trained business processing model corresponding to the target service, wherein the business processing model is trained using first image credential data corresponding to the target service, and the second module is trained using a module constructed by a preset deep learning algorithm based on sub-image credential data of different regions in the second image credential data and text description data corresponding to the second image credential data; the first module is used to perform credential parsing processing on the output of the second module; using the second module of the pre-trained business processing model, determining sub-image credential data of different regions containing text information in the image credential data, and performing text feature extraction processing on the sub-image credential data respectively to obtain text feature information corresponding to the image credential data; using the first module of the pre-trained business processing model, performing credential parsing processing on the text feature information to obtain a credential parsing result for the image credential data; and based on the credential parsing result, determining the business processing result corresponding to the target service triggered by the target user. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram of a data processing system described in this specification;

[0012] Figure 2A This is a flowchart illustrating an embodiment of a data processing method described in this specification;

[0013] Figure 2B This is a schematic diagram of the processing procedure of one data processing method described in this specification;

[0014] Figure 3 This is a schematic diagram of image voucher data in this specification;

[0015] Figure 4 This is a schematic diagram illustrating the processing procedure of another data processing method described in this specification;

[0016] Figure 5 This is a schematic diagram of the training process of one of the second modules in this specification;

[0017] Figure 6 This is a schematic diagram illustrating the processing procedure of one business processing model described in this specification.

[0018] Figure 7 This is a schematic diagram of the structure of an embodiment of a data processing device according to this specification;

[0019] Figure 8 This is a schematic diagram of the structure of a data processing device described in this specification. Detailed Implementation

[0020] This specification provides a data processing method, apparatus, and device through its embodiments.

[0021] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0022] The technical solutions in this specification can be applied to data processing systems, such as... Figure 1 As shown, the data processing system can have terminal devices and servers. The servers can be independent servers or server clusters composed of multiple servers. The terminal devices can be devices such as personal computers or mobile terminal devices such as mobile phones and tablets.

[0023] The data processing system may include n terminal devices and m servers, where n and m are positive integers greater than or equal to 1. The servers may be the backend servers of a certain application, and the terminal devices may be the client devices of the application. The application may be an application that can provide users with services such as resource transfer, video viewing, and instant messaging.

[0024] The terminal device can collect image credential data corresponding to a user triggering a certain service and send the collected image credential data to the server. The server can then obtain the image credential data corresponding to the user-triggered service and a pre-trained service processing model. Using the second module of this pre-trained model, the server performs text feature extraction on different regions of the image credential data to obtain the corresponding text feature information. Then, the server uses the first module of the pre-trained model to perform credential parsing on the text feature information, obtaining the credential parsing result for the image credential data. Finally, based on the credential parsing result, the server can determine the service processing result corresponding to the target user triggering the target service.

[0025] In addition, the server can also store image voucher data collected by terminal devices, so that when the model training cycle arrives, the business processing model can be trained using the stored image voucher data to obtain the trained business processing model.

[0026] In addition, the data processing system can also include a central server (e.g., server 1). The central server can receive historical image voucher data stored by terminal devices and / or servers. Based on this historical image voucher data, it determines second image voucher data to train the second module in the business processing model, resulting in a trained second module. Then, the central server can obtain the first image voucher data corresponding to the target business from the historical image voucher data and train the business processing model (i.e., the business processing model constructed from the first module and the pre-trained second module) using this first image voucher data, resulting in a trained business processing model. The central server can then send the model parameters of the trained business processing model to other servers in the data processing system. These other servers can update their local business processing models based on the received model parameters, obtaining trained business processing models. These trained models can then be used to perform voucher parsing processing on the image voucher data corresponding to the target business, obtaining the voucher parsing results. Finally, the business processing result is determined based on the voucher parsing results. This avoids business interruptions caused by the need to train the business processing model, thus meeting the user's business needs.

[0027] Based on the above data processing system architecture, the data processing methods in the following embodiments can be implemented.

[0028] Example 1

[0029] like Figure 2A and Figure 2B As shown in the embodiments of this specification, a data processing method is provided. The execution subject of this method can be a server, which can be a standalone server or a server cluster composed of multiple servers. The method specifically includes the following steps:

[0030] In S202, obtain the image credential data corresponding to the target user triggering the execution of the target business.

[0031] The target business can be any business that may involve data risks such as leakage of user privacy data. For example, the target business can be a resource transfer business, an identity verification business, an account registration business, or a resource accounting business. The image credential data can be credential data of an image type that can indicate the target user's triggering of the target business. For example, the image credential data can include image-type credential data used to indicate the target user's identity, image-type credential data used to indicate the processing process or result of the target business, etc. Specifically, taking the target business as a resource transfer business as an example, the image credential data can include the target user's identity verification image (i.e., the image entered by the target user that can prove their identity), and the resource transfer page corresponding to the target user's triggering of the resource transfer business (i.e., the page containing information such as the number of resources transferred, the method of resource transfer, and the corresponding information of resource transfer), etc. Taking the target business as an account registration business as an example, the image credential data can include the target user's identity verification image data, and the account registration page corresponding to the target user's triggering of the account registration business (i.e., the page containing information such as the registered account, registered username, registered password, and verification method), etc.

[0032] In practice, with the rapid development of the internet industry, network operators are providing users with an increasing variety and number of services. A key focus for network operators is how to parse user credentials used in these services to better provide them (e.g., fast and accurate authentication based on credential parsing results to protect user privacy). While preset credential parsing rules corresponding to the service being processed can be used to parse credentials (e.g., to check for illegal keywords), the increasing variety of service types leads to inaccurate and inefficient parsing. Therefore, a solution is needed to improve the accuracy and efficiency of credential parsing for accurate service processing. This specification provides a technical solution to address these issues, as detailed below.

[0033] Taking the target service as a resource transfer service as an example, the target user can trigger the resource transfer service through a resource transfer application installed on the terminal device. That is, when the terminal device detects that the target user has triggered the resource transfer service through a resource transfer application, the terminal device can collect the target user's identity verification image data (such as the user login page when the target user logs into the resource transfer application, the document image data entered by the target user that can be used for identity verification, etc.), as well as the resource transfer page corresponding to the target user triggering the execution of the resource transfer service.

[0034] The terminal device can identify the collected image data as image credential data corresponding to the target user triggering the execution of the target service, and send the image credential data to the server. In other words, the server can receive the image credential data corresponding to the target user triggering the execution of the target service.

[0035] Furthermore, the above-mentioned method for obtaining image voucher data is an optional and feasible method of determination. In actual application scenarios, there can be a variety of different methods of obtaining data, which may vary depending on the actual application scenario. This specification does not specifically limit the embodiments in this way.

[0036] In S204, obtain the pre-trained business processing model corresponding to the target business.

[0037] The business processing model can be obtained by training the business processing model constructed by the first module and the pre-trained second module using the first image voucher data corresponding to the target business. The second module can be obtained by training the module constructed by the preset deep learning algorithm based on the sub-image voucher data of different regions in the second image voucher data and the text description data corresponding to the second image voucher data. The first module can be used to perform voucher parsing processing on the output results of the second module.

[0038] In implementation, the server can select second image voucher data from pre-stored historical image voucher data based on a preset model update cycle (e.g., historical image voucher data can be determined as the second image voucher data). Then, the server can pre-train the second module based on the second image voucher data to obtain the trained second module.

[0039] When training the second module, the server can divide the second image credential data into regions to obtain multiple sub-image credential data. For example, the server can divide the second image credential data into regions based on the type of elements contained within it. Specifically, taking the second image credential data as the resource transfer page corresponding to the resource transfer service triggered by the target user as an example... Figure 3 As shown, the second image voucher data can contain image type elements, text type elements, video type elements, and audio type elements. Based on the different elements mentioned above, the server can divide the second image voucher data into sub-image voucher data 1 corresponding to region 1, sub-image voucher data 2 corresponding to region 2, sub-image voucher data 3 corresponding to region 3, and sub-image voucher data 4 corresponding to region 4.

[0040] The method for determining the sub-image voucher data described above is an optional and implementable method. In actual application scenarios, there can be a variety of different acquisition methods, which may vary depending on the actual application scenario. This specification does not specifically limit the methods used in this embodiment.

[0041] The server can train the second module using sub-image voucher data from different regions of the second image voucher data, as well as the corresponding text description data. This allows the sub-image voucher data for each region of the second image voucher data to be aligned with the corresponding sub-text description data in the text description data. This improves the second module's ability to perceive various types of information in the image voucher data, thus enhancing the alignment between the second image voucher data and the text description data at a finer granular level. As a result, the trained second module can extract more detailed elements from the image voucher data.

[0042] After obtaining the trained second module, the server can select the first image voucher data corresponding to the target business from the historical image data, and use the first image voucher data to train the business processing model to obtain the trained business processing model.

[0043] In S206, the second module of the pre-trained business processing model determines the sub-image voucher data containing text information in different regions of the image voucher data, and performs text feature extraction processing on the sub-image voucher data to obtain the text feature information corresponding to the image voucher data.

[0044] In implementation, since the second module is trained based on sub-image voucher data from different regions of the second image voucher data and the corresponding text description data, the trained second module can extract more refined elements from the image voucher data. Therefore, the server can use the second module of the pre-trained business processing model to determine the sub-image voucher data from different regions containing text information in the image voucher data, and perform text feature extraction processing on the sub-image voucher data to obtain the text feature information corresponding to the image voucher data.

[0045] In S208, the first module of the pre-trained business processing model performs voucher parsing processing on the text feature information to obtain the voucher parsing result for the image voucher data.

[0046] The first module can be a module built based on a preset machine learning algorithm for credential parsing of text feature information. The credential parsing result can include the result obtained by parsing the image credential data based on the business processing requirements corresponding to the target business. For example, if the target business is a resource transfer business, the business processing requirement corresponding to the target business can be a risk detection requirement. In this case, the credential parsing result can include the result obtained by risk detection processing of the image credential data. Or, if the target business is an account registration business, the business processing requirements corresponding to the target business can be a risk detection requirement and an account information extraction requirement. In this case, the credential parsing result can include the result obtained by risk detection processing of the image credential data and the result obtained by account information extraction processing of the image credential data.

[0047] In implementation, the server can build a corresponding first module based on the business processing requirements of the target business. For example, if the business processing requirements of the target business can be risk detection requirements, then the server can build a first module based on a preset classification algorithm. That is, the first module can be used to perform classification processing based on text feature information to obtain the risk classification result (i.e., the voucher parsing result) of the image voucher data.

[0048] In addition, the business processing requirements of the target business can be varied. The corresponding first module can be constructed according to the actual business processing requirements of the target business. This specification does not specifically limit this in the embodiments.

[0049] In S210, based on the credential parsing result, the business processing result corresponding to the target business triggered for the target user is determined.

[0050] In practice, taking the risk classification result of image credential data as the result of credential parsing as an example, if it is determined that the image credential data is at risk based on the risk classification result, then the server can suspend the execution of the target business and determine the result of the suspension as the business processing result that triggers the execution of the target business for the target user.

[0051] Alternatively, if the risk classification results determine that the image credential data does not pose a risk, the server can continue to execute the target business based on the image credential data and obtain the corresponding business processing result.

[0052] This specification provides a data processing method that acquires image credential data corresponding to a target user triggering the execution of a target service, and acquires a pre-trained business processing model corresponding to the target service. The business processing model is obtained by training a business processing model constructed from a first module and a pre-trained second module using the first image credential data corresponding to the target service. The second module is obtained by training a module constructed from a preset deep learning algorithm based on sub-image credential data of different regions in the second image credential data and text description data corresponding to the second image credential data. The first module is used to perform credential parsing processing on the output of the second module. Through the second module of the pre-trained business processing model, the sub-image credential data of different regions containing text information in the image credential data are determined, and text feature extraction processing is performed on the sub-image credential data to obtain the text feature information corresponding to the image credential data. Through the first module of the pre-trained business processing model, credential parsing processing is performed on the text feature information to obtain the credential parsing result for the image credential data. Based on the credential parsing result, the business processing result corresponding to the target user triggering the execution of the target service is determined. Since the second module is trained based on sub-image voucher data from different regions of the second image voucher data and corresponding text description data, it can extract more refined elements from the image voucher data. This allows the pre-trained business processing model to perform text feature extraction on these refined elements, obtaining text feature information and improving the accuracy of subsequent voucher parsing by the first module. Furthermore, after obtaining the trained second module, different training image voucher data can be acquired for different business operations to train the business processing model. The server can use the first image voucher data corresponding to the target business to train the business processing model constructed from the first module and the pre-trained second module. This improves the training efficiency of the business processing model corresponding to the target business, enhances the efficiency and accuracy of voucher parsing, and ultimately improves the efficiency and accuracy of determining the business processing results triggered for the target user.

[0053] Example 2

[0054] like Figure 4 As shown in the embodiments of this specification, a data processing method is provided. The execution subject of this method can be a server, which can be a standalone server or a server cluster composed of multiple servers. The method specifically includes the following steps:

[0055] In S202, obtain the image credential data corresponding to the target user triggering the execution of the target business.

[0056] In S402, the second image voucher data and the corresponding text description data are obtained.

[0057] The second image voucher data can be larger than the first image voucher data. For example, the first image voucher data can be image voucher data corresponding to the target business, while the second image voucher data can be image voucher data corresponding to multiple different businesses. This means the second module can be pre-trained using a large amount of training sample data to improve its training performance. The text description data can be data used to describe the content contained in the second image voucher data. For example, assuming the second image voucher data is as follows... Figure 3 The resource transfer page shown can have the following text description for the second image credential data: "This is a screenshot that says: Welcome to the resource transfer application, the number of resources transferred is xx, the resource transfer object is xx, and a video ad and a voice ad are playing."

[0058] In S404, text recognition processing is performed on the second image voucher data to obtain the text recognition result. Based on the text recognition result, the sub-image data of different regions containing text information in the second image voucher data are determined as the sub-image voucher data in the second image voucher data.

[0059] In implementation, the server can perform text recognition processing on the second image voucher data based on a preset text recognition algorithm to obtain the text recognition result. Then, the server can perform entity recognition processing, segmentation processing, or classification processing on the text recognition result, and divide the second image voucher data into multiple sub-image data. Finally, the server can determine the sub-image data containing different regions of text information in the second image voucher data as the sub-image voucher data in the second image voucher data.

[0060] For example, using the second image voucher data as... Figure 3 Taking the image data shown as an example, assume that the text recognition results for the second image credential data include "Welcome to the Resource Transfer Application", "Please enter:", "Resource Transfer Quantity", "Resource Transfer Object", "Video Playback Area", and "Audio Playback Area". The server can determine the sub-image corresponding to each of the above text recognition results as the sub-image credential data in the second image credential data. That is, the sub-image credential data may include sub-image credential data 1 corresponding to area 1, sub-image credential data 2 corresponding to area 2, sub-image credential data 3 corresponding to area 3, and sub-image credential data 4 corresponding to area 4.

[0061] Alternatively, the server can classify the above text recognition results. The resulting classification results could be Category 1, which includes "Welcome to the Resource Transfer Application"; Category 2, which includes "Please enter:", "Resource Transfer Quantity", "Resource Transfer Object", and "Video Playback Area"; and Category 3, which includes both "Video Playback Area" and "Audio Playback Area". The server can obtain the region corresponding to each category in the second image credential data and determine the sub-image corresponding to that region as the sub-image credential data in the second image credential data. That is, the sub-image credential data can include sub-image credential data 1 corresponding to region 1, sub-image credential data 2 corresponding to region 2, and sub-image credential data 3 corresponding to regions 3 and 4.

[0062] The method for determining the sub-image voucher data described above is an optional and implementable method. In actual application scenarios, there can be a variety of different acquisition methods, which may vary depending on the actual application scenario. This specification does not specifically limit the methods used in this embodiment.

[0063] In S406, the second module performs text feature extraction processing on the sub-image voucher data in the second image voucher data to obtain the first text feature.

[0064] In S408, the third module performs feature extraction processing on the text description data to obtain the second text features.

[0065] In S410, a first training error value is determined based on the first text feature and the second text feature, and based on the first training error value, it is determined whether the second module has converged. If the second module has not converged, the second module and the third module are trained again based on the sub-image voucher data in the second image voucher data and the text description data until the second module converges, and the trained second module is obtained.

[0066] In implementation, taking the model built using the second and third modules as a base model as an example, the second module can perform text feature extraction through visual encoding, and the third module can perform feature extraction through text encoding. For example... Figure 5 As shown, the server can input the sub-image voucher data from the second image voucher data into the second module to obtain the first text feature, and input the text description data into the third module to obtain the second text feature corresponding to the text description data.

[0067] In this way, by using the first training error value determined by the first and second text features to compare and learn the second and third modules, the alignment between the sub-image voucher data and the text description data can be improved, thereby enhancing the ability of the trained second module to extract text features from the image voucher data at a finer granular level.

[0068] In S412, the first image voucher data and the first voucher parsing result corresponding to the first image voucher data are obtained.

[0069] In implementation, the server can obtain the first image credential data corresponding to the business type of the target business, as well as the first credential parsing result corresponding to the first image credential data, based on the business type of the target business.

[0070] In S414, based on the pre-trained second module, text feature extraction is performed on the sub-image voucher data of different regions in the first image voucher data to obtain the third text feature.

[0071] In implementation, because the second module is trained through comparative learning using sub-image voucher data and corresponding text description data, it possesses the ability to extract text features from image voucher data at a finer granularity. Therefore, as... Figure 6 As shown, the server can input the first image data into the second module, so that the second module can perform text feature extraction processing on the sub-image voucher data of different regions in the first image data to obtain the parsing result of the first voucher.

[0072] Alternatively, the server can perform text recognition processing on the first image voucher data to obtain the text recognition result. Based on the text recognition result, the server can identify the sub-image data containing text information in different regions of the first image voucher data as sub-image voucher data within the first image voucher data. Then, the server can input the sub-image voucher data from the first image voucher data into the second module to obtain the first voucher parsing result.

[0073] In S416, based on the first module, the third text feature is processed by voucher parsing to obtain the second voucher parsing result, and based on the first voucher parsing result and the second voucher parsing result, the second training error value is determined.

[0074] In implementation and practical application, different first modules can be constructed for different business operations. For example, the first module may include a first sub-module for extracting and processing business information and a second sub-module for detecting image tampering. The credential parsing result may include the business information extraction result and the image tampering detection result. Therefore, the processing method of S416 above can be referred to the following steps one to three:

[0075] Step 1: Based on the first sub-module of the first module, perform business information extraction processing on the third text features to obtain the first business information extraction result in the second voucher parsing result.

[0076] The first business information extraction result may include the business information required for processing the target business from the first image voucher data. For example, assuming the first image voucher data is as follows: Figure 3 The image data shown indicates that the target business is resource transfer. Therefore, the first business information extraction result can include the number of resources transferred, the time of resource transfer, and the object of resource transfer.

[0077] Step two: Based on the second sub-module of the first module, perform image tampering detection processing on the third text features to obtain the first image tampering detection result in the second credential parsing result.

[0078] Specifically, if the second submodule determines that the first image credential data has been tampered with, the first image tampering detection result may include the image data of the tampered area within the first image credential data. If the second submodule determines that the first image credential data has not been tampered with, the first image tampering detection result may include the untampered portion of the first image credential data. For example, suppose the first image credential data is as follows... Figure 3 If the image data shown is processed by the image tampering detection process of the second submodule and it is determined that the first image credential data has been tampered with, and that region 2 is the tampered region, then the first image tampering detection result output by the second submodule can be the image data corresponding to region 2.

[0079] In implementation, such as Figure 6 As shown, the first submodule can perform business information extraction processing on the third text features output by the second module to obtain the first business information extraction result, and the second submodule can perform image tampering detection processing on the third text features output by the second module to obtain the first image tampering detection result.

[0080] Step 3: Based on the results of the first business information extraction, the first image tampering detection, and the first voucher parsing, determine the second training error value.

[0081] In implementation, the server can determine a first sub-error value based on the first business information extraction result and the business information extraction result contained in the first credential parsing result, and determine a second sub-error value based on the first image tampering detection result and the image tampering detection result in the first credential parsing result. Finally, the server can determine a second training error value based on the first and second sub-error values.

[0082] In S418, based on the second training error value, it is determined whether the business processing model has converged. If the business processing model has not converged, the business processing model is trained again based on the first image voucher data and the parsing result of the first voucher until the business processing model converges, and the trained business processing model is obtained.

[0083] In S204, obtain the pre-trained business processing model corresponding to the target business.

[0084] In S206, the second module of the pre-trained business processing model determines the sub-image voucher data containing text information in different regions of the image voucher data, and performs text feature extraction processing on the sub-image voucher data to obtain the text feature information corresponding to the image voucher data.

[0085] In S208, the first module of the pre-trained business processing model performs voucher parsing processing on the text feature information to obtain the voucher parsing result for the image voucher data.

[0086] In S422, based on the image tampering detection results in the credential parsing results, the risk detection results for the image credential data are determined.

[0087] In implementation and practical applications, the processing method of S422 can vary. The following is one optional implementation method, which can be found in steps one through three below:

[0088] Step 1: If the image credential data is determined to be tampered based on the image tampering detection results, then the tampered area in the image credential data is identified based on the image tampering detection results.

[0089] Step 2: Obtain the business information corresponding to the tampered area from the business information extraction results.

[0090] Step 3: Based on the tampered area and the corresponding business information, perform risk detection processing on the image credential data to obtain the risk detection results for the image credential data.

[0091] In implementation, assuming the first image voucher data is as follows: Figure 3 If the image data shown is processed by the image tampering detection process of the second submodule, and it is determined that the first image credential data has been tampered with, and that region 2 is the tampered region, then the first image tampering detection result output by the second submodule can be the image data corresponding to region 2. The business information extraction result output by the second submodule for the image credential result can include the number of resources transferred, the time of resource transfer, and the object of resource transfer.

[0092] The server can obtain the business information corresponding to the tampered area from the business information extraction results, that is, the information in the image data corresponding to area 2 that matches the business information extraction results, namely the number of resources transferred and the resource transfer objects.

[0093] The server can perform risk detection on image credential data based on the quantity and object of resource transfers, and obtain a risk detection result for the image credential data. For example, the server can obtain the resource transfer threshold set by the target user for the resource transfer object, and determine the risk detection result based on the resource transfer threshold and the quantity of resource transfers. If the quantity of resource transfers is greater than the resource transfer threshold, the risk detection result for the image credential data can be considered as having risk; if the quantity of resource transfers is not greater than the resource transfer threshold, the risk detection result for the image credential data can be considered as not having risk.

[0094] The above-described method for risk detection processing of image voucher data is an optional and feasible risk detection processing method. In actual application scenarios, there are many other risk detection processing methods. Different risk detection processing methods can be selected according to different actual application scenarios. This specification does not specifically limit this method in the embodiments.

[0095] In S424, based on the business information extraction results from the risk detection results and the credential parsing results, the business processing result corresponding to the target business triggered for the target user is determined.

[0096] In practice, if the server determines that the image credential data does not pose a risk based on the risk detection results, it can extract the business information and execute the target business to obtain the business processing result corresponding to the target user.

[0097] This specification provides a data processing method that acquires image credential data corresponding to a target user triggering the execution of a target service, and acquires a pre-trained business processing model corresponding to the target service. The business processing model is obtained by training a business processing model constructed from a first module and a pre-trained second module using the first image credential data corresponding to the target service. The second module is obtained by training a module constructed from a preset deep learning algorithm based on sub-image credential data of different regions in the second image credential data and text description data corresponding to the second image credential data. The first module is used to perform credential parsing processing on the output of the second module. Through the second module of the pre-trained business processing model, the sub-image credential data of different regions containing text information in the image credential data are determined, and text feature extraction processing is performed on the sub-image credential data to obtain the text feature information corresponding to the image credential data. Through the first module of the pre-trained business processing model, credential parsing processing is performed on the text feature information to obtain the credential parsing result for the image credential data. Based on the credential parsing result, the business processing result corresponding to the target user triggering the execution of the target service is determined. Since the second module is trained based on sub-image voucher data from different regions of the second image voucher data and corresponding text description data, it can extract more refined elements from the image voucher data. This allows the pre-trained business processing model to perform text feature extraction on these refined elements, obtaining text feature information and improving the accuracy of subsequent voucher parsing by the first module. Furthermore, after obtaining the trained second module, different training image voucher data can be acquired for different business operations to train the business processing model. The server can use the first image voucher data corresponding to the target business to train the business processing model constructed from the first module and the pre-trained second module. This improves the training efficiency of the business processing model corresponding to the target business, enhances the efficiency and accuracy of voucher parsing, and ultimately improves the efficiency and accuracy of determining the business processing results triggered for the target user.

[0098] Example 3

[0099] The above describes the data processing method provided in the embodiments of this specification. Based on the same idea, the embodiments of this specification also provide a data processing device, such as... Figure 7 As shown.

[0100] The data processing device includes: a first acquisition module 701, a model acquisition module 702, a first processing module 703, a second processing module 704, and a result determination module 705, wherein:

[0101] The first acquisition module 701 is used to acquire image credential data corresponding to the target user triggering the execution of the target business;

[0102] The model acquisition module 702 is used to acquire a pre-trained business processing model corresponding to the target business. The business processing model is obtained by training a business processing model constructed by a first module and a pre-trained second module using first image voucher data corresponding to the target business. The second module is obtained by training a module constructed by a preset deep learning algorithm based on sub-image voucher data of different regions in the second image voucher data and text description data corresponding to the second image voucher data. The first module is used to perform voucher parsing processing on the output result of the second module.

[0103] The first processing module 703 is used to determine the sub-image voucher data containing text information in different regions of the image voucher data through the second module of the pre-trained business processing model, and to perform text feature extraction processing on the sub-image voucher data respectively to obtain the text feature information corresponding to the image voucher data.

[0104] The second processing module 704 is used to perform voucher parsing processing on the text feature information through the first module of the pre-trained business processing model to obtain the voucher parsing result for the image voucher data.

[0105] The result determination module 705 is used to determine the business processing result corresponding to triggering the execution of the target business for the target user based on the credential parsing result.

[0106] In the embodiments described in this specification, the device further includes:

[0107] The second acquisition module is used to acquire the second image voucher data and the text description data corresponding to the second image voucher data;

[0108] The image partitioning module is used to perform text recognition processing on the second image voucher data to obtain text recognition results, and based on the text recognition results, to determine the sub-image data of different areas containing text information in the second image voucher data as sub-image voucher data in the second image voucher data;

[0109] The third processing module is used to perform text feature extraction processing on the sub-image voucher data in the second image voucher data through the second module to obtain the first text feature;

[0110] The fourth processing module is used to perform feature extraction processing on the text description data through the third module to obtain the second text features;

[0111] The first training module is used to determine a first training error value based on the first text feature and the second text feature, and to determine whether the second module has converged based on the first training error value. If the second module has not converged, the second module and the third module are trained again based on the sub-image voucher data in the second image voucher data and the text description data until the second module converges, thus obtaining the trained second module.

[0112] In the embodiments described in this specification, the device further includes:

[0113] The third acquisition module is used to acquire the first image voucher data and the first voucher parsing result corresponding to the first image voucher data;

[0114] The fifth processing module is used to extract text features from the sub-image voucher data of different regions in the first image voucher data based on the pre-trained second module, so as to obtain the third text features.

[0115] An error determination module is used to perform voucher parsing processing on the third text features based on the first module to obtain a second voucher parsing result, and to determine a second training error value based on the first voucher parsing result and the second voucher parsing result;

[0116] The second training module is used to determine whether the business processing model has converged based on the second training error value. If the business processing model has not converged, the business processing model is trained again based on the first image voucher data and the parsing result of the first voucher until the business processing model converges, thus obtaining the trained business processing model.

[0117] In this embodiment of the specification, the data volume of the second image voucher data is greater than the data volume of the first image voucher data.

[0118] In this embodiment of the specification, the first module includes a first submodule for extracting and processing business information and a second submodule for detecting image tampering. The credential parsing result includes the business information extraction result and the image tampering detection result. The error determination module is used for:

[0119] Based on the first sub-module of the first module, business information extraction processing is performed on the third text features to obtain the first business information extraction result in the second voucher parsing result;

[0120] Based on the second sub-module of the first module, image tampering detection processing is performed on the third text features to obtain the first image tampering detection result in the second credential parsing result;

[0121] Based on the results of the first business information extraction, the first image tampering detection, and the first credential parsing, the second training error value is determined.

[0122] In the embodiments of this specification, the result determination module 705 is used for:

[0123] Based on the image tampering detection results in the credential parsing results, the risk detection results for the image credential data are determined;

[0124] Based on the risk detection results and the business information extraction results from the credential parsing results, the business processing result corresponding to the target business is determined for the target user.

[0125] In the embodiments of this specification, the result determination module 705 is used for:

[0126] If the image credential data is determined to be tampered image data based on the image tampering detection result, the tampered area in the image credential data is determined based on the image tampering detection result.

[0127] Obtain the business information corresponding to the tampered area from the business information extraction results;

[0128] Based on the tampered area and the corresponding business information, risk detection processing is performed on the image credential data to obtain the risk detection result for the image credential data.

[0129] In the embodiments of this specification, the result determination module 705 is used for:

[0130] If the risk detection results determine that the image credential data does not pose a risk, the target business is executed based on the business information extraction results to obtain the business processing result corresponding to the execution of the target business for the target user.

[0131] This specification provides a data processing device that acquires image credential data corresponding to a target user triggering the execution of a target service, and acquires a pre-trained business processing model corresponding to the target service. The business processing model is obtained by training a business processing model constructed from a first module and a pre-trained second module using first image credential data corresponding to the target service. The second module is obtained by training a module constructed from a preset deep learning algorithm based on sub-image credential data of different regions in the second image credential data and text description data corresponding to the second image credential data. The first module is used to perform credential parsing processing on the output result of the second module. Through the second module of the pre-trained business processing model, the sub-image credential data of different regions containing text information in the image credential data are determined, and text feature extraction processing is performed on the sub-image credential data to obtain text feature information corresponding to the image credential data. Through the first module of the pre-trained business processing model, credential parsing processing is performed on the text feature information to obtain credential parsing results for the image credential data. Based on the credential parsing results, the business processing result corresponding to the target user triggering the execution of the target service is determined. Since the second module is trained based on sub-image voucher data from different regions of the second image voucher data and corresponding text description data, it can extract more refined elements from the image voucher data. This allows the pre-trained business processing model to perform text feature extraction on these refined elements, obtaining text feature information and improving the accuracy of subsequent voucher parsing by the first module. Furthermore, after obtaining the trained second module, different training image voucher data can be acquired for different business operations to train the business processing model. The server can use the first image voucher data corresponding to the target business to train the business processing model constructed from the first module and the pre-trained second module. This improves the training efficiency of the business processing model corresponding to the target business, enhances the efficiency and accuracy of voucher parsing, and ultimately improves the efficiency and accuracy of determining the business processing results triggered for the target user.

[0132] Example 4

[0133] Following the same line of thought, embodiments of this specification also provide a data processing device, such as... Figure 8 As shown.

[0134] Data processing devices can vary considerably due to differences in configuration or performance. They may include one or more processors 801 and memory 802, with memory 802 storing one or more application programs or data. Memory 802 can be temporary or persistent storage. The application programs stored in memory 802 may include one or more modules (not shown), each module including a series of computer-executable instructions for the data processing device. Furthermore, processor 801 may be configured to communicate with memory 802 and execute the series of computer-executable instructions stored in memory 802 on the data processing device. The data processing device may also include one or more power supplies 803, one or more wired or wireless network interfaces 804, one or more input / output interfaces 805, and one or more keyboards 806.

[0135] Specifically, in this embodiment, the data processing device includes a memory and one or more programs, wherein one or more programs are stored in the memory, and one or more programs may include one or more modules, and each module may include a series of computer-executable instructions for the data processing device, and is configured to be executed by one or more processors. The one or more programs include computer-executable instructions for performing the following:

[0136] Retrieve the image credentials data corresponding to the target user triggering the execution of the target service;

[0137] A pre-trained business processing model corresponding to the target business is obtained. The business processing model is obtained by training a business processing model constructed by a first module and a pre-trained second module using first image voucher data corresponding to the target business. The second module is obtained by training a module constructed by a preset deep learning algorithm based on sub-image voucher data of different regions in the second image voucher data and text description data corresponding to the second image voucher data. The first module is used to perform voucher parsing processing on the output result of the second module.

[0138] The second module of the pre-trained business processing model determines the sub-image voucher data containing text information in different regions of the image voucher data, and performs text feature extraction processing on the sub-image voucher data to obtain the text feature information corresponding to the image voucher data.

[0139] The first module of the pre-trained business processing model performs credential parsing processing on the text feature information to obtain credential parsing results for the image credential data.

[0140] Based on the credential parsing result, the business processing result corresponding to the target business triggered for the target user is determined.

[0141] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the data processing device embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0142] This specification provides a data processing device that acquires image credential data corresponding to a target user triggering the execution of a target service, and acquires a pre-trained business processing model corresponding to the target service. The business processing model is obtained by training a business processing model constructed from a first module and a pre-trained second module using first image credential data corresponding to the target service. The second module is obtained by training a module constructed from a preset deep learning algorithm based on sub-image credential data of different regions in the second image credential data and text description data corresponding to the second image credential data. The first module is used to perform credential parsing processing on the output result of the second module. Through the second module of the pre-trained business processing model, the device determines the sub-image credential data of different regions containing text information in the image credential data, and performs text feature extraction processing on the sub-image credential data to obtain the text feature information corresponding to the image credential data. Through the first module of the pre-trained business processing model, the device performs credential parsing processing on the text feature information to obtain the credential parsing result for the image credential data. Based on the credential parsing result, the device determines the business processing result corresponding to the target user triggering the execution of the target service. Since the second module is trained based on sub-image voucher data from different regions of the second image voucher data and corresponding text description data, it can extract more refined elements from the image voucher data. This allows the pre-trained business processing model to perform text feature extraction on these refined elements, obtaining text feature information and improving the accuracy of subsequent voucher parsing by the first module. Furthermore, after obtaining the trained second module, different training image voucher data can be acquired for different business operations to train the business processing model. The server can use the first image voucher data corresponding to the target business to train the business processing model constructed from the first module and the pre-trained second module. This improves the training efficiency of the business processing model corresponding to the target business, enhances the efficiency and accuracy of voucher parsing, and ultimately improves the efficiency and accuracy of determining the business processing results triggered for the target user.

[0143] Example 5

[0144] This specification also provides a computer-readable storage medium storing a computer program. When executed by a processor, this computer program implements the various processes of the above-described data processing method embodiments and achieves the same technical effects. To avoid repetition, it will not be described again here. The computer-readable storage medium may include, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0145] This specification provides a computer-readable storage medium that acquires image credential data corresponding to a target service triggered by a target user, and acquires a pre-trained business processing model corresponding to the target service. The business processing model is obtained by training a business processing model constructed from a first module and a pre-trained second module using first image credential data corresponding to the target service. The second module is obtained by training a module constructed from a preset deep learning algorithm based on sub-image credential data of different regions in the second image credential data and text description data corresponding to the second image credential data. The first module is used to perform credential parsing processing on the output of the second module. Through the second module of the pre-trained business processing model, the sub-image credential data of different regions containing text information in the image credential data are determined, and text feature extraction processing is performed on the sub-image credential data to obtain text feature information corresponding to the image credential data. Through the first module of the pre-trained business processing model, credential parsing processing is performed on the text feature information to obtain credential parsing results for the image credential data. Based on the credential parsing results, the business processing result corresponding to the target service triggered by the target user is determined. Since the second module is trained based on sub-image voucher data from different regions of the second image voucher data and corresponding text description data, it can extract more refined elements from the image voucher data. This allows the pre-trained business processing model to perform text feature extraction on these refined elements, obtaining text feature information and improving the accuracy of subsequent voucher parsing by the first module. Furthermore, after obtaining the trained second module, different training image voucher data can be acquired for different business operations to train the business processing model. The server can use the first image voucher data corresponding to the target business to train the business processing model constructed from the first module and the pre-trained second module. This improves the training efficiency of the business processing model corresponding to the target business, enhances the efficiency and accuracy of voucher parsing, and ultimately improves the efficiency and accuracy of determining the business processing results triggered for the target user.

[0146] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0147] In the 1990s, improvements to a technology could be clearly distinguished as either hardware improvements (e.g., improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the methodology). However, with technological advancements, many methodological improvements today can be considered direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved methodology into the hardware circuit. Therefore, it cannot be said that a methodological improvement cannot be implemented using a hardware physical module. For example, a Programmable Logic Device (PLD) (e.g., a Field Programmable Gate Array (FPGA)) is such an integrated circuit whose logic function is determined by the user programming the device. Designers can program a digital system themselves to "integrate" it onto a PLD, without needing chip manufacturers to design and manufacture dedicated integrated circuit chips. Furthermore, nowadays, instead of manually manufacturing integrated circuit chips, this programming is mostly implemented using "logic compiler" software. Similar to the software compiler used in program development, the original code before compilation must be written in a specific programming language, called a Hardware Description Language (HDL). There are many HDLs, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, and RHDL (Ruby Hardware Description Language). Currently, the most commonly used are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should understand that by simply performing some logic programming on the method flow using one of these hardware description languages ​​and programming it into an integrated circuit, the hardware circuit implementing the logical method flow can be easily obtained.

[0148] The controller can be implemented in any suitable manner. For example, it can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicon Labs C8051F320. A memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also recognize that, in addition to implementing the controller in purely computer-readable program code form, the same functionality can be achieved by logically programming the method steps to make the controller take the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the means included therein for implementing various functions can also be considered as structures within the hardware component. Alternatively, the means for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0149] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0150] For ease of description, the above apparatus is described by dividing it into various functional units. Of course, when implementing one or more embodiments of this specification, the functions of each unit can be implemented in one or more software and / or hardware.

[0151] Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, one or more embodiments of this specification may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, one or more embodiments of this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0152] The embodiments described herein are illustrated with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this specification. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0153] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0154] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0155] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0156] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0157] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0158] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0159] Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, one or more embodiments of this specification may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, one or more embodiments of this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0160] One or more embodiments of this specification can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a particular task or implement a particular abstract data type. One or more embodiments of this specification can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0161] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0162] The above description is merely an embodiment of this specification and is not intended to limit this specification. Various modifications and variations can be made to this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this specification should be included within the scope of the claims of this specification.

Claims

1. A data processing method, comprising: Retrieve the image credentials data corresponding to the target user triggering the execution of the target service; A pre-trained business processing model corresponding to the target business is obtained. This business processing model is obtained by training a business processing model constructed from a first module and a pre-trained second module using first image voucher data corresponding to the target business. The second module is obtained by training a module constructed using a preset deep learning algorithm based on a first training error value determined by first and second text features. The first text feature is obtained by extracting text features from sub-image voucher data containing text information in different regions of the second image voucher data corresponding to various businesses. The second text feature is obtained by extracting features from text description data corresponding to the second image voucher data. The first module is used to perform voucher parsing processing on the output of the second module. The second module of the pre-trained business processing model determines the sub-image voucher data containing text information in different regions of the image voucher data, and performs text feature extraction processing on the sub-image voucher data to obtain the text feature information corresponding to the image voucher data. The first module of the pre-trained business processing model performs credential parsing processing on the text feature information to obtain credential parsing results for the image credential data. Based on the credential parsing result, the business processing result corresponding to the target business triggered for the target user is determined.

2. The method according to claim 1, further comprising, before obtaining the pre-trained business processing model corresponding to the target business: Obtain the second image voucher data and the text description data corresponding to the second image voucher data; The second image voucher data is subjected to text recognition processing to obtain text recognition results. Based on the text recognition results, the sub-image data of different areas containing text information in the second image voucher data are determined as sub-image voucher data in the second image voucher data. The second module performs text feature extraction processing on the sub-image voucher data in the second image voucher data to obtain the first text feature; The third module performs feature extraction processing on the text description data to obtain the second text features; Based on the first text feature and the second text feature, a first training error value is determined, and based on the first training error value, it is determined whether the second module has converged. If the second module has not converged, the second module and the third module are trained again based on the sub-image voucher data in the second image voucher data and the text description data until the second module converges, thus obtaining the trained second module.

3. The method according to claim 2, further comprising: Obtain the first image voucher data and the first voucher parsing result corresponding to the first image voucher data. fruit; Based on the pre-trained second module, text feature extraction is performed on the sub-image voucher data of different regions in the first image voucher data to obtain the third text feature. Based on the first module, the third text features are processed by voucher parsing to obtain a second voucher parsing result, and based on the first voucher parsing result and the second voucher parsing result, a second training error value is determined; Based on the second training error value, it is determined whether the business processing model has converged. If the business processing model has not converged, the business processing model is trained again based on the first image voucher data and the parsing result of the first voucher until the business processing model converges, thus obtaining the trained business processing model.

4. The method according to claim 3, wherein the data volume of the second image voucher data is greater than the data volume of the first image voucher data.

5. The method according to claim 4, wherein the first module comprises a first sub-module for performing business information extraction processing and a second sub-module for performing image tampering detection, the credential parsing result comprises a business information extraction result and an image tampering detection result, and the step of performing credential parsing processing on the third text feature based on the first module to obtain a second credential parsing result, and determining a second training error value based on the first credential parsing result and the second credential parsing result, comprises: Based on the first sub-module of the first module, business information extraction processing is performed on the third text features to obtain the first business information extraction result in the second voucher parsing result; Based on the second sub-module of the first module, image tampering detection processing is performed on the third text features to obtain the first image tampering detection result in the second credential parsing result; Based on the results of the first business information extraction, the first image tampering detection, and the first credential parsing, the second training error value is determined.

6. The method according to claim 5, wherein determining the business processing result corresponding to triggering the execution of the target business for the target user based on the credential parsing result includes: Based on the image tampering detection results in the credential parsing results, the risk detection results for the image credential data are determined; Based on the risk detection results and the business information extraction results from the credential parsing results, the business processing result corresponding to the target business is determined for the target user.

7. The method according to claim 6, wherein determining the risk detection result for the image tampering detection result based on the credential parsing result includes: If the image credential data is determined to be tampered image data based on the image tampering detection result, the tampered area in the image credential data is determined based on the image tampering detection result. Obtain the business information corresponding to the tampered area from the business information extraction results; Based on the tampered area and the corresponding business information, the image credential data is subjected to risk detection processing. The risk detection results for the image credential data are obtained.

8. The method according to claim 7, wherein determining the business processing result corresponding to triggering the execution of the target business for the target user based on the business information extraction result from the risk detection result and the credential parsing result includes: If the risk detection results determine that the image credential data does not pose a risk, the target business is executed based on the business information extraction results to obtain the business processing result corresponding to the execution of the target business for the target user.

9. A data processing apparatus, comprising: The first acquisition module is used to acquire image credential data corresponding to the target user triggering the execution of the target business; The first training module is used to determine a first training error value based on the first text feature and the second text feature, and to train the module constructed by the preset deep learning algorithm to obtain the second module; wherein, the first text feature is obtained by extracting text features from the sub-image voucher data of different regions containing text information in the second image voucher data corresponding to various businesses, and the second text feature is obtained by extracting features from the text description data corresponding to the second image voucher data. The model acquisition module is used to acquire a pre-trained business processing model corresponding to the target business. The business processing model is obtained by training the business processing model constructed by the first module and the pre-trained second module with the first image voucher data corresponding to the target business. The first module is used to perform voucher parsing processing on the output result of the second module. The first processing module is used to determine the sub-image voucher data containing text information in different regions of the image voucher data through the second module of the pre-trained business processing model, and to perform text feature extraction processing on the sub-image voucher data respectively to obtain the text feature information corresponding to the image voucher data. The second processing module is used to perform voucher parsing processing on the text feature information through the first module of the pre-trained business processing model to obtain the voucher parsing result for the image voucher data; The result determination module is used to determine the business processing result corresponding to triggering the execution of the target business for the target user based on the credential parsing result.

10. A data processing apparatus, the data processing apparatus comprising: processor; as well as A memory configured to store computer-executable instructions, which, when executed, cause the processor to: Retrieve the image credentials data corresponding to the target user triggering the execution of the target service; A pre-trained business processing model corresponding to the target business is obtained. This business processing model is obtained by training a business processing model constructed from a first module and a pre-trained second module using first image voucher data corresponding to the target business. The second module is obtained by training a module constructed using a preset deep learning algorithm based on a first training error value determined by first and second text features. The first text feature is obtained by extracting text features from sub-image voucher data containing text information in different regions of the second image voucher data corresponding to various businesses. The second text feature is obtained by extracting features from text description data corresponding to the second image voucher data. The first module is used to perform voucher parsing processing on the output of the second module. The second module of the pre-trained business processing model determines the sub-image voucher data containing text information in different regions of the image voucher data, and performs text feature extraction processing on the sub-image voucher data to obtain the text feature information corresponding to the image voucher data. The first module of the pre-trained business processing model performs credential parsing processing on the text feature information to obtain credential parsing results for the image credential data. Based on the credential parsing result, the business processing result corresponding to the target business triggered for the target user is determined.

Images