Data auditing method, device and equipment, medium and program product

By identifying the business and rules of the data to be reviewed, selecting and calling a special audit tool for review, the problem of low data audit accuracy in the existing technology is solved, and the selection and use of automated tools is realized, and the accuracy of data identification is improved.

CN119090517APending Publication Date: 2024-12-06TENCENT TECHNOLOGY (SHENZHEN) CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411200985.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-28
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

It is difficult for existing technology to improve accuracy in data review, especially in online products and social scenarios. How to effectively review the authenticity and qualifications of data is an urgent problem.

Method used

By obtaining the data to be reviewed, determining its corresponding audit business and rules, selecting a target audit tool collection based on this information, and calling these tools to audit the data, thereby realizing automated tool selection and use and improving the accuracy of data identification.

Benefits of technology

It realizes automated selection and use of audit tools, and understands the uses and methods of the tools according to audit rules, thereby improving the accuracy of data identification. It is suitable for scenarios such as e-commerce audit, online business audit and Internet ecological security testing.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119090517A_ABST
    Figure CN119090517A_ABST
Patent Text Reader

Abstract

The embodiment of the invention provides a data auditing method and device, equipment, a medium and a program product, which are used for improving the accuracy of data auditing. The method can be applied to the fields of computers, artificial intelligence, e-commerce and the like. Comprising the steps of obtaining to-be-audited data; determining an auditing service corresponding to the to-be-audited data; determining a corresponding auditing rule based on the auditing service; a target auditing tool set corresponding to the to-be-audited data is determined based on the auditing rule and an auditing tool list, the auditing tool list is used for indicating a created auditing tool set, and the target auditing tool set is included in the auditing tool list; each auditing tool in the target auditing tool set is used for processing auditing data corresponding to the auditing rule; and calling the target auditing tool set to audit the to-be-audited data to obtain an auditing result of the to-be-audited data.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of computers, and in particular to a data audit method, device, equipment, medium and program product. Background Art

[0002] In recent years, with the rapid development of Internet technology, people's daily lives have increasingly shifted to the Internet, such as online shopping, online social networking, etc. Therefore, how to ensure the authenticity and effectiveness of online products and online social networking is an urgent problem to be solved in the development of Internet technology.

[0003] In the above scenarios, it is necessary to conduct qualification or authenticity review of data providers (such as product suppliers or interactive objects). Therefore, how to improve the accuracy of data review is an urgent problem that needs to be solved. Summary of the invention

[0004] Embodiments of the present application provide a data audit method, apparatus, device, medium, and program product for improving the accuracy of data audit.

[0005] In view of this, on the one hand, the present application provides a data audit method, including: obtaining data to be audited; determining the audit business corresponding to the data to be audited; determining the corresponding audit rules based on the audit business; determining a target audit tool set corresponding to the data to be audited based on the audit rules and an audit tool list, the audit tool list being used to indicate a created audit tool set, the target audit tool set being included in the audit tool list, and each audit tool in the target audit tool set being used to process the audit data corresponding to the audit rules; calling the target audit tool set to audit the data to be audited to obtain an audit result of the data to be audited.

[0006] On the other hand, the present application provides a data audit device, including: an acquisition module, used to acquire data to be audited;

[0007] A processing module is used to determine the audit business corresponding to the data to be audited; determine the corresponding audit rules based on the audit business; determine the target audit tool set corresponding to the data to be audited based on the audit rules and the audit tool list, the audit tool list is used to indicate the created audit tool set, the target audit tool set is included in the audit tool list, and each audit tool in the target audit tool set is used to process the audit data corresponding to the audit rule; call the target audit tool set to audit the data to be audited to obtain the audit result of the data to be audited.

[0008] In one possible design, in another implementation of another aspect of the embodiment of the present application, the processing module is used to perform data preprocessing on the data to be audited based on the audit rules to obtain audit data of the data to be audited, and the audit data serves as input data of the target audit tool set.

[0009] In one possible design, in another implementation of another aspect of the embodiment of the present application, the processing module is used to determine a preprocessing prompt template based on the audit rule, and the preprocessing prompt template includes preprocessing prompt information, the audit rule and the data to be audited;

[0010] The data to be reviewed is filtered based on the preprocessing prompt template to obtain review data of the data to be reviewed.

[0011] In one possible design, in another implementation of another aspect of the embodiment of the present application, the processing module is used to determine the use order and input parameters of each audit tool in the target audit tool set according to the audit rules of the audit business, and the input parameters are used to indicate the input data required by the audit tool;

[0012] The data to be audited is audited according to the usage sequence and the input parameters corresponding to each audit tool to obtain the audit result of the data to be audited.

[0013] In one possible design, in another implementation of another aspect of the embodiment of the present application, the processing module is used to obtain a mapping list corresponding to the audit tool list, the mapping list being used to indicate a correspondence between the audit tools in the audit tool list and the input parameters;

[0014] The input parameters corresponding to each audit tool in the target audit tool set are determined according to the target audit tool set and the mapping list.

[0015] In one possible design, in another implementation of another aspect of the embodiment of the present application, the processing module is used to create a target audit tool and store it in the audit tool list, and the target audit tool includes a tool name, a tool description and a tool execution process, wherein the tool description is used to indicate the purpose of the target audit tool.

[0016] In one possible design, in another implementation of another aspect of the embodiment of the present application, a processing module is used to call a first model to identify the audit rules and the audit tool list to determine a target audit tool set corresponding to the data to be audited, and the first model is obtained by training an initial first model based on a training sample, wherein the training sample includes training instructions, training input data, and training output data, the training instructions are used to indicate the training target of the initial tool recognition model, the training input data includes audit rules and a tool list, and the training output data is an alternative list of audit tools.

[0017] In one possible design, in another implementation of another aspect of the embodiment of the present application, the acquisition module is used to acquire an initial first model and a training sample, wherein the training sample includes a training instruction, training input data, and training output data, the training instruction is used to indicate a training target of the initial tool recognition model, the training input data includes an audit rule and a tool list, and the training output data is an audit tool candidate list;

[0018] The processing module is used to call the initial first model to perform prediction processing on the training sample to obtain a prediction review tool candidate list; determine a first training loss value based on the prediction review tool candidate list and the review tool candidate list; and train the initial first model based on the first training loss value to obtain the first model.

[0019] In one possible design, in another implementation of another aspect of the embodiment of the present application, the processing module is used to determine the first training loss value according to the prediction review tool candidate list and the review tool candidate list according to the gradient descent algorithm and the cross entropy loss function.

[0020] In one possible design, in another implementation of another aspect of the embodiment of the present application, the initial first model is a pre-trained large model.

[0021] In one possible design, in another implementation of another aspect of the embodiment of the present application, when the audit tool list includes a multimodal audit tool, the acquisition module is used to acquire an initial multimodal audit tool and a training sample, the training sample includes an audit rule, input data, and output data, wherein the input data is used to indicate the audit sample, and the output data is the audit result;

[0022] The processing module is used to call the initial multimodal audit tool to perform identification processing on the training sample to obtain a predicted audit result; determine a second training loss value based on the predicted audit result and the audit result; and train the initial multimodal audit tool based on the second training loss value to obtain the multimodal audit tool.

[0023] In one possible design, in another implementation of another aspect of the embodiment of the present application, the processing module is used to determine the second training loss value based on the predicted audit result and the audit result using contrastive learning or a similarity algorithm.

[0024] Another aspect of the present application provides a computer device, including: a memory, a processor, and a bus system;

[0025] Wherein, the memory is used to store programs;

[0026] The processor is used to execute the program in the memory, and the processor is used to execute the above-mentioned methods according to the instructions in the program code;

[0027] The bus system is used to connect the memory and the processor so that the memory and the processor can communicate with each other.

[0028] Another aspect of the present application provides a computer-readable storage medium, in which instructions are stored. When the computer-readable storage medium is run on a computer, the computer is enabled to execute the above-mentioned methods.

[0029] Another aspect of the present application provides a computer program product or a computer program, the computer program product or the computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the methods provided by the above aspects.

[0030] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages: identifying the audit business, audit rules and special audit tools corresponding to the audit rules of the data to be audited, and then using the special audit tools to audit the data to be audited. In this way, the functions of automatically selecting tools and using tools can be realized, and at the same time, the purpose and use of the audit tools can be understood according to the audit rules, thereby improving the accuracy of data recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 A schematic diagram of the architecture of a data audit system in an embodiment of the present application;

[0032] Figure 2 A schematic diagram of a flow chart of a data audit method in an embodiment of the present application;

[0033] Figure 3 This is a schematic diagram of an embodiment of the data audit method in the embodiment of the present application;

[0034] Figure 4 A schematic diagram of a flow chart of a data audit method in an embodiment of the present application;

[0035] Figure 5 A schematic diagram of an embodiment of a data audit device in an embodiment of the present application;

[0036] Figure 6 This is a schematic diagram of an embodiment of a server in an embodiment of the present application;

[0037] Figure 7 This is a schematic diagram of an embodiment of a terminal device in an embodiment of the present application. DETAILED DESCRIPTION

[0038] Embodiments of the present application provide a data audit method, apparatus, device, medium, and program product for improving the accuracy of data audit.

[0039] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein, for example. In addition, the terms "including" and "corresponding to" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0040] In the embodiments of the present application, the term "module" or "unit" refers to a computer program or a part of a computer program with a predetermined function, and works together with other related parts to achieve a predetermined goal, and can be implemented in whole or in part by using software, hardware (such as processing circuits or memories) or a combination thereof. Similarly, a processor (or multiple processors or memories) can be used to implement one or more modules or units. In addition, each module or unit can be part of an overall module or unit that includes the function of the module or unit.

[0041] In recent years, with the rapid development of Internet technology, people's daily lives have increasingly shifted to the Internet, such as online shopping, online social networking, etc. Therefore, how to ensure the authenticity and effectiveness of online goods and online social networking is an urgent problem to be solved in the development of Internet technology. For example, in the field of e-commerce, the quality problems of counterfeit and inferior goods have become increasingly prominent, becoming a major obstacle to the sustainable and healthy development of e-commerce, and also a problem that is of particular concern and urgently needs to be solved in the development of e-commerce. In the process of e-commerce transactions, the qualification review of merchants by e-commerce platforms plays a very important role in improving the quality of platform products. What qualifications should merchants of different categories have to settle in e-commerce platforms? How to review these qualifications of merchants to ensure that merchants settled in e-commerce platforms meet the requirements of relevant laws and regulations, and to ensure the quality of settled merchants from the source, is important for ensuring the quality of goods. Therefore, the settlement of merchants on e-commerce platforms needs to be strictly reviewed in accordance with the review specifications. This is of great significance for e-commerce platform companies to control the source of merchants and ensure the quality of goods. At present, data audit usually adopts the following schemes: Scheme 1: A method based on keyword matching. This method uses a keyword filtering method to identify the text content containing abusive, vulgar and other illegal words in product information, merchant information, etc. as low-quality, and requires rectification or rejection; Scheme 2: A method based on various text similarity calculations to calculate the similarity between product information, merchant information and control items. When the similarity is greater than the threshold, the corresponding text is identified as illegal, and rectification or rejection is required; Scheme 3: A method based on machine learning and deep learning, using large-scale labeled data for statistics and analysis, training machine learning or deep learning models, and identifying illegal merchants; Scheme 4: A method based on a large model, allowing the large model to simulate the human thinking process, and according to the audit rules, directly replace manual work in some simple and easy-to-understand tasks. However, the above methods all have corresponding shortcomings, so there is an urgent need for a solution that can improve the accuracy of data audit.

[0042] In order to solve the above technical problems, the present application provides the following technical solutions: obtaining the data to be audited; determining the audit business corresponding to the data to be audited; determining the audit rules corresponding to the audit business; determining the target audit tool set corresponding to the data to be audited based on the audit rules and the audit tool list, the audit tool list is used to indicate the audit tool set that has been created, the target audit tool set is included in the audit tool list, and each audit tool in the target audit tool set is used to process the audit data corresponding to the audit rule; calling the target audit tool set to audit the data to be audited to obtain the audit result of the data to be audited. In this way, the audit business, audit rules and special audit tools corresponding to the audit rules of the data to be audited are identified, and then the special audit tools are used to audit the data to be audited. In this way, the functions of automatically selecting tools and using tools can be realized, and at the same time, the purpose and use of the audit tools can be understood according to the audit rules, thereby improving the recognition accuracy of data.

[0043] The following describes several exemplary embodiments to illustrate the technical solutions of the embodiments of the present application and the technical effects produced by the technical solutions of the present application. It should be noted that the following embodiments can refer to, draw on or combine with each other, and the same terms, similar features and similar implementation steps in different embodiments will not be described repeatedly.

[0044] The embodiments of the present application can be applied to scenarios such as e-commerce audit, online business audit, online social interaction object information audit, product testing, and Internet ecological security testing. The following is an example of e-commerce audit. An exemplary architecture of this scenario can be found in Figure 1 The architecture includes a terminal 100, a server 200, a database 300 and a network 400. The terminal 100, the server 200 and the database 300 are connected via the network 400. The database 300 is used to store various materials uploaded by users for review (such as invoices, e-commerce store names, e-commerce business licenses, e-commerce account subject identity information, etc.). Figure 1 The number of terminals, servers, and databases in the system shown is for example only. For example, the number of terminals, servers, and databases may be multiple, and the embodiments of the present application do not limit the number of terminals, servers, and databases.

[0045] The terminal 100 communicates with the server 200 through a network, and the database 300 can be integrated on the server 200, or placed on a cloud or other server. In the above data processing process, the terminal 100 and the server 200 can interact with each other. For example, a user uploads corresponding data to be reviewed through the terminal 100, and the terminal 100 sends the data to be reviewed to the server 200.

[0046] The terminal 100 is a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, an intelligent voice interaction device, a smart home appliance, a vehicle-mounted terminal, etc., but is not limited thereto. The server 200 is an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDN), big data and artificial intelligence platforms.

[0047] Database 300, in short, can be regarded as an electronic filing cabinet - a place to store electronic files, where users can add, query, update, delete, etc. the data in the files. The so-called "database" is a collection of data that is stored together in a certain way, can be shared with multiple users, has as little redundancy as possible, and is independent of the application program. Database Management System (DBMS) is a computer software system designed for managing databases, generally with basic functions such as storage, interception, security, and backup. Database management systems can be classified according to the database model it supports, such as relational, Extensible Markup Language (XML); or according to the type of computer supported, such as server clusters, mobile phones; or according to the query language used, such as Structured Query Language (SQL), XQuery; or according to the performance impulse focus, such as maximum scale, maximum operating speed; or other classification methods. Regardless of the classification method used, some DBMS can cross categories, for example, supporting multiple query languages ​​at the same time.

[0048] In an exemplary solution, the technical solution provided by this application can be applied to the e-commerce audit process. The audit of merchants currently mainly includes three subtasks: invoice audit, qualification audit, and naming audit. Based on the above three subtasks, it can cover about 32 audit rules. The following is an explanation of each audit task:

[0049] Invoice review: For e-commerce platforms, invoices are officially recognized transaction vouchers, and each transaction must have a corresponding invoice as proof. This is to ensure the compliance of merchants and prevent illegal transactions. For consumers, invoices are an important basis for consumer rights protection. If there is a problem with the goods or services purchased by consumers, invoices can be used as vouchers for returns, exchanges or repairs. For merchants, accurate and complete invoice information can help merchants to correctly declare taxes and avoid tax risks caused by incorrect declarations. At the same time, invoice information helps merchants to conduct financial accounting, profit and loss analysis, etc., and helps companies make better business decisions. In addition, through strict invoice review, merchants can be prevented from forging, tampering with invoices and other fraudulent behaviors, protecting the rights and interests of e-commerce platforms and consumers.

[0050] Qualification review: Qualification review can ensure the legitimacy of the merchant. By checking its business license, tax registration certificate and other relevant documents, it can verify whether the merchant has legal business qualifications. It can also prevent counterfeit and shoddy goods from entering the market, protect the rights and interests of consumers, and improve consumers' shopping experience and satisfaction. Conducting qualification review on merchants can prevent illegal merchants from conducting illegal transactions on the platform and maintain the image and reputation of the platform. Qualification review can also promote the improvement of standards in the entire e-commerce industry and enhance the competitiveness of the industry. Through qualification review, the credit of merchants can be ensured and a safe trading environment can be provided for consumers. Therefore, qualification review is an important part of the e-commerce platform's review of merchants, and it plays an important role in maintaining the normal operation of the e-commerce platform, protecting consumer rights, and improving industry standards.

[0051] Naming review: Naming review is mainly to ensure that the name of each merchant is unique, avoid confusion and misunderstanding, and protect the interests of consumers. Names are an important part of a brand. Reviewing the names of merchants can protect consumers and other merchants from malicious merchants who impersonate or counterfeit well-known brands. In addition, naming review can avoid the appearance of merchant names containing inappropriate or illegal content, such as defamation, hatred, violence, etc. Through review, it is also ensured that the naming of all merchants complies with the regulations and standards of the platform to ensure a fair competition environment on the platform. For users, a merchant name that complies with regulations, is clear and easy to understand, can help consumers better understand and remember merchants, thereby improving user experience.

[0052] Based on the above description, in this application, the model can be used to identify the audit business of the data to be audited and select the audit tool. Figure 2 The flowchart shown explains the data review of this application:

[0053] like Figure 2As shown, after the user inputs the data to be reviewed, the big model is called to identify and process the data to be reviewed; if the review result can be directly identified by the big model, the review result is directly returned. For example, if an online review requires identity information to be provided, and the data to be reviewed is detected as identity information, the review result can be directly returned, indicating that the identity information has been provided and the review is passed; if the review result cannot be directly identified by the big model (for example, an online review requires reviewing whether the invoice is compliant, that is, further review is required for the invoice in the data to be reviewed, and the result of invoice compliance cannot be directly returned), the review tool set is selected according to the big model; then, the data to be reviewed is extracted according to the input parameters corresponding to the data to be reviewed and the review tool for inputting the review data of the review tool; finally, the review tool is called to review the review data to obtain the review result. It should be understood that when a data to be reviewed has multiple review businesses, different review tools can be called in parallel for review (that is, different review businesses are conducted at the same time) or different review tools can be called in series for review (that is, multiple review businesses are sorted and executed in a certain order), which is not limited here.

[0054] It is understandable that in the specific implementation of this application, related data such as data to be reviewed is involved. When the above embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of relevant data need to comply with relevant laws, regulations and standards of relevant countries and regions.

[0055] Combined with the above introduction, the following is an introduction to the data review method in this application, with the server as the execution subject. Please refer to Figure 3 , an embodiment of the data audit method in the embodiment of the present application includes:

[0056] 301. Get the data to be reviewed.

[0057] In this embodiment, the server receives the data to be reviewed uploaded by the user through the user terminal.

[0058] In an exemplary solution, in an e-commerce audit scenario, the data to be audited may be invoices, business licenses, tax registration certificates, business licenses and other related documents. During the upload process, the data to be audited may be in text, image or video form, which is not limited here.

[0059] In an exemplary solution, in a product review scenario, the data to be reviewed may be product packaging images, product images, product introductions, product advertisements, and other information, which is not specifically limited here.

[0060] 302. Determine the audit business corresponding to the data to be audited.

[0061] In this embodiment, the server may call the fine-tuned large model to identify the data to be reviewed to determine the review business corresponding to the data to be reviewed; or, the server may also determine the review business corresponding to the data to be reviewed through a mapping table.

[0062] In this embodiment, the data to be reviewed may correspond to one or more review services. For example, when the data to be reviewed is a paper invoice or an electronic invoice, the review service may include reviewing the invoice code, reviewing the buyer information and seller information of the invoice, reviewing the verification code, reviewing the invoice number, reviewing the seal, etc.

[0063] It should be understood that the fine-tuned large model can be a model obtained by fine-tuning the pre-trained large model based on the training sample, wherein the training sample can include the data to be reviewed and the sample label, wherein the sample label is used to indicate the review business corresponding to the data to be reviewed. During the training process, the pre-trained large model is called to identify the training sample to obtain a predicted recognition result; then the training loss value is determined based on the predicted recognition result and the sample label; finally, the parameters of the pre-trained large model are optimized based on the training loss value to obtain the fine-tuned large model. Among them, the training loss value can be a cross entropy loss value or other loss value, which is not limited here.

[0064] 303. Determine the corresponding audit rules based on the audit business.

[0065] In this embodiment, after obtaining the audit service of the data to be audited, the server determines the corresponding audit rules according to the audit service. For example, when auditing an invoice, if the audit service is to audit the invoice code, the audit rule may be to audit whether the invoice code is unique; if the audit service is to audit the purchaser information and the seller information of the invoice, the audit rule may be to audit whether the purchaser information and the seller information are normal; if the audit service is to audit the verification code, the audit rule may be to audit whether the verification code is legal; if the audit service is to audit the invoice number, the audit rule may be to audit whether the invoice number is continuous; if the audit service is to audit the seal, the audit rule may be to audit whether the seal is clear.

[0066] 304. Determine the target audit tool set corresponding to the data to be audited based on the audit rule and the audit tool list. The audit tool list is used to indicate the audit tool set that has been created. The target audit tool set is included in the audit tool list. Each audit tool in the target audit tool set is used to process the audit data corresponding to the audit rule.

[0067] In this embodiment, after determining the audit rules of each audit business, the server can call the fine-tuned large model (i.e., the first model in this application) to recall the target audit tool set corresponding to the data to be audited from the audit tool list according to the audit rules. For example, the data to be audited is an invoice, and its corresponding audit business and audit rules can be as follows: when auditing an invoice, if the audit business is to audit the invoice code, the audit rule can be to audit whether the invoice code is unique; if the audit business is to audit the purchaser information and the seller information of the invoice, the audit rule can be to audit whether the purchaser information and the seller information are normal; if the audit business is to audit the verification code, the audit rule can be to audit whether the verification code is legal; if the audit business is to audit the invoice number, the audit rule can be to audit whether the invoice number is continuous; if the audit business is to audit the seal, the audit rule can be to audit whether the seal is clear. Then call the first model to identify the above-mentioned audit rules, and then recall the audit tool set corresponding to each audit business from the audit tool list. For example, to audit the invoice code, the optical character recognition (OCR) tool and the string comparison tool can be obtained through the first model recognition. In another exemplary scheme, in the qualification review stage, for the review rule that "the current name of the account in the user input information and the company name in the business license image need to be consistent", it is necessary to use the OCR tool in the review tool list to extract the text in the business license image. At the same time, in order to facilitate subsequent review, the first model can also structure the text extracted by the OCR tool to obtain the business license information. Finally, use the string comparison tool to extract the "current name of the account" and "company name" from the "user input information" and "business license information" respectively for matching. Therefore, for the review rule that "the current name of the account in the user input information and the company name in the business license image need to be consistent", the first model can output "OCR tool" and "string comparison tool" as the target review tool set.

[0068] It should be understood that the first model can be a model obtained by fine-tuning the pre-trained large model based on the training sample. The training sample can include training instructions, training input data and training output data, the training instructions are used to indicate the training target of the initial tool recognition model, the training input data include audit rules and tool lists, and the training output data is the audit tool candidate list; the initial first model is called to perform prediction processing on the training sample to obtain the predicted audit tool candidate list; the first training loss value is determined according to the predicted audit tool candidate list and the audit tool candidate list; the initial first model is trained based on the first training loss value to obtain the first model. In an exemplary scheme, the data form of the training sample can be {instruction, input, output}, where the instruction part can be fixed as "task description: according to the audit rules provided below and the currently known tool list, determine which tools in the tool list are needed for the audit rule."; the input part is "the following are the audit rules: {audit rules}; the following are the currently available tool list: {tool list}"; the output part is "{tool candidate list}". Then the pre-trained model is trained by fine-tuning the instructions. In this embodiment, the gradient descent algorithm and the cross entropy loss function can be used to optimize the parameters of the pre-trained model. The specific formula is as follows:

[0069]

[0070] Where M is the number of categories, y ij is an indicator function, which is 1 when the true label of the i-th sample is the j-th class, otherwise it is 0. ij is the probability that the model predicts that the i-th sample is of the j-th class. In the task of this invention, M is the number of all words in the vocabulary, y ij It is used to indicate whether the true label of the i-th word in the sentence is the j-th word in the vocabulary. ijIt is the probability that the model predicts that the i-th word in the sentence is the j-th word in the vocabulary. The ultimate optimization goal is to make the value of this loss smaller and smaller. In this embodiment, the initial first model can be a pre-trained large model, and its specific structure can be a generative pre-trained transformer (GPT), a general language model (GLM), Gemini, and Claude. Among them, the Gemini model is famous for its advanced deep learning architecture and wide application fields. It is trained on massive text data, can generate high-quality and coherent text content, and supports multiple languages, and can perform cross-language translation and text generation. In addition, Gemini's multimodal capabilities are also very prominent. It can process multiple data types including text, images, audio and video. The Claude model has been improved in terms of enhanced multilingual capabilities, long text processing capabilities, visual processing capabilities, real-time response capabilities, reduced rejection replies, improved accuracy, structured output, and easier use.

[0071] Optionally, each audit tool in the audit tool list can be created by the user. In this embodiment, when creating the audit tool, a general structure and format can be used to customize and expand the new audit tool. In an exemplary solution, the general structure and format can include three attributes: tool name (name), tool description (description), and tool execution process (run). In an exemplary solution, the code of the tool can be as follows:

[0072]

[0073] Furthermore, in order to ensure that the audit of various data has a high accuracy, a multimodal audit tool can also be created when creating an audit tool. Among them, the multimodal audit tool can increase its different modal functions according to actual applications. And the multimodal tool can be obtained through alignment training. The training method can be as follows: obtain an initial multimodal audit tool and a training sample, the training sample includes audit rules, input data and output data, wherein the input data is used to indicate the audit sample, and the output data is the audit result; call the initial multimodal audit tool to perform recognition processing on the training sample to obtain a predicted audit result; determine a second training loss value based on the predicted audit result and the audit result; train the initial multimodal audit tool based on the second training loss value to obtain the multimodal audit tool. An exemplary scheme is, for example, assuming that the multimodal audit tool needs to audit text (ie, audit rules) and images at the same time, then the multimodal audit tool can provide a text recognition module and an image recognition model. Among them, the text recognition module can be used to understand the audit rules, summarize the required information and make the final judgment; the image recognition module is used to extract image features; then the text features and image features are aligned and mapped to the same feature space. The specific process can be expressed as follows:

[0074] TF x ,IF x =Alignment(TF o ,IF o )

[0075] Among them TF x IF x Represents the aligned text and image features, TF o IF o Represents the original text and image features. The aligned features can be used to identify the image by calculating the similarity. In the alignment training, contrastive learning can be used to train the multimodal tool. In the contrastive learning process, its loss function can be expressed as follows:

[0076]

[0077] The training idea in the contrastive learning process is that if the samples do not belong to the same distribution, the loss function will maximize the distance between the samples, and if the samples belong to the same distribution, the loss function will minimize the distance between them. i ” and “S j " is the corresponding label "y i ” and “y j ”, “θ” is the embedding network, and “∈” is a hyperparameter that defines the lower bound distance between samples of different categories.

[0078] Optionally, when training a multimodal audit tool, a similarity algorithm may also be used for training, and the specific method is not limited here.

[0079] 305. Call the target audit tool set to audit the data to be audited to obtain the audit result of the data to be audited.

[0080] The server determines the order of use and input parameters of each audit tool in the target audit tool set through the audit rules, wherein the input parameters are used to indicate the input data required by the audit tool; then the data to be audited is audited according to the order of use and the input parameters corresponding to each audit tool to obtain the audit result of the data to be audited. The input parameters can be obtained by the server through a mapping list query, wherein the mapping list is used to indicate the correspondence between the audit tool and the input parameters. For example, the OCR tool needs to input the URL of the image, the string comparison tool needs to input two or more strings, the trademark network query tool needs to input the trademark number, and so on.

[0081] When the target audit tool is called for auditing, the data to be audited can also be preprocessed to obtain corresponding audit data. That is, the server can preprocess the data to be audited based on the audit rules to obtain audit data of the data to be audited, and the audit data is used as input data of the target audit tool set.

[0082] In an exemplary solution, the server may determine a preprocessing prompt template based on the audit rule, the preprocessing prompt template including preprocessing prompt information, the audit rule and the data to be audited; and filter the data to be audited based on the preprocessing prompt template to obtain audit data of the data to be audited. The preprocessing prompt template may be designed as follows:

[0083] “Task description: Based on the audit rules and user input information provided below, determine which user input fields the audit rules need to use and filter out irrelevant information. Audit rules: {}; User input: {}.

[0084] After obtaining the audit results, the server can also feed back the audit results to the terminal device and prompt the user with specific modification directions according to the audit results. For example, if the invoice does not comply with the rules, the user can be prompted that the invoice is not compliant and asks the user to provide a compliant invoice.

[0085] Below Figure 4 The merchant data review process shown below explains the data review in this application:

[0086] The merchant inputs information, wherein the merchant information may include qualification information and naming information; then the server (i.e. Figure 4 The agent in the example inputs the merchant information and tool information (it should be understood that the tool information is used to indicate the tool information of the audit tool that has been created in the tool pool at the current moment) into a trained big model (the big model can be used to identify the audit business, audit rules, and corresponding audit tools and input parameters corresponding to the audit tools corresponding to the merchant information); then the big model outputs the target tool information and the input parameters corresponding to the target tool; the server calls the target tool based on the target tool information and the input parameters corresponding to the target tool to audit the merchant information to obtain the audit result; the server feeds back the audit result to the merchant.

[0087] The data audit device in this application is described in detail below. Figure 5 , Figure 5 This is a schematic diagram of an embodiment of a data audit device in an embodiment of the present application. The data audit device 20 includes:

[0088] The acquisition module 201 is used to acquire the data to be reviewed;

[0089] Processing module 202 is used to determine the audit business corresponding to the data to be audited; determine the corresponding audit rules based on the audit business; determine the target audit tool set corresponding to the data to be audited based on the audit rules and the audit tool list, the audit tool list is used to indicate the created audit tool set, the target audit tool set is included in the audit tool list, and each audit tool in the target audit tool set is used to process the audit data corresponding to the audit rule; call the target audit tool set to audit the data to be audited to obtain the audit result of the data to be audited.

[0090] In the embodiment of the present application, a data audit device is provided. The above device is used to identify the audit business, audit rules and special audit tools corresponding to the audit rules of the data to be audited, and then the special audit tools are used to audit the data to be audited. In this way, the functions of automatically selecting tools and using tools can be realized, and the purpose and use of the audit tools can be understood according to the audit rules, thereby improving the recognition accuracy of data.

[0091] Optionally, in the above Figure 5 On the basis of the corresponding embodiment, in another embodiment of the data audit device 20 provided in the embodiment of the present application,

[0092] The processing module 202 is used to perform data preprocessing on the data to be reviewed based on the review rule to obtain review data of the data to be reviewed, and the review data is used as input data of the target review tool set.

[0093] In the embodiment of the present application, a data audit device is provided. With the above device, after the audit business and audit rules are determined, the audit data can be pre-processed according to the audit rules, which can reduce redundant data and thus shorten the model processing time, while also improving the accuracy of data audit.

[0094] Optionally, in the above Figure 5 On the basis of the corresponding embodiment, in another embodiment of the data audit device 20 provided in the embodiment of the present application, the processing module 202 is used to determine a preprocessing prompt template based on the audit rule, and the preprocessing prompt template includes preprocessing prompt information, the audit rule and the data to be audited;

[0095] The data to be reviewed is filtered based on the preprocessing prompt template to obtain review data of the data to be reviewed.

[0096] In an embodiment of the present application, a data audit device is provided. By using the above device and using a prompt template to pre-process the data to be audited, the accuracy of the data pre-processing can be effectively increased.

[0097] Optionally, in the above Figure 5 On the basis of the corresponding embodiment, in another embodiment of the data audit device 20 provided in the embodiment of the present application,

[0098] The processing module 202 is used to determine the use order and input parameters of each audit tool in the target audit tool set according to the audit rules of the audit business, and the input parameters are used to indicate the input data required by the audit tool;

[0099] The data to be audited is audited according to the usage sequence and the input parameters corresponding to each audit tool to obtain the audit result of the data to be audited.

[0100] In an embodiment of the present application, a data audit device is provided. By using the above device, the order of use of each audit tool is determined according to the audit order provided in the audit rule, so that the audit order in the data audit can be correct, thereby ensuring the accuracy of the data audit.

[0101] Optionally, in the above Figure 5 On the basis of the corresponding embodiment, in another embodiment of the data audit device 20 provided in the embodiment of the present application,

[0102] The processing module 202 is used to obtain a mapping list corresponding to the audit tool list, where the mapping list is used to indicate the corresponding relationship between the audit tools in the audit tool list and the input parameters;

[0103] The input parameters corresponding to each audit tool in the target audit tool set are determined according to the target audit tool set and the mapping list.

[0104] In an embodiment of the present application, a data audit device is provided. With the above device, after selecting the audit tool, the corresponding audit parameters are determined for each audit tool according to the different usage modes of the audit tool, and provided to the model, so that the audit data can be obtained more effectively, thereby reducing the model processing time and increasing the accuracy of data audit.

[0105] Optionally, in the above Figure 5 On the basis of the corresponding embodiment, in another embodiment of the data audit device 20 provided in the embodiment of the present application, the processing module 202 is used to create a target audit tool and store it in the audit tool list, and the target audit tool includes a tool name, a tool description and a tool execution process, wherein the tool description is used to indicate the purpose of the target audit tool.

[0106] In an embodiment of the present application, a data audit device is provided. The device is used to implement customized extension of audit tools using a common structure and format, thereby achieving scalability and generalization of data audit.

[0107] Optionally, in the above Figure 5 On the basis of the corresponding embodiment, in another embodiment of the data audit device 20 provided in the embodiment of the present application, the processing module 202 is used to call the first model to identify the audit rules and the audit tool list to determine the target audit tool set corresponding to the data to be audited, and the first model is obtained by training an initial first model based on a training sample, wherein the training sample includes training instructions, training input data and training output data, the training instructions are used to indicate the training target of the initial tool recognition model, the training input data includes audit rules and a tool list, and the training output data is an alternative list of audit tools.

[0108] In an embodiment of the present application, a data audit device is provided. The above device is used to identify the audit business, audit rules and special audit tools corresponding to the audit rules of the data to be audited, and then the special audit tools are used to audit the data to be audited. In this way, by using the tool to audit the data to be audited, the model can understand the purpose and use of the tool, and the model can automatically select and use the tool according to the actual audit business, so that the recognition effect can be more accurate.

[0109] Optionally, in the above Figure 5On the basis of the corresponding embodiment, in another embodiment of the data audit device 20 provided in the embodiment of the present application, the acquisition module 201 is used to acquire the initial first model and training samples, wherein the training samples include training instructions, training input data and training output data, the training instructions are used to indicate the training target of the initial tool recognition model, the training input data includes audit rules and a tool list, and the training output data is an alternative list of audit tools;

[0110] The processing module 202 is used to call the initial first model to perform prediction processing on the training sample to obtain a prediction review tool candidate list; determine a first training loss value based on the prediction review tool candidate list and the review tool candidate list; and train the initial first model based on the first training loss value to obtain the first model.

[0111] In the embodiment of the present application, a data audit device is provided. The above device is used to construct training data using audit rules, tool lists and candidate lists, and the training data is used to train the model, so that the model can understand the purpose and use of the tool, and the model can automatically select and use the tool according to the actual audit business, so that the recognition effect can be more accurate.

[0112] Optionally, in the above Figure 5 On the basis of the corresponding embodiment, in another embodiment of the data audit device 20 provided in the embodiment of the present application,

[0113] The processing module 202 is used to determine the first training loss value according to the prediction audit tool candidate list and the audit tool candidate list according to the gradient descent algorithm and the cross entropy loss function.

[0114] In an embodiment of the present application, a data audit device is provided. The above device is used to calculate the training loss value using a gradient descent algorithm and a cross entropy loss function to improve the feasibility of the solution.

[0115] Optionally, in the above Figure 5 On the basis of the corresponding embodiment, in another embodiment of the data audit device 20 provided in the embodiment of the present application,

[0116] The initial first model is a pre-trained large model.

[0117] In the embodiment of the present application, a data audit device is provided. By using the above device, a pre-trained large model is used as the initial model of the first model, which can effectively reduce the time of model training, thereby increasing the feasibility of the solution.

[0118] Optionally, in the above Figure 5On the basis of the corresponding embodiment, in another embodiment of the data audit device 20 provided in the embodiment of the present application,

[0119] When the audit tool list includes a multimodal audit tool, the acquisition module 201 is used to acquire an initial multimodal audit tool and a training sample, the training sample includes an audit rule, input data and output data, wherein the input data is used to indicate the audit sample, and the output data is the audit result;

[0120] The processing module 202 is used to call the initial multimodal audit tool to perform identification processing on the training sample to obtain a predicted audit result; determine a second training loss value based on the predicted audit result and the audit result; and train the initial multimodal audit tool based on the second training loss value to obtain the multimodal audit tool.

[0121] In an embodiment of the present application, a data audit device is provided. By using the above device, for a multi-modal audit service, a corresponding multi-modal audit tool is trained, so that the accuracy of data audit can be increased.

[0122] Optionally, in the above Figure 5 On the basis of the corresponding embodiment, in another embodiment of the data audit device 20 provided in the embodiment of the present application,

[0123] The processing module 202 is used to determine the second training loss value by contrastive learning according to the predicted audit result and the audit result.

[0124] In the embodiment of the present application, a data audit device is provided. By using the above device, a comparative learning scheme is provided to determine the training loss value, thereby increasing the feasibility of the scheme.

[0125] The data audit device provided by this application can be used for the server, please refer to Figure 6 , Figure 6: is a schematic diagram of a server structure provided in an embodiment of the present application. The server 300 may have relatively large differences due to different configurations or performances, and may include one or more central processing units (CPU) 322 (for example, one or more processors) and memory 332, and one or more storage media 330 (for example, one or more mass storage devices) storing application programs 342 or data 344. Among them, the memory 332 and the storage medium 330 can be short-term storage or permanent storage. The program stored in the storage medium 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the server. Furthermore, the central processing unit 322 can be configured to communicate with the storage medium 330 to execute a series of instruction operations in the storage medium 330 on the server 300.

[0126] The server 300 may also include one or more power supplies 326, one or more wired or wireless network interfaces 350, one or more input and output interfaces 358, and / or one or more operating systems 341, such as Windows Server 2000. TM , Mac OS X TM , Unix TM ,Linux TM , FreeBSD TM etc.

[0127] The steps performed by the server in the above embodiment can be based on the Figure 6 The server structure shown.

[0128] The data audit device provided in this application can be used for terminal equipment, please refer to Figure 7 For the sake of convenience, only the parts related to the embodiments of the present application are shown. For specific technical details not disclosed, please refer to the method part of the embodiments of the present application. In the embodiments of the present application, the terminal device is a smart phone as an example for explanation:

[0129] Figure 7 The block diagram shows a partial structure of a smart phone related to the terminal device provided in the embodiment of the present application. Figure 7 The smart phone includes: a radio frequency (RF) circuit 410, a memory 420, an input unit 430, a display unit 440, a sensor 450, an audio circuit 460, a wireless fidelity (WiFi) module 470, a processor 480, and a power supply 490. Those skilled in the art will appreciate that Figure 7The structure of the smartphone shown in the figure does not constitute a limitation of the smartphone, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0130] Combine the following Figure 7 A detailed introduction to the various components of a smartphone:

[0131] The RF circuit 410 can be used for receiving and sending signals during the process of sending and receiving information or making calls. In particular, after receiving the downlink information of the base station, it is sent to the processor 480 for processing; in addition, the designed uplink data is sent to the base station. Generally, the RF circuit 410 includes but is not limited to an antenna, at least one amplifier, a transceiver, a coupler, a low noise amplifier (LNA), a duplexer, etc. In addition, the RF circuit 410 can also communicate with the network and other devices through wireless communication. The above-mentioned wireless communication can use any communication standard or protocol, including but not limited to the global system of mobile communication (GSM), general packet radio service (GPRS), code division multiple access (CDMA), wideband code division multiple access (WCDMA), long term evolution (LTE), email, short messaging service (SMS), etc.

[0132] The memory 420 can be used to store software programs and modules. The processor 480 executes various functional applications and data processing of the smartphone by running the software programs and modules stored in the memory 420. The memory 420 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area may store data created according to the use of the smartphone (such as audio data, a phone book, etc.), etc. In addition, the memory 420 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0133] The input unit 430 can be used to receive input digital or character information, and to generate key signal input related to the user settings and function control of the smart phone. Specifically, the input unit 430 may include a touch panel 431 and other input devices 432. The touch panel 431, also known as a touch screen, can collect the user's touch operation on or near it (such as the user's operation on the touch panel 431 or near the touch panel 431 using any suitable object or accessory such as a finger, stylus, etc.), and drive the corresponding connection device according to a pre-set program. Optionally, the touch panel 431 may include two parts: a touch detection device and a touch controller. Among them, the touch detection device detects the user's touch orientation, detects the signal brought by the touch operation, and transmits the signal to the touch controller; the touch controller receives the touch information from the touch detection device, converts it into contact coordinates, and then sends it to the processor 480, and can receive and execute commands sent by the processor 480. In addition, the touch panel 431 can be implemented in various types such as resistive, capacitive, infrared, and surface acoustic waves. In addition to the touch panel 431, the input unit 430 may also include other input devices 432. Specifically, other input devices 432 may include but are not limited to one or more of a physical keyboard, function keys (such as volume control keys, switch keys, etc.), a trackball, a mouse, a joystick, etc.

[0134] The display unit 440 can be used to display information input by the user or information provided to the user and various menus of the smartphone. The display unit 440 may include a display panel 441. Optionally, the display panel 441 may be configured in the form of a liquid crystal display (LCD), an organic light-emitting diode (OLED), or the like. Further, the touch panel 431 may cover the display panel 441. When the touch panel 431 detects a touch operation on or near it, it is transmitted to the processor 480 to determine the type of touch event. Subsequently, the processor 480 provides a corresponding visual output on the display panel 441 according to the type of touch event. Although in Figure 7 In the embodiment, the touch panel 431 and the display panel 441 are used as two independent components to implement the input and output functions of the smartphone, but in some embodiments, the touch panel 431 and the display panel 441 can be integrated to implement the input and output functions of the smartphone.

[0135] The smartphone may also include at least one sensor 450, such as a light sensor, a motion sensor, and other sensors. Specifically, the light sensor may include an ambient light sensor and a proximity sensor, wherein the ambient light sensor may adjust the brightness of the display panel 441 according to the brightness of the ambient light, and the proximity sensor may turn off the display panel 441 and / or the backlight when the smartphone is moved to the ear. As a type of motion sensor, the accelerometer sensor can detect the magnitude of acceleration in all directions (generally three axes), and can detect the magnitude and direction of gravity when stationary. It can be used for applications that identify the posture of the smartphone (such as horizontal and vertical screen switching, related games, magnetometer posture calibration), vibration recognition related functions (such as pedometer, tapping), etc.; as for other sensors that can be configured in smartphones, such as gyroscopes, barometers, hygrometers, thermometers, infrared sensors, etc., they will not be repeated here.

[0136] The audio circuit 460, the speaker 461, and the microphone 462 can provide an audio interface between the user and the smartphone. The audio circuit 460 can transmit the received audio data to the speaker 461 after converting the received audio data into an electrical signal, which is converted into a sound signal for output; on the other hand, the microphone 462 converts the collected sound signal into an electrical signal, which is received by the audio circuit 460 and converted into audio data, and then the audio data is output to the processor 480 for processing, and then sent to another smartphone through the RF circuit 410, or the audio data is output to the memory 420 for further processing.

[0137] WiFi is a short-range wireless transmission technology. Smartphones can help users send and receive emails, browse web pages, and access streaming media through WiFi module 470, which provides users with wireless broadband Internet access. Figure 7 A WiFi module 470 is shown, but it is understandable that it is not an essential component of the smartphone and can be omitted as needed without changing the essence of the invention.

[0138] Processor 480 is the control center of the smartphone, which uses various interfaces and lines to connect various parts of the entire smartphone, and executes various functions of the smartphone and processes data by running or executing software programs and / or modules stored in memory 420, and calling data stored in memory 420, thereby monitoring the smartphone as a whole. Optionally, processor 480 may include one or more processing units; optionally, processor 480 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into processor 480.

[0139] The smart phone also includes a power source 490 (such as a battery) for supplying power to various components. Optionally, the power source can be logically connected to the processor 480 through a power management system, thereby managing functions such as charging, discharging, and power consumption management through the power management system.

[0140] Although not shown, the smartphone may also include a camera, a Bluetooth module, etc., which will not be described in detail here.

[0141] The steps performed by the terminal device in the above embodiment can be based on the Figure 7 The terminal device structure shown.

[0142] A computer-readable storage medium is also provided in an embodiment of the present application. The computer-readable storage medium stores a computer program, which, when executed on a computer, enables the computer to execute the methods described in the aforementioned embodiments.

[0143] The present application also provides a computer program product including a program, which, when executed on a computer, enables the computer to execute the methods described in the aforementioned embodiments.

[0144] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0145] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0146] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0147] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0148] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0149] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A data audit method, characterized in that: include: Get the data to be reviewed; Determine the audit business corresponding to the data to be audited and determine the audit rules corresponding to the audit business based on the audit business; Determine a target audit tool set corresponding to the data to be audited based on the audit rules and the audit tool list, wherein the audit tool list is used to indicate a created audit tool set, the target audit tool set is included in the audit tool list, and each audit tool in the target audit tool set is used to process the audit data corresponding to the audit rules; The target audit tool set is called to audit the data to be audited to obtain the audit result of the data to be audited.

2. The method according to claim 1, characterized in that Determining the target audit tool set corresponding to the data to be audited based on the audit rule and the audit tool list includes: The first model is called to identify the audit rules and the audit tool list to determine the target audit tool set corresponding to the data to be audited, wherein the first model is obtained by training an initial first model based on a training sample, wherein the training sample includes training instructions, training input data and training output data, the training instructions are used to indicate the training target of the initial tool recognition model, the training input data includes audit rules and a tool list, and the training output data is an alternative list of audit tools.

3. The method according to claim 2, characterized in that The method further comprises: Acquire an initial first model and a training sample, wherein the training sample includes a training instruction, training input data, and training output data, the training instruction is used to indicate the training target of the initial tool recognition model, the training input data includes an audit rule and a tool list, and the training output data is an audit tool candidate list; Calling the initial first model to perform prediction processing on the training sample to obtain a candidate list of prediction review tools; Determine a first training loss value according to the prediction audit tool candidate list and the audit tool candidate list; The initial first model is trained based on the first training loss value to obtain the first model.

4. The method according to claim 3, characterized in that Determining the first training loss value according to the prediction audit tool candidate list and the audit tool candidate list includes: The first training loss value is determined according to the prediction audit tool candidate list and the audit tool candidate list using a gradient descent algorithm and a cross entropy loss function.

5. The method according to any one of claims 1 to 4, characterized in that The calling of the target audit tool set to audit the data to be audited to obtain the audit result of the data to be audited includes: Determining the order of use and input parameters of each audit tool in the target audit tool set according to the audit rules of the audit business, wherein the input parameters are used to indicate the input data required by the audit tool; The data to be audited is audited according to the usage sequence and the input parameters corresponding to each audit tool to obtain the audit result of the data to be audited.

6. The method according to claim 5, characterized in that The step of determining the input parameters of each audit tool in the target audit tool set according to the audit rules of the audit business includes: Obtaining a mapping list corresponding to the audit tool list, wherein the mapping list is used to indicate a corresponding relationship between the audit tools in the audit tool list and input parameters; The input parameters corresponding to each audit tool in the target audit tool set are determined according to the target audit tool set and the mapping list.

7. The method according to any one of claims 1 to 4 and 6, characterized in that: The method further comprises: A target audit tool is created and stored in the audit tool list, wherein the target audit tool includes a tool name, a tool description, and a tool execution process, wherein the tool description is used to indicate the purpose of the target audit tool.

8. The method according to any one of claims 1 to 4 and 6, characterized in that: After determining the audit rules corresponding to the audit business, before calling the target audit tool set to audit the data to be audited to obtain the audit results of the data to be audited, the method further includes: The data to be audited is preprocessed based on the audit rules to obtain audit data of the data to be audited, and the audit data serves as input data of the target audit tool set.

9. The method according to claim 8, characterized in that The preprocessing of the data to be audited based on the audit rule to obtain audit data of the data to be audited includes: Determine a preprocessing prompt template based on the audit rule, wherein the preprocessing prompt template includes preprocessing prompt information, the audit rule and the data to be audited; The data to be reviewed is filtered based on the preprocessing prompt template to obtain review data of the data to be reviewed.

10. The method according to any one of claims 1 to 4 and 6, characterized in that: When the list of audit tools includes a multimodal audit tool, the method further includes: Acquire an initial multimodal audit tool and a training sample, wherein the training sample includes an audit rule, input data, and output data, wherein the input data is used to indicate the audit sample, and the output data is the audit result; Calling the initial multimodal audit tool to perform recognition processing on the training samples to obtain a predicted audit result; Determine a second training loss value according to the predicted audit result and the audit result; The initial multimodal audit tool is trained according to the second training loss value to obtain the multimodal audit tool.

11. The method according to claim 10, characterized in that Determining the second training loss value according to the predicted audit result and the audit result includes: The second training loss value is determined according to the predicted audit result and the audit result using contrastive learning or a similarity algorithm.

12. A data audit device, characterized in that: include: The acquisition module is used to obtain the data to be reviewed; A processing module, used to determine the audit business corresponding to the data to be audited; Determine the corresponding audit rules based on the audit business; determine the target audit tool set corresponding to the data to be audited based on the audit rules and the audit tool list, the audit tool list is used to indicate the created audit tool set, the target audit tool set is included in the audit tool list, and each audit tool in the target audit tool set is used to process the audit data corresponding to the audit rules; call the target audit tool set to audit the data to be audited to obtain the audit result of the data to be audited.

13. A computer device, characterized in that: include: Memory, processor, and bus system; Wherein, the memory is used to store programs; The processor is used to execute the program in the memory, and the processor is used to execute the method according to any one of claims 1 to 11 according to the instructions in the program code; The bus system is used to connect the memory and the processor so that the memory and the processor can communicate with each other.

14. A computer-readable storage medium comprising instructions, which, when executed on a computer, causes the computer to perform the method according to any one of claims 1 to 11.

15. A computer program product comprising a computer program, characterized in that The computer program is used by a processor to execute the method according to any one of claims 1 to 11.

Citation Information

Cited By

  • E-commerce product name review method and apparatus, and electronic device and storage medium

    WO2026129863A1