Advertisement content isolation auditing method, system and device and medium

Abstract

The invention is suitable for the field of advertisement auditing, and discloses an advertisement content isolation auditing method, system and device and a medium, and the method comprises the steps: receiving advertisement content; the advertisement content is classified, and text advertisements, picture advertisements and video advertisements are identified and classified; the advertisement contents of different classifications are audited respectively; wherein when the picture advertisements are classified, digital watermarks are embedded in the picture advertisements, and the digital watermarks comprise category information of the picture advertisements. In the picture advertisement classification and auditing process, the digital watermarking technology is innovatively adopted in the picture advertisement classification and auditing method. By embedding the category information of the picture advertisement into the digital watermark and adding the watermark into the picture advertisement, quick classification and efficient management of the picture advertisement are realized, the accuracy and efficiency of picture advertisement classification are improved, convenience is provided for subsequent auditing work, and the user experience is improved. And the overall quality and efficiency of advertisement auditing are further improved.

Description

Technical Field

[0001] The present invention relates to the field of advertisement review, and in particular to an isolated review method, system, equipment and medium for advertisement content. Background Art

[0002] In modern advertising systems, content review is an important part of ensuring the legality and compliance of advertising content. Existing advertising review technologies mainly rely on manual review and simple rule matching. These methods are inefficient when dealing with large-scale advertising content and are prone to missed reviews and misreviews. In addition, faced with massive and diverse advertising content, especially image ads, traditional classification methods are often inefficient and inaccurate. Summary of the invention

[0003] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the present invention proposes a method, system, device and medium for isolating and reviewing advertising content.

[0004] In a first aspect, the present application proposes a method for isolating and reviewing advertising content, comprising: receiving advertising content;

[0005] Classify advertising content and identify and categorize text ads, image ads, and video ads;

[0006] Review the contents of the advertisements in different categories respectively;

[0007] Wherein, when the picture advertisement is classified, a digital watermark is embedded in the picture advertisement, and the digital watermark includes category information of the picture advertisement.

[0008] More specifically, in the above technical solution, the category information is embedded into the picture advertisement by using digital watermark technology to generate a watermark picture carrying advertisement metadata;

[0009] Analyzing the watermark image using an image recognition algorithm, extracting the advertisement metadata carried in the watermark image, and determining the advertisement category to which the image advertisement belongs;

[0010] A corresponding review strategy is adopted according to the advertisement category to which the image advertisement belongs.

[0011] More specifically, in the above technical solution, when reviewing the advertisement content:

[0012] Obtaining a video advertisement to be reviewed, extracting key frames from the video advertisement, and obtaining a key frame sequence of the video advertisement;

[0013] Based on the key frame sequence, the original key frame is transformed to generate a plurality of variant key frames through operations of image flipping, rotation, scaling, cropping and color transformation;

[0014] The original key frame and the variant key frame are merged to obtain an expanded key frame sequence, and the expanded key frame sequence is used as an expanded representation of the video advertisement.

[0015] More specifically, in the above technical solution, whether the video advertisement contains illegal content is determined according to a preset illegal content label system;

[0016] If any key frame of the video advertisement is determined to be in violation, determining that the video advertisement is an in violation video advertisement;

[0017] Based on the video advertisement determined to be in violation, determining the infringing segments in the video advertisement by locating the infringing key frames;

[0018] The video advertisement is processed accordingly according to the type and severity of the offending segment.

[0019] More specifically, in the above technical solution, when reviewing the advertisement content:

[0020] Obtain the video frame sequence and audio data of the video advertisement to be analyzed through the video dual stream technology;

[0021] Based on the video frame sequence, a pre-trained visual feature extraction model is used to extract a visual feature vector of each frame;

[0022] Based on the audio data, using a pre-trained audio feature extraction model to extract an audio feature vector of the audio;

[0023] Inputting the extracted visual feature vector and the extracted audio feature vector into a multimodal fusion network for feature fusion to obtain a fused multimodal feature representation;

[0024] Inputting the multimodal feature representation into a pre-trained video advertisement illegal content recognition model to determine whether the video advertisement contains illegal content;

[0025] If the video advertisement contains illegal content, outputting the illegal category;

[0026] A review result of the video advertisement is determined according to the violation category.

[0027] More specifically, in the above technical solution, the method further includes:

[0028] The audit result is fed back to the advertisement delivery system, and the advertisement delivery system delivers the advertisement content based on the audit result.

[0029] More specifically, in the above technical solution, after the advertisement delivery system receives the review result, if the advertisement content passes the review, the advertisement delivery system adds the reviewed advertisement content to the delivery queue; if the advertisement content has failed the review for multiple times, the advertisement delivery system reduces its delivery priority or restricts its delivery channel;

[0030] After the advertisement content is delivered, the advertisement delivery system monitors the delivered advertisement content in real time, and the advertisement delivery system adjusts the delivery strategy based on the monitoring result.

[0031] In a second aspect, the present application proposes an isolated review system for advertising content, comprising:

[0032] The first processing module: receiving advertisement content;

[0033] The second processing module: classifies the advertisement content and identifies and classifies the advertisements into text advertisements, picture advertisements and video advertisements;

[0034] The third processing module: reviewing the advertisement contents of different categories respectively;

[0035] Wherein, when the picture advertisement is classified, a digital watermark is embedded in the picture advertisement, and the digital watermark includes category information of the picture advertisement.

[0036] In a third aspect, the present application further proposes a terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements any of the methods described above when executing the computer program.

[0037] In a fourth aspect, the present application further proposes a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method as described in any one of the above items is implemented.

[0038] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0039] 1. By analyzing the watermark image through the image recognition algorithm, the advertising category metadata carried in the watermark can be quickly extracted, so as to quickly determine the advertising category to which the image advertisement belongs, which greatly improves the classification speed and reduces the workload of manual review; since digital watermark technology is highly stable and concealed, it can ensure the accuracy and completeness of the advertising category metadata. Therefore, the watermark-based classification technology can more accurately identify the classification information of the image advertisement and reduce the classification error rate; through digital watermark technology, it is also possible to quickly retrieve and locate the image advertisement. Once a certain type of advertisement needs to be modified or deleted, the relevant advertisement can be quickly found and processed, which improves the efficiency of advertising management and reduces management costs.

[0040] 2. When reviewing video ads, use video content augmentation technology to expand and enhance the key frames of video ads to generate more diverse video content representations, improve the generalization ability of the video ad review model, and enable it to identify more variants of illegal video content.

[0041] 3. Use video dual-stream technology to process video ads. Video dual-stream technology can process the visual information and audio information of the video at the same time. By extracting video frames and audio features, it can achieve a comprehensive analysis of the video ad content. This multimodal ad review method can more comprehensively and accurately identify illegal content in video ads. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0043] Figure 1 It is a flow chart of a method for isolating and reviewing advertisement content provided by an embodiment of the present invention;

[0044] Figure 2 It is a structural diagram of an isolated review system for advertising content provided by an embodiment of the present invention;

[0045] Figure 3 It is a schematic diagram of the structure of the terminal device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0046] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0047] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, wholes, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or combinations thereof.

[0048] It should also be understood that the term “and / or” used in the specification and appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0049] As used in the specification and appended claims of this application, the term "if" can be interpreted as "when" or "uponce" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "uponce it is determined" or "in response to determining" or "uponce [described condition or event] is detected" or "in response to detecting [described condition or event]", depending on the context.

[0050] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0051] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0052] See also Figure 1The present application proposes a method for isolating and reviewing advertising content, including: S101 receiving advertising content; S102 classifying the advertising content, identifying and classifying text advertisements, picture advertisements and video advertisements; S103 embedding a digital watermark in the picture advertisement when classifying the picture advertisement, the digital watermark including the category information of the picture advertisement; S104 reviewing the advertising content of different categories respectively.

[0053] By adding invisible digital watermarks to image ads and carrying metadata of ad categories, rapid classification of image ads can be achieved. This watermark-based ad classification technology can improve the efficiency and accuracy of image ad classification. It includes:

[0054] Acquire the received advertisement content data, and determine the type of the advertisement content according to the format of the advertisement content data;

[0055] If the advertisement content is in a picture format, determining the advertisement content as a picture advertisement;

[0056] For the picture advertisement, digital watermark technology is used to embed advertisement category information into the picture advertisement to generate a watermark image carrying advertisement metadata;

[0057] Analyzing the watermark image using an image recognition algorithm, extracting the advertisement category metadata carried in the watermark image, and determining the advertisement category to which the image advertisement belongs based on the advertisement category metadata;

[0058] According to the classification result of the advertisement content, a corresponding audit strategy is adopted to audit the advertisement content.

[0059] Optionally, during the review process, natural language processing and image recognition technology are used to perform semantic analysis and scene recognition on the advertising content to determine whether the advertising content complies with preset laws, regulations and social ethical standards. The review results are fed back to the advertiser, and for advertising content that fails the review, the advertiser is informed of specific review opinions to guide the advertiser to modify and optimize the advertising content.

[0060] Specifically, in the advertising content review system, the type of advertisement is first determined by the content format. For example, if the advertising content is in a picture format such as JPG or PNG, it is determined to be a picture advertisement. For picture advertisements, digital watermarking technology is used to embed the advertising category information into the picture. In specific implementation, the LSB algorithm can be used to encode the advertising category metadata into a binary bit stream, replace the least significant bit of the picture pixel, and form a picture carrying a watermark. The watermarked picture is analyzed by image recognition algorithms such as CNN, and the 64-bit metadata carried in the watermark is extracted and matched with the preset advertising category mapping table to quickly determine the advertising category to which the picture belongs, such as beauty, automobiles, travel, or food. According to the classification results, the corresponding review strategy is adopted.

[0061] For example, if the image is a beauty advertisement, the focus will be on reviewing whether there is false propaganda, banned cosmetics and other content. During the review process, the cosmetics brand logo in the image is identified through target detection algorithms such as YOLOv3 to determine whether it is a well-known brand. Combined with the OCR text recognition results, the authenticity of the advertisement is comprehensively determined. For advertising images that fail the review, feedback will be given to the advertiser, such as "The image is suspected of exaggerating the product effect, please provide product efficacy data", etc., to guide the advertiser to optimize the advertising content and improve the quality of the advertisement. For example, if the image is an automobile advertisement, the focus will be on reviewing whether it involves false propaganda, such as exaggeration in vehicle performance, fuel consumption, safety level, etc. In addition, attention should be paid to whether it involves illegal content, such as unauthorized vehicle modification, theft of other people's vehicle images, etc.; vehicle recognition algorithms (such as deep learning models) are used to identify the model in the advertisement, and combined with public vehicle performance data to compare whether the description in the advertisement is true. At the same time, OCR technology is used to identify text information in advertisements, and compliance checks are carried out against relevant national laws and regulations. For automobile advertisements that fail the review, feedback is given to advertisers, such as "The fuel consumption data of the vehicle in the advertisement is inconsistent with the actual situation. Please provide authoritative data proof" and other review opinions, and advertisers are required to make corrections. For example, if the image belongs to a food advertisement, the focus of the review is whether it involves food safety issues, such as whether the food ingredients are safe, whether the production date and shelf life are accurate, and whether there are illegal additives. At the same time, attention should also be paid to whether there is any behavior that exaggerates the efficacy of the product or misleads consumers; the appearance and packaging of the food are detected by image recognition technology, and compared with national food safety standards and relevant regulations to check whether there are any violations; for the text description part, natural language processing technology is used for semantic analysis to ensure the authenticity and accuracy of the information; for food advertisements that fail the review, feedback is given to advertisers, such as "The production date and shelf life are not marked in the advertisement, which does not comply with food safety regulations. Please correct it immediately" and other review opinions, and advertisers are required to supplement relevant information. For example, if the image is an education and training advertisement, the focus will be on reviewing whether it involves false promises (such as guaranteeing passing exams, quickly mastering skills, etc.), and whether it complies with relevant regulations and requirements of the education department; at the same time, attention should also be paid to whether key information such as the real name, address, and faculty of the training institution are marked in the advertisement; use the education institution identification database to verify the authenticity and legality of the training institution in the advertisement, and combine the relevant regulations of the education department to review whether the promises and publicity in the advertisement are compliant; at the same time, use OCR technology to extract the text content in the advertisement, conduct detailed review and analysis, and ensure the authenticity and accuracy of the advertising information.

[0062] It is understandable that the advertising content can be received via HTTPAP I and stored in a MySQL database.

[0063] In some embodiments, when reviewing video ads, video content expansion technology is used to expand and enhance the key frames of video ads to generate more diverse video content representations. This video content expansion technology can improve the generalization ability of the video ad review model, enabling it to identify more variants of illegal video content. For video ads that have passed the review, the expanded key frame sequence can be used to generate diversified preview images and promotional materials for the video ads, enriching the display form of the video ads and improving the user experience. Including:

[0064] Acquire a video advertisement to be reviewed, extract key frames from the video advertisement, and obtain a key frame sequence of the video advertisement;

[0065] For the extracted key frame sequence, the key frame is transformed by using data enhancement technology, and multiple variant key frames are generated through image flipping, rotation, scaling, cropping, and color transformation operations;

[0066] Merging the original key frame and the variant key frame to obtain an expanded key frame sequence as an expanded representation of the video advertisement;

[0067] According to a preset illegal content label system, multi-label classification is performed on the expanded key frame sequence to determine whether the video advertisement contains illegal content;

[0068] If any key frame of the video advertisement is determined to be illegal, then the video advertisement is determined to be an illegal video advertisement;

[0069] For the video advertisement determined to be in violation, the infringing segments in the video advertisement are determined by locating the infringing key frames, and the video advertisement is processed accordingly according to the type and severity of the infringing content;

[0070] Optionally, the expanded key frame sequence is input into a video advertisement review model for training, and the expanded video content representation is used to improve the generalization ability and recognition performance of the video advertisement review model.

[0071] Specifically, during the video advertisement review process, representative key frames are first extracted from the video through key frame extraction algorithms such as K-Means clustering or KTS (Kernel Temporal Segmentation), and generally one key frame is extracted every 5-10 seconds. Then, data enhancement technology is used to randomly transform the key frames, such as applying the OpenCV library to perform horizontal flipping, vertical flipping, 90° rotation, 8-2 times scaling, random cropping of 80% area, and adjusting the hue ±10° in the HSV color space. Each key frame generates 5 variants to expand the diversity of video representation. Next, based on CNN networks such as ResNet-50, the expanded key frames are multi-label classified to determine whether they contain illegal content such as pornography, violence, and contraband. The classification confidence threshold is set to 7. If any frame violates the rules, the entire video is judged to be in violation and the illegal segment is located. According to the type of violation and confidence, corresponding measures are taken, such as directly deleting pornographic content and blurring the smoking scene. The expanded key frames are input into the video review model for continuous training, and 50 epochs are iterated on 50,000 samples to enable the model to identify more changing violation features. For normal videos, the three key frames with the highest confidence and their variants are selected as preview images and promotional materials to enhance the attractiveness of video ads. Through the above process, a complete closed loop from key frame extraction, data enhancement, violation identification to content disposal is established, forming a standardized and efficient video advertising review mechanism.

[0072] According to the type and severity of the illegal content, the video advertisement is handled accordingly, which may include: first, establishing a clear classification system for illegal content. For example, illegal content can be divided into multiple categories such as pornography, political sensitivity, advertising violations, false information, malicious attacks, and illegal and illegal content. Among them, technical means such as keyword matching and semantic analysis are used to compare the identified illegal content with the standards in the classification system to determine its category. In the case of possible misjudgment in automatic machine classification, manual review can be performed to ensure the accuracy of the classification.

[0073] Secondly, determine the corresponding evaluation indicators based on the nature and severity of the illegal content. For example, pornographic and vulgar content can be evaluated for its degree of explicitness and vulgarity; politically sensitive content can be evaluated for whether it violates national laws and regulations and whether it has caused social controversy. Conduct quantitative evaluation of the identified illegal content, such as counting the number, frequency, and duration of illegal content, so as to more accurately understand the violation. Combine the classification of illegal content with the quantitative evaluation results to conduct a comprehensive evaluation.

[0074] Finally, based on the evaluation results, determine how to deal with illegal content. For example, for illegal content that clearly violates platform regulations or laws and regulations, it should be directly deleted or refused to be published, including but not limited to serious illegal content such as pornography, violence, propaganda of contraband, false propaganda, etc.; for minor illegal or editable content, the creator can be asked to modify it and resubmit it for review after modification, such as inappropriate language expression, minor video content violations, etc.; for some content that may be controversial or require further evaluation, it can be temporarily shelved and its development trend can be observed. At the same time, communication can be conducted with the creator to understand its creative intention and background in order to more accurately evaluate its compliance.

[0075] In some embodiments, video ads are processed using dual-stream video technology. Dual-stream video technology can process visual information and audio information of a video at the same time, and can achieve a comprehensive analysis of the content of a video ad by extracting video frames and audio features. This multimodal ad review method can more comprehensively and accurately identify illegal content in video ads. It includes:

[0076] Acquire a video frame sequence and audio data of a video advertisement to be analyzed, wherein the video frame sequence and audio data are obtained by using a video dual-stream technology;

[0077] For the video frame sequence, a pre-trained visual feature extraction model is used to extract a multi-dimensional visual feature vector of each frame;

[0078] For the audio data, a pre-trained audio feature extraction model is used to extract a multi-dimensional feature vector of the audio;

[0079] The extracted visual feature vector sequence and audio feature vector are input into a multimodal fusion network for feature fusion to obtain a fused multimodal feature representation;

[0080] Inputting the multimodal feature representation into a pre-trained video advertisement illegal content recognition model to determine whether the video advertisement contains illegal content;

[0081] If the video advertisement contains illegal content, outputting the illegal category;

[0082] Determining a review result of the video advertisement according to the violation category;

[0083] Optionally, the review results are fed back to advertisers; sample data of manually reviewed video advertisements are continuously collected, and the visual feature extraction model, the audio feature extraction model and the illegal content recognition model are regularly fine-tuned to improve the recognition accuracy of the models.

[0084] Specifically, first, the video frame sequence of the video advertisement to be analyzed is obtained through the video decoder using the video dual-stream technology, with the frame rate set to 25fps. At the same time, the audio data of the video is extracted, and the sampling rate is set to 41kHz. Then, for the obtained video frame sequence, the pre-trained ResNet-50 visual feature extraction model is used to extract the 2048-dimensional visual feature vector of each frame. For the obtained audio data, the pre-trained VGG i sh audio feature extraction model is used to extract the 128-dimensional audio feature vector and convert it into the same length as the visual feature vector sequence. Then, the extracted visual feature vector sequence and audio feature vector are input into the multimodal Transformer fusion network for feature fusion. Through the self-attention mechanism and feedforward neural network, the interaction relationship between different modalities is learned to obtain the fused 1024-dimensional multimodal feature representation. Finally, the multimodal feature representation is input into the pre-trained video advertisement illegal content recognition model. The model uses the TextCNN structure to determine whether the video advertisement contains illegal content such as violence, pornography, false propaganda, etc. If so, the corresponding violation category is output. Based on the violation category output by the recognition model, the review result of the video ad is determined. For any violation ad, it is rejected and fed back to the advertiser for modification. At the same time, we continue to collect sample data of manually reviewed video ads, and regularly use new samples to fine-tune the visual feature extraction model, audio feature extraction model, and violation content recognition model, continuously improve the recognition accuracy of the model, and realize the automation and intelligence of video ad review.

[0085] In some embodiments, the audit results are fed back to the advertisement delivery system, and the advertisement delivery system delivers the advertisement content based on the audit results.

[0086] Develop an API interface between the review system and the advertising delivery system to ensure that the two can communicate in real time; define clear interface protocols and data formats, including status codes and error messages of review results; ensure that the advertising content in the review system is synchronized with the advertising content in the advertising delivery system, and pass the unique identifier of the advertisement (such as the advertising ID) to the review system when the advertisement is submitted for review.

[0087] After the audit system completes the audit, it will update the audit results (such as passed, rejected, pending modification, etc.) and the corresponding status code to the database in real time; the audit system will send the audit results to the advertising delivery system through the API interface. If an error occurs during the audit process (such as content violation, format error, etc.), the audit system should return detailed error information so that the advertising delivery system can handle it. After receiving the error information, the advertising delivery system should perform corresponding error processing, such as notifying the advertiser, refusing to publish the advertisement, etc.

[0088] The advertising delivery system selects the advertising content that meets the delivery requirements based on the audit results received. For the ads that pass the audit, the advertising delivery system adds them to the delivery queue; for the rejected or pending ads, the system handles them accordingly. Based on the audit results, the advertising delivery system can dynamically adjust the delivery strategy. For example, for ads that have been rejected multiple times due to content violations, the advertising delivery system can lower their delivery priority or restrict their delivery channels.

[0089] The advertising delivery system should monitor the delivered ads in real time, including delivery effects, audience reactions, etc. Based on the monitoring results, the advertising delivery system can further optimize the delivery strategy, such as adjusting the delivery time, frequency, and region.

[0090] Optionally, encryption technology should be used to ensure data security when transmitting data between the audit system and the advertising delivery system; the encryption method can be SSL / TLS protocol or other secure transmission protocols. When receiving the audit results, the advertising delivery system should perform data verification to ensure the integrity and accuracy of the data; if the data verification fails, the advertising delivery system should refuse to receive the audit results and notify the audit system to resend them. The audit system and advertising delivery system should record detailed log information, including data transmission time, status code, error information, etc.; these log information will help with subsequent troubleshooting and data analysis.

[0091] It should be noted that for text ads, first, the system receives the text ad content to be reviewed. These texts may come from direct submission by advertisers, crawling by ad platforms or other channels. After receiving the text ads, the system pre-processes them, including removing irrelevant characters (such as special symbols, HTML tags, etc.), word segmentation, removing stop words, etc., for subsequent analysis.

[0092] Next, the system performs semantic analysis on the pre-processed text ads. This may include but is not limited to: matching keywords in text ads with a preset illegal keyword library to detect whether there are sensitive words or illegal content; identifying key components such as the subject, predicate, and object in the ads by analyzing the syntactic structure of the text to further understand the semantic content of the ads; and performing sentiment analysis on text ads to determine whether they contain inappropriate remarks such as excessive exaggeration or false propaganda.

[0093] Based on semantic analysis, the system conducts compliance checks on text ads, including: ensuring that text ads do not violate national laws and regulations, such as the Advertising Law, the Consumer Rights Protection Law, etc.; checking whether text ads comply with social ethics and public order and morality, and avoiding the use of vulgar, discriminatory or offensive language.

[0094] Based on the results of semantic analysis and compliance checks, the system will review the text ads. This can include: if the content of the text ad is legal and compliant, and there is no inappropriate speech, it will be considered as passed; rejection of review: if the text ad contains illegal content or inappropriate speech, it will be considered as rejected for review, and the corresponding reasons for the violation and suggestions will be given; if the text ad has problems that can be modified, such as some inaccurate expressions or may cause misunderstandings, it will be considered to be pending modification, and modification suggestions will be given.

[0095] Finally, the system will feed back the review results to the advertising delivery system. The advertising delivery system will make delivery decisions on the advertising content based on the review results: for ads that pass the review, they will be added to the delivery queue and delivered according to the preset delivery strategy; for ads that refuse to be reviewed, they will be rejected and the review results and modification suggestions will be fed back to the advertiser; for ads to be modified, they will be returned to the advertiser for modification and submitted for review again after modification.

[0096] See also Figure 2 , the application also proposes an isolated review system for advertising content, including:

[0097] First processing module 201: receiving advertisement content;

[0098] The second processing module 202: classifies the advertisement content, and identifies and classifies the advertisements into text advertisements, picture advertisements and video advertisements;

[0099] The third processing module 203: reviewing the advertisement contents of different categories respectively;

[0100] Wherein, when the picture advertisement is classified, a digital watermark is embedded in the picture advertisement, and the digital watermark includes category information of the picture advertisement.

[0101] It is understandable that if Figure 1 The contents of the embodiment of the isolation and audit method of the advertising content shown in the figure are applicable to the embodiment of the isolation and audit system of the advertising content. The functions specifically implemented by the embodiment of the isolation and audit system of the advertising content are similar to those of the embodiment of the advertising content isolation and audit system shown in the figure. Figure 1 The embodiment of the method for isolating and reviewing the advertisement content shown in the figure is the same as that in the embodiment of the method for isolating and reviewing the advertisement content shown in the figure, and the beneficial effects achieved are the same as those achieved in the embodiment of the method for isolating and reviewing the advertisement content shown in the figure. Figure 1 The beneficial effects achieved by the illustrated embodiment of the method for isolating and reviewing advertising content are also the same.

[0102] It should be noted that the information interaction, execution process and other contents between the above-mentioned systems are based on the same concept as the embodiment of the method of the present invention. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0103] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0104] See also Figure 3 The embodiment of the present invention further provides a computer device 3, comprising: a memory 302 and a processor 301 and a computer program 303 stored in the memory 302. When the computer program 303 is executed on the processor 301, the isolation and review method of the advertising content as described in any one of the above methods is implemented.

[0105] The computer device 3 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The computer device 3 may include, but is not limited to, a processor 301 and a memory 302. Those skilled in the art will appreciate that Figure 3 It is only an example of computer device 3 and does not constitute a limitation on computer device 3. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components, for example, it may also include input and output devices, network access devices, etc.

[0106] The processor 301 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0107] In some embodiments, the memory 302 may be an internal storage unit of the computer device 3, such as a hard disk or memory of the computer device 3. In other embodiments, the memory 302 may also be an external storage device of the computer device 3, such as a plug-in hard disk, a smart memory card (Smart Med iaCard, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the computer device 3. Further, the memory 302 may also include both an internal storage unit of the computer device 3 and an external storage device. The memory 302 is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as the program code of the computer program, etc. The memory 302 may also be used to temporarily store data that has been output or is to be output.

[0108] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the method for isolating and reviewing advertising content as described in any one of the above methods is implemented.

[0109] In this embodiment, 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 present application implements all or part of the processes in the above-mentioned embodiment method, which can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium can at least include: any entity or device that can carry the computer program code to the camera / terminal device, recording medium, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. For example, a USB flash drive, a mobile hard disk, a magnetic disk or an optical disk. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electric carrier signals and telecommunication signals.

[0110] The embodiments described above 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, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such 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, and should all be included in the protection scope of the present application.

Claims

1. A method for isolating and reviewing advertising content, characterized in that: The method comprises: Receive advertising content; Classify advertising content and identify and categorize text ads, image ads, and video ads; Review the contents of the advertisements in different categories respectively; Wherein, when the picture advertisement is classified, a digital watermark is embedded in the picture advertisement, and the digital watermark includes category information of the picture advertisement.

2. The method for isolating and reviewing advertisement content according to claim 1, characterized in that: Using digital watermark technology, embedding the category information into the picture advertisement to generate a watermarked picture carrying advertisement metadata; Analyzing the watermark image using an image recognition algorithm, extracting the advertisement metadata carried in the watermark image, and determining the advertisement category to which the image advertisement belongs; A corresponding review strategy is adopted according to the advertisement category to which the image advertisement belongs.

3. The method for isolating and reviewing advertisement content according to claim 1, characterized in that: When reviewing the content of said advertisements: Obtaining a video advertisement to be reviewed, extracting key frames from the video advertisement, and obtaining a key frame sequence of the video advertisement; Based on the key frame sequence, the original key frame is transformed to generate a plurality of variant key frames through operations of image flipping, rotation, scaling, cropping and color transformation; The original key frame and the variant key frame are merged to obtain an expanded key frame sequence, and the expanded key frame sequence is used as an expanded representation of the video advertisement.

4. The method for isolating and reviewing advertisement content according to claim 3, characterized in that: Determine whether the video advertisement contains illegal content according to a preset illegal content label system; If any key frame of the video advertisement is determined to be in violation, determining that the video advertisement is an in violation video advertisement; Based on the video advertisement determined to be in violation, determining the infringing segments in the video advertisement by locating the infringing key frames; The video advertisement is processed accordingly according to the type and severity of the offending segment.

5. The method for isolating and reviewing advertisement content according to claim 1, characterized in that: When reviewing the content of said advertisements: Obtain the video frame sequence and audio data of the video advertisement to be analyzed through the video dual stream technology; Based on the video frame sequence, a pre-trained visual feature extraction model is used to extract a visual feature vector of each frame; Based on the audio data, using a pre-trained audio feature extraction model to extract an audio feature vector of the audio; Inputting the extracted visual feature vector and the extracted audio feature vector into a multimodal fusion network for feature fusion to obtain a fused multimodal feature representation; Inputting the multimodal feature representation into a pre-trained video advertisement illegal content recognition model to determine whether the video advertisement contains illegal content; If the video advertisement contains illegal content, outputting the illegal category; A review result of the video advertisement is determined according to the violation category.

6. The method for isolating and reviewing advertisement content according to any one of claims 1 to 5, characterized in that: The method further comprises: The audit result is fed back to the advertisement delivery system, and the advertisement delivery system delivers the advertisement content based on the audit result.

7. The method for isolating and reviewing advertisement content according to claim 6, characterized in that: After the advertisement delivery system receives the review result, if the advertisement content passes the review, the advertisement delivery system will add the reviewed advertisement content to the delivery queue; if the advertisement content has failed the review for multiple times, the advertisement delivery system will lower its delivery priority or limit its delivery channel; After the advertisement content is delivered, the advertisement delivery system monitors the delivered advertisement content in real time, and the advertisement delivery system adjusts the delivery strategy based on the monitoring result.

8. An isolated review system for advertising content, characterized in that: include: The first processing module: receiving advertisement content; The second processing module: classifies the advertisement content and identifies and classifies the advertisements into text advertisements, picture advertisements and video advertisements; The third processing module: reviewing the advertisement contents of different categories respectively; Wherein, when the picture advertisement is classified, a digital watermark is embedded in the picture advertisement, and the digital watermark includes category information of the picture advertisement.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images