Data auditing method and device and electronic equipment

Abstract

The invention provides a data auditing method and device and electronic equipment. The data auditing method comprises the steps of obtaining to-be-audited data and a first auditing demand; determining a second auditing demand according to the to-be-audited data and the first auditing demand, wherein the second auditing demand is used for representing an auditing demand associated with the to-be-audited data; determining a corresponding auditing instruction according to the second auditing demand; and processing the to-be-audited data based on the audit instruction to obtain a corresponding audit result. The invention relates to the technical field of data analysis, and can improve the accuracy of data auditing.

Description

Technical Field

[0001] The present application relates to the technical field of data analysis, and in particular to a data review method, device and electronic equipment. Background Art

[0002] At present, with the popularity of social media, short video platforms and instant messaging tools, the amount and types of content generated by users on Internet platforms are growing rapidly. These contents are in various forms, involving multiple modes such as text, pictures, audio and video. However, these massive contents are also mixed with a large amount of illegal and bad information, which brings huge challenges to the compliance operation of the platform. Manual review has always been the main means of data review, but it has problems of low efficiency and low accuracy. When using machine learning and deep learning models to review data, they lack multimodal fusion capabilities and can usually only process single-modal data. They cannot comprehensively analyze the correlation information between multimodal data, resulting in low accuracy of the review results. Summary of the invention

[0003] In view of the above, it is necessary to propose a data audit method, device and electronic device to solve the technical problem of low accuracy of data audit.

[0004] The present application provides a data audit method, which includes: obtaining data to be audited and a first audit requirement; determining a second audit requirement based on the data to be audited and the first audit requirement, wherein the second audit requirement is used to characterize the audit requirement associated with the data to be audited; determining a corresponding audit instruction based on the second audit requirement; and processing the data to be audited based on the audit instruction to obtain a corresponding audit result.

[0005] In some embodiments, determining the second audit requirement based on the data to be audited and the first audit requirement includes: determining corresponding modal information based on the data to be audited; determining corresponding audit standards based on the first audit requirement; and determining the second audit requirement based on the modal information and the audit standards.

[0006] In some embodiments, the method also includes: after obtaining the second audit requirement, if it is determined that the second audit requirement does not meet a preset first condition, receiving defect information of the second audit requirement; wherein the first condition is used to characterize the semantics of the first audit requirement; updating the second audit requirement according to the defect information, the modal information, and the audit standard to obtain an updated second audit requirement.

[0007] In some embodiments, determining the corresponding audit instruction based on the second audit requirement includes: determining the first audit task in the second audit requirement based on the modal information; determining the second audit task in the second audit requirement based on the audit standard and the first audit task; and determining the audit instruction based on the first audit task and the second audit task.

[0008] In some embodiments, the method further includes: sorting the first audit task and the second audit task according to the second audit requirement to obtain an execution order of the first audit task and the second audit task; and running the audit instructions corresponding to the first audit task and the second audit task according to the execution order.

[0009] In some embodiments, processing the data to be audited based on the audit instruction to obtain the corresponding audit result includes: when there are multiple audit instructions, obtaining candidate audit results output by multiple audit instructions; combining the candidate audit results to obtain the audit result corresponding to the data to be audited.

[0010] In some embodiments, the method further includes: outputting a corresponding audit result when the audit result satisfies a preset second condition; wherein the second condition is used to characterize the degree of matching between the audit result and the data to be audited.

[0011] In some embodiments, the method further includes: if the audit result does not satisfy the second condition, receiving an evaluation result corresponding to the audit result; and generating a modification strategy for the audit instruction based on the audit result and the corresponding evaluation result.

[0012] An embodiment of the present application also provides a data audit device, which includes: an acquisition module, used to acquire data to be audited and a first audit requirement; a parsing module, used to determine a second audit requirement based on the data to be audited and the first audit requirement, the second audit requirement being used to characterize the audit requirement associated with the data to be audited; a planning module, used to determine a corresponding audit instruction based on the second audit requirement; and an audit module, used to process the data to be audited based on the audit instruction to obtain a corresponding audit result.

[0013] An embodiment of the present application also provides an electronic device, which includes: a memory storing at least one instruction; and a processor executing the instruction stored in the memory to implement the data audit method.

[0014] An embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the data audit method is implemented.

[0015] It can be seen from the above technical solutions that the embodiment of the present application obtains a second audit requirement that is highly correlated with the first audit requirement and has more content by analyzing the data to be audited and the first audit requirement. Then, the corresponding audit instruction is determined according to the second audit requirement, and the granularity of data audit is improved through different audit instructions. Finally, the outputs of different audit instructions are combined to obtain the audit results corresponding to the data to be audited, thereby improving the accuracy of data audit. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is an application scenario diagram of a data audit method provided in one embodiment of the present application.

[0017] Figure 2 It is a flow chart of a data audit method provided in one embodiment of the present application.

[0018] Figure 3 This is a flowchart of a method for determining a second audit requirement provided by an embodiment of the present application.

[0019] Figure 4 This is a flowchart of a method for updating the second review requirement provided by an embodiment of the present application.

[0020] Figure 5 This is a flowchart of a method for determining an audit instruction provided by an embodiment of the present application.

[0021] Figure 6 This is a flowchart of a method for executing an audit instruction provided in one embodiment of the present application.

[0022] Figure 7 This is a flowchart of a method for determining audit results provided by an embodiment of the present application.

[0023] Figure 8 It is a flowchart of a method for generating a modification strategy provided by an embodiment of the present application.

[0024] Fig. 9 It is a functional module diagram of a data audit device provided in one embodiment of the present application.

[0025] Fig.10 It is a structural schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0026] In order to more clearly understand the purpose, features and advantages of the present application, the present application is described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that, in the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other. In the following description, many specific details are set forth to facilitate a full understanding of the present application, and the embodiments described are only a part of the embodiments of the present application, rather than all of the embodiments.

[0027] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0028] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by those skilled in the art to which the present application belongs. The terms used herein in the specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0029] The embodiment of the present application provides a data audit method, which can be applied to one or more electronic devices. The electronic device is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to a microprocessor, an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a programmable gate array (Field-Programmable Gate Array, FPGA), a digital processor (Digital Signal Processor, DSP), an embedded device, etc.

[0030] An electronic device can be any electronic product that can interact with customers, such as personal computers, tablet computers, smart phones, personal digital assistants (PDAs), game consoles, Internet Protocol Television (IPTV), smart wearable devices, etc.

[0031] The electronic device may also include a network device and / or a client device, wherein the network device includes, but is not limited to, a single network server, a server group consisting of multiple network servers, or a cloud consisting of a large number of hosts or network servers based on cloud computing.

[0032] The networks where electronic devices are located include but are not limited to the Internet, wide area network, metropolitan area network, local area network, virtual private network (VPN), etc.

[0033] like Figure 1 The figure shows an application scenario diagram of a data audit method provided by an embodiment of the present application. The data audit method provided by the present application can be applied to an electronic device 100. The electronic device 100 is used to receive the data to be audited and the first audit requirement input by the user. The electronic device 100 is used to parse the data to be audited and the first audit requirement, and determine the second audit requirement according to the data to be audited and the first audit requirement, wherein the second audit requirement is used to characterize the audit requirement associated with the data to be audited. The electronic device 100 is also used to determine a corresponding audit instruction according to the second audit requirement, process the data to be audited based on the audit instruction, and obtain a corresponding audit result. like Figure 2 , is a flow chart of a data audit method provided by an embodiment of the present application. According to different requirements, the order of the steps in the flow chart can be changed, and some steps can be omitted. A data audit method provided by an embodiment of the present application includes the following steps.

[0034] S20, obtaining the data to be reviewed and the first review requirement.

[0035] In one embodiment of the present application, in order to ensure the security of the disseminated data, the data to be reviewed and the first review requirement input by the user into the electronic device can be obtained in advance, and the data to be reviewed can be analyzed at multiple levels in combination with a pre-built database and review instructions to obtain the review results of the data to be reviewed.

[0036] Specifically, the user inputs the data to be reviewed through an interactive program running in the electronic device. The interactive program is used to provide the user with a multi-modal input interface for data interaction with the electronic device, which can receive the data to be reviewed input by the user (for example, text, pictures, videos, audio, etc.) and input the first review requirement. Through natural language processing and form options, the user can clearly express the review requirements such as inappropriate speech, violent images, sensitive information identification, etc. The system confirms the user's needs through an interactive interface or multiple rounds of dialogue.

[0037] In one embodiment of the present application, the data to be reviewed input by the user includes text data, image data, video data, and audio data. Exemplarily, the text data can be an article, a comment, or social media content, etc.; the image data can be a static picture, a screenshot, or a dynamic image in GIF format, etc.; the video data can be a video clip, or a live streaming video, etc.; the audio data can be an audio file in any format. The first review requirement input by the user is used to characterize his or her need to review the data to be reviewed. Exemplarily, the content of the first review requirement can be: whether the review content complies with policies and regulations (for example, inappropriate remarks). The content of the first review requirement can also be: identifying specific sensitive information (for example, sensitive information specified by the Internet platform, etc.).

[0038] S21, determining a second audit requirement according to the data to be audited and the first audit requirement, where the second audit requirement is used to represent an audit requirement associated with the data to be audited.

[0039] In one embodiment of the present application, in order to improve the accuracy of data audit, the electronic device retrieves the associated data related to the first audit requirement from a preset database according to the first audit requirement input by the user. And analyzes the first audit requirement and the data to be audited based on the parsing instruction with the intent parsing function to obtain the second audit requirement. Among them, the database is used to store multiple audit standards in the form of a knowledge graph, so that the audit standards and the relevant information of sensitive content can be structured, so as to regularly update the content of the audit standards to maintain its accuracy.

[0040] The second audit requirement is used to characterize the audit requirement associated with the data to be audited. The parsing instruction is used to perform natural language analysis on the first audit requirement, extract key information from the first audit requirement, and determine the audit requirement associated with the data to be audited. Specifically, the parsing instruction can be used to implement a pre-trained generative model (for example, a GPT model), use the pre-trained generative model to perform semantic analysis on the modal information of the data to be audited and the associated data of the first audit requirement, and generate the corresponding second audit requirement. Specifically, for a method of determining the second audit requirement based on the data to be audited and the first audit requirement, please refer to Figure 3 The corresponding detailed description.

[0041] In one embodiment of the present application, in order to improve the accuracy of the second audit requirement, the second audit requirement can be output to facilitate the user to evaluate the second audit requirement, and the evaluation result of the second audit requirement input by the user is received, and the second audit requirement is updated according to the evaluation result to obtain the updated second audit requirement. For specific methods of updating the second audit requirement, please refer to Figure 4 The corresponding detailed description.

[0042] S22: Determine a corresponding audit instruction according to the second audit requirement.

[0043] In one embodiment of the present application, in order to improve the accuracy of data audit, the corresponding audit instruction is also determined according to the second audit requirement. Specifically, the second audit requirement can be decomposed according to the semantics of the second audit requirement, at least one audit instruction corresponding to the second audit requirement can be determined, and the running order of the audit instruction can be determined to ensure that the running process of the audit instruction conforms to the logic of the second audit requirement. Specifically, the electronic device decomposes the more complex second audit requirement into multiple audit tasks through instructions with data disassembly function, thereby reducing the complexity of the task and determining the audit instructions that meet the second audit requirement to execute each audit task. And the electronic device module determines the execution order of the audit instruction according to the complexity of the audit task, which can optimize the allocation of computing resources, thereby improving the efficiency of data audit. Exemplarily, the electronic device runs a generative pre-training model to decompose the second audit requirement into multiple levels of audit tasks, and then plans to call different audit instructions according to the second audit requirement, and gradually and complexly calls the corresponding audit instructions to process the data to be audited. For simple data audit requirements, the electronic device directly calls an audit instruction with a certain data analysis function, such as an audit instruction with a target detection function, and an audit instruction with a human voice recognition function.

[0044] The audit instructions include multiple types of data to be audited for processing multiple different modal information, and the electronic device calls different types of audit instructions to process the data to be audited according to the second audit requirement. Specifically, the audit instructions corresponding to the data to be audited for multiple different modal information include: audit instructions for processing text data, audit instructions for processing video data, audit instructions for processing image data, and audit instructions for processing voice data.

[0045] In one embodiment, the second audit requirement can be used as the target task for the generation of the audit instruction, and the pre-trained large language model can be used to perform hierarchical task decomposition on the target task to obtain multiple sub-target tasks, and then the execution order of the review instructions corresponding to the sub-target tasks can be planned hierarchically with gradually complex logic based on the thought chain method. Among them, each sub-task can correspond to one or more audit instructions. In order to reduce the complexity of the task and be able to call the corresponding audit model (which can be a different intelligent agent with a pre-trained model) based on the audit instruction to conduct a collaborative audit on the corresponding part of the corresponding data to be audited, it can also reasonably allocate computing resources to prevent excessive computing overhead.

[0046] Exemplarily, the audit instruction for processing text data may be a code for implementing an optical character recognition (OCR) algorithm, which is used to extract text information from images or video frames in order to further audit the audited data; it may also be any text parsing tool based on natural language processing (NLP), which is responsible for understanding the semantics in the text and determining whether it contains sensitive information. The audit instruction for processing video data may be a code for implementing a CogVLM2 model, which is used for understanding and analyzing video content, and can determine whether the scenes, objects, characters, etc. in the video meet the audit standards; it may also be a code for implementing a YOLO (You Only Look Once) model, a widely used object detection algorithm, which is suitable for object recognition in videos and pictures, and can detect sensitive objects in videos in real time; action recognition model: can recognize actions in videos. Application scenario: When a user submits a video, when the task is relatively simple, the system will use YOLO to detect whether the objects in the video involve sensitive content (such as weapons, etc.), and when the video content is relatively complex, the video understanding model will be called to have a deeper understanding of the video content. The audit instructions for processing image data can be instructions for implementing face recognition functions, used to identify face information in images and determine whether specific sensitive people appear; they can also be instructions for implementing image classification, which classifies images through classification models to determine whether images belong to sensitive categories, and can also call the ChatGLM-4V multimodal large model to gain a deeper understanding of images. The audit instructions for processing voice data can be instructions for implementing the Automatic Speech Recognition (ASR) algorithm, which is used to convert the voice in the audio into text, thereby classifying the audio content and determining whether it involves sensitive topics or inappropriate content.

[0047] As another embodiment, when there are multiple audit requirements, it is necessary to simultaneously audit multiple types of violation elements in the audited data. Therefore, in this embodiment, hierarchical task decomposition can be performed based on the audit requirements and the corresponding data to be audited, and multiple audit models corresponding to the audit instructions can be generated according to the decomposed sub-tasks, so as to perform collaborative audits on the corresponding parts of the corresponding data to be audited. At the same time, computing resources can be reasonably allocated to prevent excessive computing overhead from calling different audit models to collaboratively process the audited data, that is, multiple audit instructions are generated, and each audit instruction can call an audit model to audit a mode or a subtask in the multimodal data. For example, when the user needs to handle more complex tasks, such as inputting multimodal data, it is necessary to audit violence, pornography, and hate speech at the same time. , terrorism, false information, copyright infringement and personal privacy, etc., can be audited by planning operations: one of the multiple audit instructions generated calls the CogVLM-2 multimodal video understanding model as a model for processing videos, and another audit instruction in the multiple audit instructions generated calls the ChatGLM-4 text processing model as a model for processing texts. At the same time, further, when processing a certain type of data, the current data of the same type can be further decomposed based on the thinking chain method, and the tasks are decomposed layer by layer according to the gradually complicated logic, and then the corresponding audit instructions are generated according to the decomposed task sounds, and the audit instruction call order is planned, and then the corresponding processing model is called based on the audit instruction and the call order to process the corresponding task. Exemplarily, when it is necessary to identify whether there are certain actions in the video, the target detection model can be called first to identify people or animals. If it is not identified, it indicates that the action does not exist. If a person or an animal is detected, the action in the person or animal is further identified to determine whether it is a target action. In this way, the corresponding part of the corresponding data to be audited can be collaboratively audited, and computing resources can be reasonably allocated to prevent excessive computing overhead.

[0048] For details on how to determine the audit instructions, see Figure 5 For detailed instructions on how to execute the audit instructions, please refer to Figure 6 The corresponding detailed description.

[0049] S23: Process the data to be audited based on the audit instruction to obtain a corresponding audit result.

[0050] In an embodiment of the present application, when there are multiple audit instructions, each audit instruction can be run in sequence according to the execution order of the audit instructions to obtain the candidate audit results output by each audit instruction, and then determined according to the candidate audit results. Figure 7 The corresponding detailed description.

[0051] In one embodiment of the present application, after combining the candidate audit results output by different audit instructions to obtain the audit result corresponding to the data to be audited, it is also determined whether the audit result meets the preset second condition based on the evaluation result corresponding to the audit result, and the output audit result or the generation of the modification strategy is determined based on the judgment result. Among them, the evaluation result can be an evaluation result for the audit result input by the user, and the second condition is used to characterize the degree of match between the audit result and the data to be audited. Exemplarily, the evaluation result can be text data input by the user into the electronic device according to the audit result, for example, "the audit result does not meet the audit requirements", "the audit result has an error" or "the audit result has a format error".

[0052] In one embodiment of the present application, when the audit result satisfies the second condition, indicating that the audit result output by the electronic device has a high degree of match with the data to be audited, the audit result can be directly output. The data to be audited, the first audit requirement, the audit result and the audit instruction are stored in a database, which can ensure that when the electronic device receives new data to be audited and audit requirements in the future, it can determine the relevant audit instructions based on the stored data to be audited and audit requirements, thereby improving the efficiency of data audit. Among them, the database is used to store the data to be audited provided by the user and the interaction information between the user and the electronic device during the audit. In future data audit tasks, the audit instructions corresponding to future audit tasks can be determined based on the data to be audited and the corresponding audit results stored in the database, thereby improving the efficiency of data audit.

[0053] Exemplarily, the electronic device can input the audit results and the evaluation results input by the user into the generative pre-trained model, and output the operation suggestions corresponding to the audit results (for example, shielding the data to be audited, deleting the data to be audited, or performing manual audit). The output results of the generative pre-trained model can be in the form of text reports or charts, which is not limited in this application. The audit results, evaluation results, and operation suggestions are then stored in the database, supporting context tracking of multiple rounds of dialogues, and optimizing the system's review strategy by analyzing historical data; data update: regularly synchronize external data and regulations to ensure that the knowledge base content is kept up to date; security: encrypt and store data to ensure the security of sensitive information while complying with data privacy regulations.

[0054] In one embodiment of the present application, if the audit result does not meet the second condition, indicating that the audit result output by the electronic device has a low degree of match with the data to be audited, a modification strategy for the audit instruction can be generated based on the evaluation result and the audit result. The modification strategy is used to modify the execution order of the audit instruction or the audit instruction itself. For example, the modification strategy can be to increase the number of calls to the audit instruction, or to change the parameters in the audit instruction, and the modification strategy of the audit instruction is stored in the database. For more details on the method of generating the modification strategy, please refer to Figure 8The corresponding detailed description.

[0055] It can be seen from the above technical solutions that the embodiment of the present application obtains a second audit requirement that is highly correlated with the first audit requirement and has more content by analyzing the data to be audited and the first audit requirement. Then, the corresponding audit instruction is determined according to the second audit requirement, and the granularity of data audit is improved through different audit instructions. Finally, the outputs of different audit instructions are combined to obtain the audit results corresponding to the data to be audited, thereby improving the accuracy of data audit.

[0056] like Figure 3 , is a flow chart of a method for determining a second audit requirement provided in an embodiment of the present application. According to different requirements, the order of the steps in the flow chart can be changed, and some steps can be omitted. The method for determining a second audit requirement provided in an embodiment of the present application includes the following steps.

[0057] S30: Determine corresponding modal information according to the data to be reviewed.

[0058] In one embodiment of the present application, in order to improve the accuracy of the review of the data to be reviewed, the corresponding modal information can be first determined based on the data to be reviewed, so as to facilitate the subsequent detailed classification of the content in the data to be reviewed according to different modal information, thereby improving the accuracy of data review.

[0059] S31: Determine a corresponding audit standard according to the first audit requirement.

[0060] In one embodiment of the present application, in order to improve the accuracy of data audit, the electronic device retrieves associated data related to the voice of the first audit requirement from a preset database according to the first audit requirement input by the user. The associated data includes the audit standard corresponding to the first audit requirement. Specifically, the audit standard is used to characterize the standard for evaluating multiple dimensions of the data to be audited. For example, the audit standard can be a security standard, a compliance standard, etc., which is not limited in the present application.

[0061] S32: Determine a second audit requirement according to the modal information and the audit standard.

[0062] In one embodiment of the present application, prompt data can be constructed according to the modal information and the audit standard, and the prompt data can be input into the generative pre-training model to obtain the second audit requirement. The generative pre-training model can be a ChatGLM-4 model or an LLM model, which is not limited in the present application.

[0063] In this way, by determining the corresponding second audit requirement through the audit standard associated with the first audit requirement and the modal information of the data to be audited in the database, the richness of the information contained in the data audit requirement can be improved, thereby improving the accuracy of data audit.

[0064] like Figure 4 , is a flow chart of a method for updating the second audit requirement provided in an embodiment of the present application. According to different requirements, the order of the steps in the flow chart can be changed, and some steps can be omitted. The method for updating the second audit requirement provided in an embodiment of the present application includes the following steps.

[0065] S40, after obtaining the second audit requirement, if it is determined that the second audit requirement does not meet a preset first condition, receiving defect information of the second audit requirement; wherein the first condition is used to characterize the semantics of the first audit requirement.

[0066] In one embodiment of the present application, the first condition can be used to characterize the semantics of the first audit requirement. If the second audit requirement satisfies the first condition, indicating that the semantics of the second audit requirement is consistent with the first audit requirement input by the user, the second audit requirement can be output so that the user can evaluate the second audit requirement and obtain defect information of the second audit requirement.

[0067] Exemplarily, the first audit requirement input by the user into the electronic device may be "whether there is any illegal content in the data to be audited". First, the modal information corresponding to the data to be audited may be parsed, and the corresponding audit standard may be retrieved in the database according to the first audit requirement. The audit standard and modal information obtained are input into the generative pre-training model to obtain a second audit requirement with more information. The specific audit requirement is output and the quality of the generated result is manually judged. The second audit requirement corresponding to the above first audit requirement may be: "Task prompt: content violation audit; target prompt: analyze the given content to determine whether there are any illegal elements; audit scope prompt: video content, text content, image content, link content; audit standard: violent content, hate speech, false information, infringing content, personal privacy information, etc.; output prompt: indicate whether illegal content is found, if there is any violation, point out the specific illegal part and its type, and output the assessment of the degree of violation (minor / medium / serious) for handling suggestions".

[0068] S41, updating the second audit requirement according to the defect information, the modal information, and the audit standard to obtain an updated second audit requirement.

[0069] In one embodiment of the present application, the electronic device also receives defect information input by the user, reconstructs prompt data according to the defect information, modal information, audit standards and second audit requirements, and inputs the reconstructed prompt data into the generative pre-trained model to obtain an updated second audit requirement.

[0070] like Figure 5, is a flow chart of a method for determining an audit instruction provided in an embodiment of the present application. According to different requirements, the order of the steps in the flow chart can be changed, and some steps can be omitted. The method for determining an audit instruction provided in an embodiment of the present application includes the following steps.

[0071] S50: Determine a first audit task in the second audit requirement according to the modal information.

[0072] In one embodiment of the present application, in order to implement different audit methods for content with different modal information in the data to be audited, the first audit task in the second audit requirement can be determined based on the modal information. Among them, each type of modal information corresponds to a first audit task. Exemplarily, when the modal information of the data to be audited is "text data", the corresponding first audit task can be a "text audit task"; when the modal information of the data to be audited is "image data", the corresponding first audit task can be an "image audit task".

[0073] S51: Determine a second audit task in the second audit requirement according to the audit standard and the first audit task.

[0074] In one embodiment of the present application, after determining the first audit task corresponding to different modal information, the first audit task can be further refined according to the audit standard to obtain a second audit task. For example, when the audit standard is "violent content" and the first audit task is "image audit task", the second audit task can be determined as "determining violent content in the image".

[0075] S52: Determine an audit instruction according to the first audit task and the second audit task.

[0076] In one embodiment of the present application, a plurality of corresponding audit instructions may be determined according to the first audit task, and then the audit instruction for processing the second audit task may be determined from the plurality of audit instructions. Exemplarily, when the first audit task is an "image audit task", it may be determined that the corresponding audit instruction is an instruction with a video parsing function (for example, an instruction for implementing a video analysis model), and when the second audit task is "determining violent content in an image", it may be determined that the corresponding audit instruction is an instruction with an action recognition function (for example, an instruction for implementing a CogVLM-2 multimodal video understanding model); when the first audit task is a "text audit task", it may be determined that the corresponding audit instruction is an instruction with a text parsing function (for example, any instruction with a natural language processing function), and when the second audit task is "identifying violent content in text", the corresponding audit instruction may be (for example, an instruction for implementing a ChatGLM-4 model for recognizing text semantics).

[0077] like Figure 6, is a flow chart of a method for executing an audit instruction provided in an embodiment of the present application. According to different requirements, the order of the steps in the flow chart can be changed, and some steps can be omitted. The method for executing an audit instruction provided in an embodiment of the present application includes the following steps.

[0078] S60: Sort the first audit task and the second audit task according to the second audit requirement to obtain an execution order of the first audit task and the second audit task.

[0079] In one embodiment of the present application, in order to improve the accuracy of data auditing, the logic represented by the semantics in the data to be audited can be identified during the auditing process of the data to be audited. And the first audit task and the second audit task are sorted according to the semantic logic in the second audit requirement to determine the execution order of the first audit task and the second audit task. For example, when it is necessary to process a more complex second audit requirement, the data to be audited may include multiple modal information (for example, text, image and video), and the audit criteria may include violent content, hate speech, false information, infringement content and other content. In order to improve the accuracy of the data auditing process, the execution order of the first audit task and the second audit task can be determined according to the semantic information of the second audit requirement. For example, the first audit task includes "video audit task", "image audit task" and "text audit task", and the second audit task includes "audit of violent content in video", "audit of false information in image" and "audit of infringement content in text", then the execution order of the first audit task and the second audit task from front to back is: "video audit task", "audit of violent content in video", "image audit task", "audit of false information in image", "text audit task", "audit of infringement content in text".

[0080] S61, executing the audit instructions corresponding to the first audit task and the second audit task according to the execution order.

[0081] In one embodiment of the present application, the audit instructions corresponding to the first audit task and the second audit task are run according to the execution order, so that the alignment of the data to be audited can be achieved during the data audit process, specifically, including operations such as time, semantics, and physical space position alignment of the data to be audited. For example, when the data audit requirement is to audit the movement of the target in the video, the audit instructions corresponding to the first audit task and the audit instructions corresponding to the second audit task are run according to the execution order. First, the audit instructions with target recognition function are called to process the video data, and several frames containing the target in the output video data are used. The audit instructions with target detection function are used to process multiple video frames to obtain the audit result of the movement of the target in the video. Among them, the audit result can be "the target in the video is running."

[0082] like Figure 7 As shown, it is a flow chart of a method for determining an audit result provided in an embodiment of the present application. According to different requirements, the order of the steps in the flow chart can be changed, and some steps can be omitted. The method for determining an audit result provided in an embodiment of the present application includes the following steps.

[0083] S70, when there are multiple audit instructions, obtaining candidate audit results output by multiple audit instructions.

[0084] In one embodiment of the present application, when there are multiple audit instructions, each audit instruction outputs at least one candidate audit result. For example, the candidate audit result output by the audit instruction with target recognition function may be "the target exists in the third video frame to the seventh video frame"; the candidate audit result output by the audit instruction with action recognition function may be "the target in the fourth video frame to the sixth video frame makes a running action".

[0085] S71, combining the candidate audit results to obtain the audit result corresponding to the data to be audited.

[0086] In one embodiment of the present application, multiple candidate audit results can be combined to obtain the audit result corresponding to the data to be audited. For example, when the candidate audit results include "the target exists in the third video frame to the seventh video frame" and "the target in the fourth video frame to the sixth video frame makes a running action", after combining the candidate audit results, the audit result obtained is "the target exists in the third video frame to the seventh video frame in the video data, and the target makes a running action in the fourth video frame to the sixth video frame".

[0087] like Figure 8 As shown, it is a flow chart of a method for generating a modification strategy provided in an embodiment of the present application. According to different requirements, the order of the steps in the flow chart can be changed, and some steps can be omitted. The method for generating a modification strategy provided in an embodiment of the present application includes the following steps.

[0088] S80: When the audit result does not satisfy the second condition, receive an evaluation result corresponding to the audit result.

[0089] In one embodiment of the present application, if the audit result does not meet the second condition, indicating that the audit result output by the electronic device has a low degree of match with the data to be audited, then an evaluation result corresponding to the audit result may be received. The evaluation result may be text information input into the electronic device after the user evaluates the audit result. For example, the evaluation result may be "the audit result is inaccurate, and the target recognition may be performed on the video multiple times."

[0090] S81, generating a modification strategy for the audit instruction according to the audit result and the corresponding evaluation result.

[0091] In one embodiment of the present application, a modification strategy for the audit instruction can be generated based on the evaluation results and the audit results. The modification strategy is used to modify the execution order of the audit instruction or the audit instruction itself. For example, the modification strategy can be to increase the number of calls to the audit instruction, the modification strategy can also be to change the parameters in the audit instruction, and the modification strategy of the audit instruction is stored in the database. Specifically, prompt data can be constructed based on the evaluation results and the audit results, and the prompt data can be input into the generative pre-trained model to obtain the modification strategy. For example, the modification strategy can be "repeat the audit instruction with target recognition function three times."

[0092] See also Fig. 9 , Fig. 9 1 is a functional module diagram of a data audit device provided in an embodiment of the present application. The data audit device 91 includes an acquisition module 910, a parsing module 911, a planning module 912 and an audit module 913. The module / unit referred to in the present application refers to a series of computer-readable instruction segments that can be executed by the processor 13 and can complete fixed functions, which are stored in the memory 12. In this embodiment, the functions of each module / unit will be described in detail in subsequent embodiments.

[0093] The acquisition module 910 is used to acquire the data to be reviewed and the first review requirement.

[0094] The analysis module 911 is used to determine a second audit requirement according to the data to be audited and the first audit requirement, where the second audit requirement is used to represent the audit requirement associated with the data to be audited.

[0095] The planning module 912 is used to determine the corresponding audit instruction according to the second audit requirement.

[0096] The audit module 913 is used to process the data to be audited based on the audit instruction to obtain a corresponding audit result.

[0097] The analysis module 911 determines the second audit requirement according to the data to be audited and the first audit requirement, including: determining corresponding modal information according to the data to be audited; determining corresponding audit standards according to the first audit requirement; and determining the second audit requirement according to the modal information and the audit standards.

[0098] After the parsing module 911 obtains the second audit requirement, if it is determined that the second audit requirement does not meet the preset first condition, it receives the defect information of the second audit requirement; wherein the first condition is used to characterize the semantics of the first audit requirement; the second audit requirement is updated according to the defect information, the modal information, and the audit standard to obtain an updated second audit requirement.

[0099] The planning module 912 determines the corresponding audit instructions according to the second audit requirement, including: determining the first audit task in the second audit requirement according to the modal information; determining the second audit task in the second audit requirement according to the audit standard and the first audit task; and determining the audit instructions according to the first audit task and the second audit task.

[0100] The planning module 912 is also used to sort the first audit task and the second audit task according to the second audit requirement to obtain the execution order of the first audit task and the second audit task; and run the audit instructions corresponding to the first audit task and the second audit task according to the execution order.

[0101] The audit module 913 processes the data to be audited based on the audit instruction to obtain the corresponding audit result, including: when there are multiple audit instructions, obtaining candidate audit results output by multiple audit instructions; combining the candidate audit results to obtain the audit result corresponding to the data to be audited.

[0102] The audit module 913 is further configured to output a corresponding audit result when the audit result satisfies a preset second condition; wherein the second condition is configured to characterize the degree of matching between the audit result and the data to be audited.

[0103] The audit module 913 is further configured to, if the audit result does not satisfy the second condition, receive an evaluation result corresponding to the audit result; and generate a modification strategy for the audit instruction according to the audit result and the corresponding evaluation result.

[0104] See also Fig.10 , is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device 100 includes a memory 12 and a processor 13. The memory 12 is used to store computer-readable instructions, and the processor 13 is used to execute the computer-readable instructions stored in the memory to implement a data audit method described in any of the above embodiments.

[0105] In an embodiment of the present application, the electronic device 100 further includes a bus, and a computer program stored in the memory 12 and executable on the processor 13, such as a data audit program.

[0106] Fig.10 Only the electronic device 100 having the memory 12 and the processor 13 is shown, and those skilled in the art can understand that Fig.10 The structure shown does not constitute a limitation on the electronic device 100 , and the electronic device 100 may include fewer or more components than shown in the figure, or combine some components, or arrange the components differently.

[0107] Combination Figure 2 The memory 12 in the electronic device 100 stores a plurality of computer-readable instructions to implement the data audit method, and the processor 13 can execute the plurality of instructions to implement: obtaining the data to be audited and a first audit requirement; determining a second audit requirement based on the data to be audited and the first audit requirement, wherein the second audit requirement is used to characterize the audit requirement associated with the data to be audited; determining a corresponding audit instruction based on the second audit requirement; and processing the data to be audited based on the audit instruction to obtain a corresponding audit result.

[0108] Specifically, the specific implementation method of the processor 13 for the above instructions can refer to Figure 2 The description of the relevant steps in the corresponding embodiments will not be repeated here.

[0109] Those skilled in the art will appreciate that the schematic diagram is merely an example of the electronic device 100 and does not constitute a limitation on the electronic device 100. The electronic device 100 may be a bus-type structure or a star-type structure. The electronic device 100 may also include more or less other hardware or software than shown in the diagram, or a different arrangement of components. For example, the electronic device 100 may also include input and output devices, network access devices, etc.

[0110] It should be noted that the electronic device 100 is only an example, and other existing or future electronic products that are suitable for the present application should also be included in the protection scope of the present application and included here by reference.

[0111] Among them, the memory 12 includes at least one type of readable storage medium, and the readable storage medium can be non-volatile or volatile. The readable storage medium includes flash memory, mobile hard disk, multimedia card, card-type memory (for example: SD or DX memory, etc.), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 12 can be an internal storage unit of the electronic device 100, such as a mobile hard disk of the electronic device 100. In other embodiments, the memory 12 can also be an external storage device of the electronic device 100, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. equipped on the electronic device 100. The memory 12 can not only be used to store application software and various types of data installed in the electronic device 100, such as a code of a data audit program, etc., but also can be used to temporarily store data that has been output or is to be output.

[0112] In some embodiments, the processor 13 may be composed of an integrated circuit, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and combinations of various control chips. The processor 13 is the control core (Control Unit) of the electronic device 100, and uses various interfaces and lines to connect various components of the entire electronic device 100, and executes various functions of the electronic device 100 and processes data by running or executing programs or modules stored in the memory 12 (for example, executing data audit programs, etc.), and calling data stored in the memory 12.

[0113] The processor 13 executes the operating system of the electronic device 100 and various installed applications. The processor 13 executes the applications to implement the steps in the above-mentioned various data audit method embodiments, for example Figure 2 Steps shown.

[0114] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory 12 and executed by the processor 13 to complete the present application. The one or more modules / units may be a series of computer-readable instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program in the electronic device 100. For example, the computer program may be divided into an acquisition module 910, a parsing module 911, a planning module 912, and an audit module 913.

[0115] The above-mentioned integrated unit implemented in the form of a software function module can be stored in a computer-readable storage medium. The above-mentioned software function module is stored in a storage medium, and includes a number of instructions for enabling a computer device (which can be a personal computer, a computer device, or a network device, etc.) or a processor to execute a part of a data audit method described in each embodiment of the present application.

[0116] If the module / unit integrated in the electronic device 100 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 present application implements all or part of the processes in the above-mentioned embodiment method, and can also instruct the relevant hardware devices to complete through a computer program. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by the processor, the steps of each of the above-mentioned method embodiments can be implemented.

[0117] The computer program includes computer program code, which may be in source code form, object code form, executable file or some intermediate form, etc. The computer readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory and other memory, etc.

[0118] Furthermore, the computer-readable storage medium 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, etc.; the data storage area may store data created according to the use of the blockchain node, etc.

[0119] The bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. Fig.10 Only one arrow is used in the figure, but it does not mean that there is only one bus or one type of bus. The bus is configured to realize the connection and communication between the memory 12 and at least one processor 13, etc.

[0120] An embodiment of the present application also provides a computer-readable storage medium (not shown), in which computer-readable storage medium is stored computer-readable instructions, and the computer-readable instructions are executed by a processor in an electronic device to implement the data review method described in any of the above embodiments.

[0121] In the several embodiments provided in this 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 illustrative, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0122] The modules described as separate components may or may not be physically separated, and the components shown as modules 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 modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0123] In addition, each functional module in each embodiment of the present application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0124] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the specification can also be implemented by one unit or device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any specific order.

[0125] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present application and are not intended to limit it. Although the present application has been described in detail with reference to the preferred embodiments, a person of ordinary skill in the art should understand that the technical solution of the present application may be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present application.

Claims

1. A data audit method, characterized in that: The method comprises: Obtain the data to be reviewed and the first review requirements; Determining a second audit requirement according to the data to be audited and the first audit requirement, wherein the second audit requirement is used to represent the audit requirement associated with the data to be audited; Determining a corresponding audit instruction according to the second audit requirement; The data to be audited is processed based on the audit instruction to obtain a corresponding audit result.

2. The data audit method according to claim 1, characterized in that: Determining the second audit requirement according to the to-be-audited data and the first audit requirement includes: Determining corresponding modal information according to the data to be reviewed; Determine the corresponding audit standard according to the first audit requirement; A second audit requirement is determined based on the modal information and the audit standard.

3. The data audit method according to claim 2, characterized in that: The method further comprises: After obtaining the second audit requirement, if it is determined that the second audit requirement does not meet a preset first condition, receiving defect information of the second audit requirement; wherein the first condition is used to characterize the semantics of the first audit requirement; The second audit requirement is updated according to the defect information, the modal information, and the audit standard to obtain an updated second audit requirement.

4. The data audit method according to claim 2, characterized in that: Determining the corresponding audit instruction according to the second audit requirement includes: Determining a first audit task in the second audit requirement according to the modal information; Determine a second audit task in the second audit requirement according to the audit standard and the first audit task; An audit instruction is determined according to the first audit task and the second audit task.

5. The data audit method according to claim 4, characterized in that: The method further comprises: According to the second audit requirement, the first audit task and the second audit task are sorted to obtain an execution order of the first audit task and the second audit task; The audit instructions corresponding to the first audit task and the second audit task are executed according to the execution order.

6. The data audit method according to claim 1, characterized in that: Processing the data to be audited based on the audit instruction to obtain a corresponding audit result includes: In the case where there are multiple audit instructions, obtaining candidate audit results output by the multiple audit instructions; The candidate audit results are combined to obtain the audit result corresponding to the data to be audited.

7. The data audit method according to claim 1, characterized in that: The method further comprises: When the audit result satisfies a preset second condition, the corresponding audit result is output; wherein the second condition is used to characterize the matching degree between the audit result and the data to be audited.

8. The data audit method according to claim 7, characterized in that: The method further comprises: If the audit result does not satisfy the second condition, receiving an evaluation result corresponding to the audit result; A modification strategy for the audit instruction is generated according to the audit result and the corresponding evaluation result.

9. A data audit device, characterized in that: The device comprises: An acquisition module is used to acquire the data to be reviewed and the first review requirements; A parsing module, configured to determine a second audit requirement according to the data to be audited and the first audit requirement, wherein the second audit requirement is used to represent the audit requirement associated with the data to be audited; A planning module, used to determine corresponding audit instructions according to the second audit requirement; The audit module is used to process the data to be audited based on the audit instruction to obtain a corresponding audit result.

10. An electronic device, characterized in that: The electronic device comprises a processor and a memory, and the processor is used to implement the data audit method according to any one of claims 1 to 8 when executing a computer program stored in the memory.

Images