Method and apparatus for optimizing advertisement based on cultural compliance, electronic device and medium

Abstract

The application relates to the technical field of advertisement optimization based on cultural compliance, and discloses an advertisement optimization method and device based on cultural compliance, an electronic device and a medium, wherein the method comprises the following steps: quantitatively analyzing by converting advertisement materials into multi-dimensional taboo vectors, and then automatically screening high-risk content based on an objective evaluation standard of a cultural conflict index, so as to avoid subjectivity and inefficiency of manual screening, and finally generating and executing a compliance alternative solution through a cross-modal adaptation engine. The application has the beneficial effect of realizing collaborative correction of graphic and text content, completely changing the backward operation mode of separately processing and manually adjusting graphic and text content in the original mode, significantly improving the accuracy of auditing, realizing full automation and high efficiency of the advertisement compliance review process, and greatly reducing the labor cost and compliance risk of enterprises.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of advertisement optimization based on cultural compliance, and particularly relates to an advertisement optimization method and device based on cultural compliance, an electronic device and a medium. BACKGROUND

[0002] The compliance review of advertisement delivery in specific cultural regions has long been a serious challenge. First, traditional review tools rely heavily on static keyword libraries and cannot effectively analyze complex cultural metaphors and social etiquette, resulting in a large blind spot in identifying deep cultural taboos and a high misjudgment rate. Second, existing technologies lack cross-modal collaboration capabilities, and image taboos and text taboos in advertisements need to be analyzed and manually adjusted in a fragmented manner, which is time-consuming and inefficient. SUMMARY

[0003] Therefore, it is necessary to propose an advertisement optimization method and device based on cultural compliance, an electronic device and a medium to solve the existing problems of advertisement optimization based on cultural compliance.

[0004] An advertisement optimization method based on cultural compliance, the method comprising:

[0005] obtaining initial advertisement materials to be reviewed and regional information of a target delivery area thereof;

[0006] extracting cultural taboo features from the initial advertisement materials to form a multi-dimensional taboo vector;

[0007] identifying the multi-dimensional taboo vector based on a preset cultural dynamic rule library to obtain a plurality of benchmark cultural taboo contents and a cultural conflict index thereof;

[0008] marking the benchmark cultural taboo contents with the cultural conflict index exceeding a preset threshold as target cultural taboo contents;

[0009] calling a cross-modal adaptation engine to generate a compliance alternative solution for the target cultural taboo contents;

[0010] replacing the corresponding target cultural taboo contents in the initial advertisement materials with the compliance alternative solution to obtain an optimized target advertisement material.

[0011] Further, the step of calling the cross-modal adaptation engine to generate a compliance alternative solution for the target cultural taboo contents comprises:

[0012] generating a plurality of candidate alternative solutions for the target cultural taboo contents in a cross-modal semantic space based on a contrastive learning algorithm;

[0013] obtaining a subdivided cultural rule corresponding to the regional information;

[0014] Based on the detailed cultural rules, a regional suitability score is calculated for each of the candidate alternatives;

[0015] Based on the regional suitability score, the optimal compliant alternative is dynamically selected from the multiple candidate alternatives.

[0016] Furthermore, the step of obtaining the subdivided cultural rules corresponding to the regional information includes:

[0017] Collect comment text data from social media platforms in the target delivery area within a preset time window;

[0018] Based on a pre-defined cultural dimension dictionary, the comment words in the comment text data are divided into corresponding cultural dimensions;

[0019] The number of comments under each cultural dimension is counted, and the fluctuation value of the number of comments under each cultural dimension within the preset time window is calculated.

[0020] Based on the number of comments and the fluctuation value of the number of comments, a corresponding weight value is set for each cultural dimension to form a dynamically updated subdivided cultural rule for calculating the regional suitability score.

[0021] Furthermore, after the step of replacing the corresponding target cultural taboo content in the initial advertising creative with the compliant alternative to obtain the optimized target advertising creative, the method further includes:

[0022] The target advertising creative is input into a pre-built digital twin pre-screening system; wherein the digital twin pre-screening system simulates the target user group in the target advertising area;

[0023] Obtain simulated emotional feedback data output by the digital twin pre-screening system for the target advertising material;

[0024] Based on the simulated emotional feedback data, a compliance risk assessment result for the target advertising creative is generated;

[0025] Determine whether the compliance risk assessment results have achieved the expected assessment results;

[0026] If the expected evaluation results are achieved, the target advertising material will be delivered to the target delivery area.

[0027] Furthermore, after the step of placing the target advertising material into the target placement area if the expected evaluation result is achieved, the method further includes:

[0028] The system continuously scans comment data from social media platforms corresponding to the target advertising area using a pre-defined multilingual BERT model.

[0029] Monitor the frequency of occurrence of preset cultural taboo keywords from the comment data;

[0030] Monitor whether the growth rate of the occurrence frequency per unit time is greater than a preset growth rate;

[0031] When the rate of increase of the frequency of occurrence per unit time is found to be greater than the preset rate of increase, an online early warning mechanism is triggered.

[0032] Furthermore, the step of identifying the multidimensional taboo vector based on a preset cultural dynamic rule base to obtain multiple benchmark cultural taboo contents and their cultural conflict indices includes:

[0033] From the cultural dynamic rule base, multiple cultural review rules associated with the target delivery area information are retrieved; each cultural review rule corresponds to a baseline cultural taboo content and a corresponding basic severity score.

[0034] The dimension value in the multidimensional taboo vector corresponding to each of the benchmark cultural taboo contents is used as the confidence score for the existence of the benchmark cultural taboo content.

[0035] The initial conflict score is obtained by multiplying the confidence score by the corresponding baseline severity score.

[0036] Based on the target delivery area information, each triggered cultural review rule is assigned a regional weight coefficient, and the initial conflict score is multiplied by the regional weight coefficient to obtain the adjusted target conflict score.

[0037] The system detects whether there are cross-modal collaboration violations in the initial advertising materials and determines the modal superposition coefficient based on the detection results.

[0038] The target conflict scores are aggregated, and the aggregation result is multiplied by the modal superposition coefficient to obtain the cultural conflict index of the initial advertising material.

[0039] Furthermore, before the step of identifying the multidimensional taboo vector based on a preset cultural dynamic rule base to obtain multiple benchmark cultural taboo contents and their cultural conflict indices, the method further includes:

[0040] Retrieve revised regulatory data from the target deployment area at preset intervals;

[0041] The legal data was analyzed using a natural language processing model to identify newly added, modified, or repealed cultural taboo clauses.

[0042] The identified taboo clauses are transformed into structured, machine-readable rules, and version tags and effective dates are added to them;

[0043] Based on the version tag and effective time, the machine-readable rules are incorporated into the cultural dynamic rule base in an incremental update manner.

[0044] An advertising optimization device based on cultural compliance, the device comprising:

[0045] The acquisition module is used to acquire the initial advertising materials to be reviewed and their target delivery area information;

[0046] The extraction module is used to extract cultural taboo features from the initial advertising material to form a multi-dimensional taboo vector;

[0047] The identification module is used to identify the multidimensional taboo vector based on a preset cultural dynamic rule base, and obtain multiple benchmark cultural taboo contents and their cultural conflict indices;

[0048] The marking module is used to mark the benchmark cultural taboo content that exceeds the preset threshold as the target cultural taboo content;

[0049] The calling module is used to invoke the cross-modal adaptation engine to generate compliant alternatives for the target culturally taboo content;

[0050] The replacement module is used to replace the target cultural taboo content in the initial advertising material based on the compliant alternative solution, so as to obtain the optimized target advertising material.

[0051] An electronic device includes a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the following steps:

[0052] Obtain the initial advertising creatives to be reviewed and their target delivery area information;

[0053] Cultural taboo features are extracted from the initial advertising materials to form a multidimensional taboo vector;

[0054] Based on a preset cultural dynamic rule base, the multidimensional taboo vector is identified to obtain multiple benchmark cultural taboo contents and their cultural conflict indices;

[0055] The benchmark cultural taboo content that exceeds the preset threshold in the cultural conflict index is recorded as the target cultural taboo content.

[0056] The cross-modal adaptation engine is invoked to generate compliant alternatives for the target culturally taboo content;

[0057] The target cultural taboo content in the initial ad creative is replaced with the compliant alternative to obtain the optimized target ad creative.

[0058] A computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the following steps:

[0059] Obtain the initial advertising creatives to be reviewed and their target delivery area information;

[0060] Cultural taboo features are extracted from the initial advertising materials to form a multidimensional taboo vector;

[0061] Based on a preset cultural dynamic rule base, the multidimensional taboo vector is identified to obtain multiple benchmark cultural taboo contents and their cultural conflict indices;

[0062] The benchmark cultural taboo content that exceeds the preset threshold in the cultural conflict index is recorded as the target cultural taboo content.

[0063] The cross-modal adaptation engine is invoked to generate compliant alternatives for the target culturally taboo content;

[0064] The target cultural taboo content in the initial ad creative is replaced with the compliant alternative to obtain the optimized target ad creative.

[0065] The beneficial effects of this invention are as follows: By converting advertising materials into multi-dimensional taboo vectors for quantitative analysis, it overcomes the shortcomings of traditional static keyword libraries in analyzing cultural metaphors, achieving accurate identification of complex cultural taboos. Then, based on the objective evaluation standard of cultural conflict index, it can automatically filter out high-risk content, avoiding the subjectivity and inefficiency of manual screening. Finally, through the generation and execution of compliant alternatives by a cross-modal adaptation engine, it achieves collaborative correction of text and image content, completely changing the outdated operation method of requiring separate processing and manual adjustment of text and image content in the original model, significantly improving the accuracy of review, realizing the full automation and high efficiency of the advertising compliance review process, and greatly reducing the company's labor costs and compliance risks.

[0066] By converting advertising materials into multi-dimensional taboo vectors for quantitative analysis, this approach overcomes the limitations of traditional static keyword libraries in resolving complex cultural metaphors, achieving accurate identification of cultural taboos. Secondly, based on an objective evaluation standard of cultural conflict index, it can automatically filter high-risk content, avoiding the subjectivity and inefficiency of manual screening. Most importantly, it utilizes a cross-modal adaptation engine to achieve collaborative correction of text and image content. This technology allows visual elements in advertisements to be correlated with text content. Through contrastive learning algorithms, it can identify similarities and relationships between the two in a cross-modal semantic space, thus providing compliant alternatives for content that triggers cultural conflict. Specifically, this process comprehensively analyzes the characteristics of visual content and text descriptions, ensuring that the generated compliant solutions retain the core message of the advertisement in a traditional sense while eliminating potential cultural sensitivities. When the text mentions sensitive social etiquette, the accompanying image will also be matched with an appropriate expression, generating more localized advertising materials. Through this automatic collaborative correction, advertising content can ensure marketing effectiveness while conforming to local culture, effectively avoiding audience misunderstandings or cultural conflicts caused by content mismatch. Furthermore, it overcomes the technical bias of separating text and images in previous advertising production. Traditional methods usually review text and images separately, which cannot effectively handle the semantic relationship between them. This may lead to cultural taboos being ignored in a certain modality. Through the aforementioned collaborative correction method, this invention essentially achieves unified review and intelligent replacement of text and image content in terms of compliance, ensuring the consistency and adaptability of advertising content on different media, and ultimately improving the overall quality and popularity of advertising. Attached Figure Description

[0067] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0068] in:

[0069] Figure 1 This is a diagram illustrating the application environment of an advertising optimization method based on cultural compliance in one embodiment.

[0070] Figure 2 This is a flowchart of an advertising optimization method based on cultural compliance in one embodiment;

[0071] Figure 3 This is a structural block diagram of an advertising optimization device based on cultural compliance in one embodiment;

[0072] Figure 4This is a structural block diagram of an electronic device in one embodiment. Detailed Implementation

[0073] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0074] Figure 1 This is a diagram illustrating an advertising optimization application environment based on cultural compliance in one embodiment. (Refer to...) Figure 1 This cultural compliance-based advertising optimization method is applied to a cultural compliance-based advertising optimization system. The system includes a terminal 110 and a server 120. The terminal 110 and server 120 are connected via a network. The terminal 110 can be a desktop terminal or a mobile terminal, specifically a mobile phone, tablet, laptop, or other similar device. The server 120 can be a standalone server or a server cluster consisting of multiple servers. The terminal 110 is used to acquire advertising materials to be reviewed, and the server 120 is used to optimize these materials.

[0075] like Figure 2 As shown, in one embodiment, a cultural compliance-based advertising optimization method is provided. This method can be applied to both terminals and servers; this embodiment illustrates its application to terminals. The cultural compliance-based advertising optimization method specifically includes the following steps:

[0076] S1: Obtain the initial advertising creatives to be reviewed and their target delivery area information;

[0077] S2: Extract cultural taboo features from the initial advertising material to form a multi-dimensional taboo vector;

[0078] S3: Based on a preset cultural dynamic rule base, the multidimensional taboo vector is identified to obtain multiple benchmark cultural taboo contents and their cultural conflict indices;

[0079] S4: Record the benchmark cultural taboo content whose cultural conflict index exceeds a preset threshold as the target cultural taboo content;

[0080] S5: Invoke the cross-modal adaptation engine to generate compliant alternatives for the target culturally taboo content;

[0081] S6: Replace the target cultural taboo content in the initial ad creative with the compliant alternative to obtain the optimized target ad creative.

[0082] As described in step S1 above, obtain the initial advertising materials to be reviewed and their target delivery area information. The area information includes the advertisement's geographical location, cultural background, language, and related social customs. The initial advertising materials are either uploaded by relevant personnel or generated automatically by a designated tool.

[0083] As described in step S2 above, cultural taboo features are extracted from the initial advertising material to form a multi-dimensional taboo vector. This involves converting various features of the advertising material (such as keywords in the text and visual elements in the image) into a multi-dimensional taboo vector. Specifically, natural language processing and computer vision technologies can be used to identify potential cultural taboos in the advertisement, such as specific colors, postures, or word choices. In the resulting multi-dimensional taboo vector, each dimension corresponds to a specific cultural taboo, and the value of the vector represents the potential risk or severity of the corresponding taboo. Multidimensional taboo vectors are created by explicitly classifying potentially offensive cultural taboos in advertising materials and using specific algorithms to encode information such as photos and text into a multidimensional array for subsequent analysis. Dimensions can include cultural symbols, colors, gestures, textual expressions, and image content. Extraction methods can utilize deep learning algorithms, natural language processing models, and fusion learning. Specifically, to obtain a complete multidimensional taboo vector, image and text features are fused and analyzed simultaneously. Features of text and images can be integrated into the final multidimensional taboo vector through feature fusion techniques (such as weighting and splicing), ensuring logical consistency and semantic complementarity between the text and images.

[0084] As described in step S3 above, the multidimensional taboo vector is identified based on a preset cultural dynamic rule base to obtain multiple benchmark cultural taboo contents and their cultural conflict indices. The cultural dynamic rule base is a knowledge base designed for specific cultures and regions, containing various cultural taboos, social etiquette, and applicable rules. This base is constantly updated to reflect new taboos arising from policy changes or evolving social customs. Each dimension value in the multidimensional taboo vector is compared with the benchmark contents in the rule base to derive each benchmark cultural taboo content and its corresponding cultural conflict index. The conflict index is a quantitative indicator reflecting the degree of conflict between a portion of the advertising material and local cultural regulations. The Cultural Conflict Index (CCI) is used to quantify the compatibility between advertising content and the culture of the target region. The specific formula is as follows: ,in, Here, n represents the modal superposition coefficient, and n represents the total number of cultural taboo content items triggered. This represents the regional weight coefficient corresponding to the i-th cultural taboo. This represents the confidence score corresponding to the i-th taboo item in the multidimensional taboo vector. This represents the base severity score of the i-th taboo content (based on cultural censorship rules).

[0085] As described in step S4 above, the baseline cultural taboo content whose cultural conflict index exceeds a preset threshold is recorded as target cultural taboo content. Any cultural taboo content whose cultural conflict index exceeds the set preset threshold can be marked as target cultural taboo content. The preset threshold is set based on historical data, cultural background, and industry standards to ensure that potential compliance risks can be identified; for example, it is set to 0.7. This ensures that subsequent operations focus only on taboo content that may cause compliance issues, thereby optimizing the processing flow and reducing unnecessary manual intervention. By marking target cultural taboo content, compliance corrections can be prioritized in subsequent steps, reducing unexpected legal liabilities and business risks.

[0086] As described in step S5 above, the cross-modal adaptation engine is invoked to generate compliant alternatives for the target culturally taboo content. Multiple compliant alternatives are generated for the tagged target culturally taboo content. The cross-modal adaptation engine uses deep learning algorithms, combined with contrastive learning and semantic mapping techniques, to ensure coordination and adaptation between text and visual content. The generated alternatives may include conversions of textual expressions, replacement of image content, or reasonable modifications to other elements. This process not only aims to eliminate potential cultural taboos but also successfully achieves compliance while maintaining the original intent and appeal of the advertisement.

[0087] As described in step S6 above, the target cultural taboo content in the initial advertising creative is replaced with the compliant alternative to obtain the optimized target advertising creative. The initial advertising creative is then processed using the previously generated compliant alternative to replace the target cultural taboo content with newly generated compliant content, resulting in an optimized target advertising creative. This optimized advertisement adheres to local cultural rules while attracting the target audience, ensuring an optimal balance between legal compliance and marketing effectiveness.

[0088] By converting advertising materials into multi-dimensional taboo vectors for quantitative analysis, this approach overcomes the limitations of traditional static keyword libraries in resolving complex cultural metaphors, achieving accurate identification of cultural taboos. Secondly, based on an objective evaluation standard of cultural conflict index, it can automatically filter high-risk content, avoiding the subjectivity and inefficiency of manual screening. Most importantly, it utilizes a cross-modal adaptation engine to achieve collaborative correction of text and image content. This technology allows visual elements in advertisements to be correlated with text content. Through contrastive learning algorithms, it can identify similarities and relationships between the two in a cross-modal semantic space, thus providing compliant alternatives for content that triggers cultural conflict. Specifically, this process comprehensively analyzes the characteristics of visual content and text descriptions, ensuring that the generated compliant solutions retain the core message of the advertisement in a traditional sense while eliminating potential cultural sensitivities. When the text mentions sensitive social etiquette, the accompanying image will also be matched with an appropriate expression, generating more localized advertising materials. Through this automatic collaborative correction, advertising content can ensure marketing effectiveness while conforming to local culture, effectively avoiding audience misunderstandings or cultural conflicts caused by content mismatch.

[0089] Furthermore, it overcomes the technical bias of separating text and images in previous advertising production. Traditional methods usually review text and images separately, which cannot effectively handle the semantic relationship between them. This may lead to cultural taboos being ignored in a certain modality. Through the aforementioned collaborative correction method, this invention essentially achieves unified review and intelligent replacement of text and image content in terms of compliance, ensuring the consistency and adaptability of advertising content on different media, and ultimately improving the overall quality and popularity of advertising.

[0090] In one embodiment, step S5, which involves invoking a cross-modal adaptation engine to generate compliant alternatives for the target culturally taboo content, includes:

[0091] S501: Based on the contrastive learning algorithm, generate multiple candidate alternatives for the target cultural taboo content in the cross-modal semantic space;

[0092] S502: Obtain the subdivided cultural rules corresponding to the regional information;

[0093] S503: Calculate a regional suitability score for each candidate alternative based on the subdivided cultural rules;

[0094] S504: Based on the regional adaptability score, dynamically select the optimal compliant alternative from the multiple candidate alternatives.

[0095] As described in step S501 above, based on the contrastive learning algorithm, multiple candidate alternatives are generated for the target cultural taboo content in the cross-modal semantic space. Contrastive learning is an unsupervised learning method that improves the model's representational ability by constructing positive and negative sample pairs. In this context, the expression of the target cultural taboo content (e.g., textual descriptions or image elements) will be used as positive samples, while various suitable alternatives will be used as potential negative samples for training. Through training, the system can understand and map the similarities and differences between different modalities (e.g., text and images), thereby finding alternatives in the cross-modal semantic space that are semantically similar to the taboo content but culturally compliant. This process not only increases the diversity and effectiveness of candidate solutions but also takes into account the cultural characteristics of the target market, providing richer input for subsequent evaluation and selection. Constructing a cross-modal semantic space requires generating vector representations for both text and image content: Text modality: Natural language processing models (such as BERT and GPT) are used to convert advertising copy into context-aware text vectors. This process utilizes techniques like Word2Vec or TF-IDF to capture relationships between words and generate a description of the overall text. Image modality: Deep learning models such as Convolutional Neural Networks (CNNs) are used to analyze advertising images and convert them into visual feature vectors. This captures the content of the image, including objects, colors, and composition. After these processes, the text and image vectors can interact and be compared within the same semantic space—the cross-modal semantic space.

[0096] As described in step S502 above, the segmented cultural rules corresponding to the regional information are obtained. Segmented cultural rules refer to advertising compliance operating procedures formulated for specific regions, cultures, and social customs. These rules are usually derived from local laws, culture, and social expectations. In this process, a dynamic cultural rule base can be invoked to automatically retrieve and analyze relevant laws and social customs. By associating these cultural rules with candidate alternatives, it is possible to identify which alternatives conform to local cultural habits, thereby enhancing the localization adaptability of advertising.

[0097] As described in step S503 above, a regional suitability score is calculated for each candidate alternative based on the subdivided cultural rules. The regional suitability score is a quantitative indicator designed to measure the degree of fit between each candidate alternative and the cultural rules of the target deployment area. During this process, a comprehensive evaluation can be conducted based on multiple factors, including the fit between the candidate alternative and the subdivided cultural rules, potential cultural conflicts, and user feedback. The scoring can employ a weighted algorithm, assigning appropriate weights based on the importance of each cultural rule in a specific region, ultimately calculating a suitability value. This optimized evaluation can scientifically and reasonably determine the acceptability of candidate alternatives, laying a numerical foundation for subsequently selecting the optimal solution.

[0098] As described in step S504 above, based on the regional suitability score, the optimal compliant alternative is dynamically selected from the multiple candidate alternatives. Utilizing optimization methods from decision theory, the option with the highest suitability is sought among all feasible candidate options. By using set threshold standards, it is ensured that the selected solution not only meets cultural compliance requirements but also effectively conveys brand information and market positioning. The selection criteria can be updated and adjusted in real time; for example, after obtaining new cultural information or user feedback, the solution selection is re-evaluated and adjusted. This dynamic selection process can significantly improve the compliance and market adaptability of advertising materials, and enhance the overall advertising effect and user experience by achieving personalized and localized content delivery.

[0099] In one embodiment, step S502 of obtaining the subdivided cultural rules corresponding to the regional information includes:

[0100] S5021: Collect comment text data within a preset time window from the social media platforms in the target delivery area;

[0101] S5022: Based on a preset cultural dimension dictionary, classify the comment words in the comment text data into the corresponding cultural dimensions;

[0102] S5023: Count the number of comments under each cultural dimension, and calculate the fluctuation value of the number of comments under each cultural dimension within the preset time window;

[0103] S5024: Based on the number of comments and the fluctuation value of the number of comments, set a corresponding weight value for each cultural dimension to form a dynamically updated subdivided cultural rule for calculating the regional suitability score.

[0104] As described in step S5021 above, comment text data within a preset time window is collected from social media platforms in the target delivery area. First, the data source is determined (e.g., commonly used social media platforms, forums, and comment sections in the region), and a time window is set (e.g., the past 7 days or 30 days). Comment text and its metadata (timestamps, language identifiers, geographic tags, user basic attributes, etc.) from social media platforms are collected through legitimate APIs, web crawlers, or third-party data interfaces. The collected data undergoes cleaning operations, including: deduplication, noise reduction (removing advertisements or spam), language detection and encoding standardization, processing of emoticons / special symbols, and standardization of unstructured text. Dialect recognition and transcription are performed when necessary. Simultaneously, account credibility filtering (removing suspected bots or high-frequency posting accounts) and sampling strategies are implemented (ensuring representativeness across different regions / groups, outputting a structured comment text set, including fields such as: {Comment ID, Text, Time, Language, Geographic Range, User Credibility Tag}), for subsequent dimensional segmentation.

[0105] As described in step S5022 above, based on a preset cultural dimension dictionary, the comment words in the comment text data are divided into corresponding cultural dimensions. Using the "cultural dimension dictionary" as the core, natural language comments are mapped to predefined cultural dimensions (such as clothing, gender roles, food, gestures, etiquette, etc.). The specific process includes: segmenting / lexicalization, lemmatization, and stop word filtering for each comment; performing named entity and phrase identification (NER) to capture proper nouns or idioms; and then mapping words or phrases to one or more cultural dimension labels based on the dictionary and synonym expansion (using a thesaurus or embedding similarity thresholds). To improve recall, fuzzy matching and semantic matching (based on pre-trained word vectors or lightweight semantic similarity models) can be used to handle metaphors, slang, or variant expressions. The output is an annotated corpus with dimension labels. Each record contains: the original text, the matching word, the mapped cultural dimension, and the matching confidence, for subsequent statistics and weight calculation.

[0106] As described in step S5023 above, the number of comments under each cultural dimension is counted, and the fluctuation value of the number of comments for each cultural dimension within the preset time window is calculated. The number of comments is summarized by cultural dimension (can be divided into buckets by day or hour) to obtain a time series. In order to measure the activity and trend of the dimension within the time window, the fluctuation value needs to be calculated. Commonly used indicators include the standard deviation, variance, or relative volatility (e.g., the ratio of standard deviation to mean, CV). Short-term moving average difference or exponentially weighted moving average (EWMA) can also be used to capture sudden increases. If it is necessary to identify periodic or sudden events, the peak ratio (highest value / mean) can be further calculated or time series anomaly detection can be performed. During the statistical process, the impact of time window length and resolution on volatility should be considered, and extreme values ​​should be smoothed (e.g., truncated). The final output is a statistical vector corresponding to each cultural dimension: {total number of comments, time distribution series, mean, standard deviation, relative volatility, peak index}, to support subsequent weight calculation.

[0107] As described in step S5024 above, based on the number of comments and the fluctuation value of the number of comments, a corresponding weight value is set for each cultural dimension to form a dynamically updated segmented cultural rule for calculating the regional suitability score. The number of comments and the fluctuation value are normalized (e.g., minimum-maximum scaling or Z-score standardization) to eliminate differences in magnitude; then, a synthesis formula is set in conjunction with the business strategy, for example, weight = α·norm(number of comments) + β·norm(fluctuation value), where α and β are configurable coefficients, and α + β = 1, reflecting the relative importance of "general attention" and "recent sensitivity", for example, α = 0.6, β = 0.4. To prevent frequent fluctuations caused by noise, smoothing (such as moving average) and threshold limits (minimum / maximum weight) can be applied to the weights, and time decay factors or adaptive learning coefficients can be introduced to achieve incremental updates. Finally, the weights of each cultural dimension are written into the subdivided cultural rule entries to form a formal rule set (including weights, update timestamps, and source window descriptions). This rule set can be updated periodically and version-managed for use by the regional adaptability scoring module, thereby guiding priority and sorting with dynamic, data-driven weights when selecting replacement schemes.

[0108] In one embodiment, after step S6, which involves replacing the corresponding target cultural taboo content in the initial ad creative based on the compliance alternative to obtain the optimized target ad creative, the method further includes:

[0109] S701: Input the target advertising material into a pre-built digital twin pre-screening system; wherein, the digital twin pre-screening system simulates the target user group in the target delivery area;

[0110] S702: Obtain the simulated emotional feedback data output by the digital twin pre-screening system for the target advertising material;

[0111] S703: Based on the simulated emotional feedback data, generate a compliance risk assessment result for the target advertising material;

[0112] S704: Determine whether the compliance risk assessment results have achieved the expected assessment results;

[0113] S705: If the expected evaluation results are achieved, the target advertising material will be delivered to the target delivery area.

[0114] As described in step S701 above, the target advertising material is input into a pre-built digital twin pre-screening system. Digital twin technology is a technique that simulates real-world processes virtually to simulate the characteristics of user groups within a specific target advertising area. The construction of the digital twin pre-screening system involves several key components. First, it is necessary to collect and integrate multi-dimensional data of the target user group, including demographic information (age, gender, geographical location, etc.), behavioral data (social media activity, purchase records, etc.), and cultural background information. By analyzing historical data, a representative user profile is constructed to ensure that the system can simulate the characteristics of real users. Then, machine learning and data analysis techniques are used to establish a predictive model of user behavior. This model can predict the target user's reaction to different advertising materials based on the user profile data. This includes sentiment analysis, user interest modeling, etc., to ensure the accuracy of the simulation. To enhance the system's flexibility, the digital twin system should be able to receive feedback data from social media or market reactions in real time. This mechanism allows the system to continuously update and optimize the user profile and its corresponding model, making the pre-screening results more timely.

[0115] During the simulation, target advertising creatives to be reviewed are input into the system, including text, images, or video content. The system uses established user models to simulate the target advertisement, predicting the target users' emotional responses, interest engagement, and cultural adaptability. This process assesses the appeal and compliance of the advertising copy by calculating users' emotional ratings of the ad content. Based on simulated user responses, the system generates emotional feedback data, which is used to evaluate whether the ad content meets expected cultural standards and market acceptance. The system may output multiple emotional feedback dimensions, such as satisfaction, acceptance, and potential complaint risk. Combining the simulated feedback, the system will centrally assess the compliance risks of the ad content. This assessment gives advertising teams the opportunity to adjust and optimize ad design before actual launch to eliminate potential cultural conflicts.

[0116] As described in step S702 above, simulated emotional feedback data output by the digital twin pre-screening system for the target advertising material is obtained. This emotional feedback data is derived from simulated user group reactions and reflects users' emotional attitudes towards the target advertising material. The data includes emotional inclination (e.g., positive, negative, or neutral) and emotional intensity (e.g., rating or weight). This feedback stems from the user's cognition and perception of the advertising material. Machine learning and natural language processing algorithms are typically used to analyze the language, visual presentation, and emotional resonance of the advertisement towards users. Combined with deep learning of users' historical behavior, specific feedback results are generated. Obtaining this emotional feedback data helps advertisers assess the potential impact and acceptance of the advertisement within a specific cultural context, thereby further optimizing advertising content and communication strategies.

[0117] As described in step S703 above, a compliance risk assessment result for the target advertising material is generated based on the simulated emotional feedback data. The assessment process, based on a pre-defined risk assessment model and formula, transforms user emotional responses into compliance scores. This may include quantitative analysis of emotional feedback (such as satisfaction ratings and sentiment analysis), statistical analysis (such as frequency analysis and distribution assessment), and aggregation and visualization of user responses. Potential problems with the advertising content are identified by constructing decision trees or logistic regression models, and historical data analysis is used to predict the likelihood of compliance risks. Furthermore, the impact of background culture on advertising acceptance can be analyzed to facilitate comparison of the risk assessment results with cultural adaptability.

[0118] As described in step S704 above, it is determined whether the compliance risk assessment result has met the expected assessment result. The expected assessment result is a standard formulated based on advertising goals and market demand. It may include compliance score, user emotional satisfaction threshold, etc. This judgment process usually combines cluster analysis and decision tree algorithm to compare and classify the assessment results to determine whether the advertising material is suitable for market launch. If the assessment result meets the expected standard, the advertising material will enter the launch stage; if it does not meet the standard, a readjustment process is triggered, that is, returning to the optimization stage or content adjustment stage.

[0119] As described in step S705 above, if the expected evaluation result is achieved, the target advertising material will be placed in the target placement area. The placement process includes selecting a suitable placement platform, determining the placement time period, allocating the budget, and conducting media purchases. After confirming that the advertising material is compliant with the local cultural environment and can respond relatively positively to the emotional feedback of the target audience, the final release will be carried out. The execution process may also include a real-time monitoring mechanism for advertising effectiveness, so as to adjust the strategy and content in a timely manner based on early feedback after the advertising is placed. If the expected evaluation result is not achieved, necessary modifications or optimizations will be recommended to ensure that the advertisement can achieve the best communication effect before officially entering the market.

[0120] In one embodiment, after step S705 of placing the target advertising material into the target placement area if the expected evaluation result is achieved, the method further includes:

[0121] S7061: Continuously scan the comment data of social media corresponding to the target delivery area using a preset multilingual BERT model;

[0122] S7062: Monitor the frequency of occurrence of preset cultural taboo keywords from the comment data;

[0123] S7063: Monitor whether the growth rate of the occurrence frequency per unit time is greater than the preset growth rate;

[0124] S7064: When the growth rate of the occurrence frequency is detected to be greater than the preset growth rate within a unit time, an online early warning mechanism is triggered.

[0125] As described in step S7061 above, the system continuously scans comment data on social media platforms corresponding to the target advertising area using a preset multilingual BERT model. Feedback from users on the advertised content is obtained from multiple social platforms to promptly grasp public sentiment. The multilingual BERT model is an advanced natural language processing tool capable of processing text data in multiple languages ​​and performing semantic understanding. The system then sets the social media platforms to be monitored and automatically collects user comments and interaction data related to the advertisement using APIs or web crawling techniques. The specific scanning frequency and scope can be adjusted based on real-time requirements and resource constraints.

[0126] As described in step S7062 above, the frequency of occurrence of preset cultural taboo keywords is monitored from the comment data. Keywords are sensitive words related to local culture, involving aspects such as culture, gender, customs, and etiquette. By relying on a pre-built database of cultural taboo keywords, text matching is performed, and natural language processing techniques such as word segmentation, part-of-speech tagging, and keyword extraction are used to identify taboo words in the comments, thereby calculating their frequency in the total comments.

[0127] As described in step S7063 above, monitor whether the growth rate of the occurrence frequency within a unit of time is greater than a preset growth rate. Establish a baseline and detect changes in keyword frequency within the current time period using historical data. If the frequency of a certain keyword increases beyond a preset growth rate threshold, it indicates a sudden increase in discussions related to that keyword. This may mean fluctuations in comments regarding certain culturally taboo content. The growth rate is typically calculated using time series analysis methods, such as simple ratio calculations or more complex moving average methods, to ensure the smoothness and accuracy of the detection.

[0128] As described in step S7064 above, when the growth rate of the frequency of occurrence within a unit of time is found to be greater than the preset growth rate, an online early warning mechanism is triggered. The online early warning mechanism aims to react quickly and promptly inform the advertising team, alerting them to potential crises caused by the current advertising content. The content and methods of the early warning mechanism can be diverse, such as automatically pushing alerts, generating reports, sending email notifications, or pushing messages through applications. Through this mechanism, businesses can immediately anticipate increased risks and take measures such as reassessing advertising materials, changing advertising strategies, and adjusting keyword usage.

[0129] In one embodiment, step S3, which involves identifying the multidimensional taboo vector based on a preset cultural dynamic rule base to obtain multiple benchmark cultural taboo contents and their cultural conflict indices, includes:

[0130] S301: From the cultural dynamic rule base, call up multiple cultural review rules associated with the target delivery area information; wherein, each cultural review rule corresponds to a baseline cultural taboo content and a corresponding basic severity score;

[0131] S302: The dimension value in the multidimensional taboo vector corresponding to each of the benchmark cultural taboo contents is used as the confidence score of the existence of the benchmark cultural taboo content;

[0132] S303: Multiply the confidence score by the corresponding baseline severity score to obtain the initial conflict score;

[0133] S304: Based on the target delivery area information, assign a regional weight coefficient to each triggered cultural review rule, and multiply the initial conflict score by the regional weight coefficient to obtain the adjusted target conflict score;

[0134] S305: Detect whether there is a cross-modal collaboration violation in the initial advertising material, and determine the modal superposition coefficient based on the detection result;

[0135] S306: Aggregate the target conflict scores and multiply the aggregation result by the modal superposition coefficient to obtain the cultural conflict index of the initial advertising material.

[0136] As described in step S301 above, multiple cultural review rules associated with the target deployment area information are retrieved from the cultural dynamic rule base. Each cultural review rule corresponds to a baseline cultural taboo content and a corresponding basic severity score. Cultural review rules include descriptions and norms of local culture, customs, and other characteristics, relying on predefined correlations. The cultural review rule base needs to be dynamically updated to reflect policy changes and changes in social customs. Each cultural review rule corresponds to a baseline cultural taboo content and includes a basic severity score, indicating the importance of the taboo in the culture or the potential impact of its violation. This step ensures that cultural compliance analysis has a detailed and authoritative rule foundation, thereby providing accurate reference for subsequent taboo content identification.

[0137] As described in step S302 above, the dimension values ​​in the multidimensional taboo vector corresponding to each of the benchmark cultural taboo contents are used as the confidence score for the existence of that benchmark cultural taboo content. Each dimension of the multidimensional taboo vector represents a potential cultural taboo, and the magnitude of the value indicates the likelihood of that taboo feature existing in the advertising material. Through this mechanism, the risk level of each taboo content can be automatically quantified. This confidence score effectively identifies which cultural taboos are more likely to cause conflict in the advertising content and focuses the evaluation on these high-risk areas, thereby improving the effectiveness of the final review.

[0138] As described in step S303 above, the confidence score is multiplied by the corresponding baseline severity score to obtain the initial conflict score. The baseline severity score reflects the potential impact of cultural taboos in the advertising material, while the confidence score quantifies the likelihood of these taboos existing in the actual material. By multiplying these two values, the final result provides a comprehensive indicator for assessing the risk of cultural conflict that the advertisement may trigger.

[0139] As described in step S304 above, based on the target advertising region information, each triggered cultural censorship rule is assigned a regional weight coefficient, and the initial conflict score is multiplied by the regional weight coefficient to obtain the adjusted target conflict score. The weight coefficient is set according to specific cultural backgrounds, social customs, and the laws and regulations of the advertising region, aiming to reflect the differences in the degree of importance different regions place on the same cultural taboo. This dynamic weighting mechanism ensures a close integration of norms with geographical and cultural relevance. Multiplying the initial conflict score by the corresponding regional weight coefficient yields the adjusted target conflict score, which reflects the cultural tolerance of the target region and the urgency of potential taboos in the advertising content.

[0140] As described in step S305 above, the system detects whether cross-modal collaborative violations exist in the initial advertising material and determines a modal superposition coefficient based on the detection results. Cross-modal collaborative violations refer to mutually reinforcing conflicts between the text and visual content in an advertisement regarding cultural taboos. For example, an image might display elements consistent with cultural taboos mentioned in the text, creating a double-stimulation risk of conflict. The detection process typically utilizes deep learning algorithms to analyze the content correlation between different modalities (such as text and images) and check for synergistic effects. At this stage, the system generates a modal superposition coefficient for the synergistic probability of specific taboo content. This coefficient quantifies the degree to which the risk increases if the multimodal content of the advertisement collectively constitutes a violation.

[0141] As described in step S306 above, the target conflict scores are aggregated, and the aggregation result is multiplied by the modal superposition coefficient to obtain the cultural conflict index of the initial advertising material. The aggregation process allows the risks of individual taboos to be summarized, forming a cultural conflict index (CCI) for the overall advertising material. The final aggregation result is multiplied by the previously determined modal superposition coefficient to comprehensively assess the likelihood of cultural conflict. In this way, the resulting cultural conflict index not only reflects the individual risks of the advertising content but also fully considers the combined risk effect caused by the synergy of different content categories. This index provides a quantitative basis for subsequent decision-making, ultimately ensuring the monitoring and adjustment process for cultural compliance and providing an effective basis and standard for advertising optimization and compliance review.

[0142] In one embodiment, before step S3, which involves identifying the multidimensional taboo vector based on a preset cultural dynamic rule base to obtain multiple benchmark cultural taboo contents and their cultural conflict indices, the method further includes:

[0143] S201: Crawl revised regulatory data from the target deployment area at a preset cycle;

[0144] S202: Use a natural language processing model to parse the regulatory data and identify any newly added, modified, or repealed cultural taboo clauses.

[0145] S203: Transform the identified taboo clauses into structured, machine-readable rules and add version tags and effective dates to them;

[0146] S204: Based on the version tag and effective time, the machine-readable rules are incrementally updated and incorporated into the cultural dynamic rule base.

[0147] As described in step S201 above, revised regulatory data is crawled from the target region at a preset cycle. The system proactively crawls relevant legal and regulatory sources in the target region at preset cycles (e.g., weekly, monthly) to ensure timely acquisition of the latest regulatory data within that region. Legal information sources may include websites, legal information databases, data released by regulatory agencies, or third-party websites related to law. This process typically relies on web crawling technology, which can automatically access these information sources and extract new or revised regulatory texts. It requires reasonable conditional judgment to identify which documents are new or revised, and ensures that the acquisition process adheres to website crawling protocols and relevant legal provisions to avoid legal issues. By regularly collecting this revised regulatory data, the system can ensure the authority and timeliness of the information sources, thereby guaranteeing that the information in the dynamic rule base is up-to-date and can promptly reflect changes in the target market.

[0148] As described in step S202 above, the legal data is parsed using a natural language processing model to identify newly added, modified, or repealed cultural taboo clauses. The legal text is segmented, tagged with parts of speech, and analyzed syntactically to ensure the model accurately understands the text structure and meaning. By comparing the differences between the new regulations and the old regulations in the database, change detection algorithms (such as text similarity calculation and difference analysis) are used to identify specific modifications. During the clause identification process, the model can identify keywords, phrases, and related context to determine their relevance to cultural taboos.

[0149] As described in step S203 above, the identified taboo clauses are transformed into structured, machine-readable rules, and version tags and effective dates are added to them. Converting text-based cultural taboo clauses into entries with a clear structure facilitates subsequent machine processing and retrieval. The system will need to define specific fields for each taboo clause, such as: clause content, category tag (e.g., social etiquette), and scope of application. During the transformation process, version tags and effective dates are also added to each rule to ensure traceability and timeliness. This facilitates version management when regulations are updated and helps determine when to apply specific clauses.

[0150] As described in step S204 above, based on the version tag and effective time, the machine-readable rules are incrementally updated and incorporated into the cultural dynamic rule base. Incremental updates mean that the system does not replace the entire rule base at once, but only adds the latest, relevant, and revised clauses to the database in a way that minimizes their impact. This avoids potential system stagnation and data conflicts. By comparing and analyzing the new clauses, the system can detect whether old rules conflict with or are replaced by new rules, and then take appropriate action (such as repealing old rules or merging similar rules). This update process ensures that the cultural dynamic rule base is always up-to-date and accurate, providing reliable data support for the subsequent identification and compliance assessment of culturally prohibited content.

[0151] Reference Figure 3 The present invention also provides an advertising optimization device based on cultural compliance, the device comprising:

[0152] The acquisition module 902 is used to acquire the initial advertising materials to be reviewed and the regional information of their target delivery areas;

[0153] The extraction module 904 is used to extract cultural taboo features from the initial advertising material to form a multi-dimensional taboo vector;

[0154] The identification module 906 is used to identify the multidimensional taboo vector based on a preset cultural dynamic rule base to obtain multiple benchmark cultural taboo contents and their cultural conflict index.

[0155] The marking module 908 is used to mark the benchmark cultural taboo content whose cultural conflict index exceeds a preset threshold as the target cultural taboo content.

[0156] Module 910 is invoked to call the cross-modal adaptation engine to generate compliant alternatives for the target culturally taboo content;

[0157] Replacement module 912 is used to replace the target cultural taboo content in the initial advertising material based on the compliance alternative, so as to obtain the optimized target advertising material.

[0158] In one embodiment, calling module 910 includes:

[0159] The candidate alternative generation submodule is used to generate multiple candidate alternatives for the target cultural taboo content in a cross-modal semantic space based on a contrastive learning algorithm.

[0160] The sub-module for obtaining detailed cultural rules is used to obtain detailed cultural rules corresponding to the regional information.

[0161] The regional suitability score calculation submodule is used to calculate a regional suitability score for each of the candidate alternatives based on the subdivided cultural rules.

[0162] The compliance alternative selection submodule is used to dynamically select the optimal compliance alternative from the multiple candidate alternatives based on the regional adaptability score.

[0163] In one embodiment, the sub-module for obtaining detailed cultural rules includes:

[0164] The data acquisition unit is used to collect comment text data from social media platforms in the target delivery area within a preset time window;

[0165] The segmentation unit is used to segment the comment words in the comment text data into the corresponding cultural dimensions based on a preset cultural dimension dictionary;

[0166] The statistics unit is used to count the number of comments under each cultural dimension and calculate the fluctuation value of the number of comments under each cultural dimension within the preset time window;

[0167] The setting unit is used to set a corresponding weight value for each cultural dimension based on the number of comments and the fluctuation value of the number of comments, so as to form a dynamically updated subdivided cultural rule for calculating the regional suitability score.

[0168] In one embodiment, the advertising optimization device based on cultural compliance further includes:

[0169] An input module is used to input the target advertising material into a pre-built digital twin pre-screening system; wherein the digital twin pre-screening system simulates the target user group in the target delivery area;

[0170] The simulated emotional feedback data acquisition module is used to acquire the simulated emotional feedback data output by the digital twin pre-screening system for the target advertising material;

[0171] The compliance risk assessment result generation module is used to generate the compliance risk assessment result of the target advertising material based on the simulated emotional feedback data.

[0172] The judgment module is used to determine whether the compliance risk assessment results have achieved the expected assessment results;

[0173] The delivery module is used to deliver the target advertising material to the target delivery area if the expected evaluation results are achieved.

[0174] In one embodiment, the advertising optimization device based on cultural compliance further includes:

[0175] The scanning module is used to continuously scan the comment data of social media corresponding to the target delivery area using a preset multilingual BERT model;

[0176] The first monitoring module is used to monitor the frequency of occurrence of preset cultural taboo keywords from the comment data;

[0177] The second monitoring module is used to monitor whether the growth rate of the occurrence frequency per unit time is greater than a preset growth rate.

[0178] The triggering module is used to trigger an online early warning mechanism when the growth rate of the frequency of occurrence within a unit time is greater than a preset growth rate.

[0179] In one embodiment, the identification module 906 includes:

[0180] The cultural censorship rule invocation submodule is used to invoke multiple cultural censorship rules associated with the target delivery area information from the cultural dynamic rule base; wherein, each cultural censorship rule corresponds to a baseline cultural taboo content and a corresponding basic severity score;

[0181] The confidence score acquisition submodule is used to take the dimension value in the multidimensional taboo vector that corresponds to each of the benchmark cultural taboo contents as the confidence score of the existence of the benchmark cultural taboo contents.

[0182] The initial conflict score acquisition submodule is used to multiply the confidence score by the corresponding baseline severity score to obtain the initial conflict score;

[0183] The target conflict score acquisition submodule is used to assign a regional weight coefficient to each triggered cultural review rule based on the target delivery area information, and multiply the initial conflict score by the regional weight coefficient to obtain the adjusted target conflict score.

[0184] The modal superposition coefficient determination submodule is used to detect whether there are cross-modal collaboration violations in the initial advertising material, and to determine the modal superposition coefficient based on the detection results;

[0185] The aggregation submodule is used to aggregate the target conflict scores and multiply the aggregation result by the modal superposition coefficient to obtain the cultural conflict index of the initial advertising material.

[0186] In one embodiment, the advertising optimization device based on cultural compliance further includes:

[0187] The crawling module is used to crawl revised regulatory data from the target deployment area at preset intervals;

[0188] The cultural taboo clause identification module is used to analyze the regulatory data using a natural language processing model to identify newly added, modified, or repealed cultural taboo clauses.

[0189] The conversion module is used to transform the identified taboo clauses into structured machine-readable rules and add version tags and effective dates to them;

[0190] The incorporation module is used to incorporate the machine-readable rules into the cultural dynamic rule base in an incremental update manner based on the version tag and effective time.

[0191] Figure 4 An internal structural diagram of an electronic device in one embodiment is shown. This electronic device can specifically be a terminal or a server, and more specifically, a computer device. Figure 4 As shown, the electronic device includes a processor, a memory, and a network interface connected via a system bus. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and may also store a computer program. When executed by the processor, this computer program enables the processor to implement a culture-compliant advertising optimization method. The internal memory may also store a computer program, which, when executed by the processor, enables the processor to implement a culture-compliant advertising optimization method. Those skilled in the art will understand that... Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the electronic device to which the present application is applied. The specific electronic device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0192] In one embodiment, an electronic device is provided, including a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the following steps:

[0193] Obtain the initial advertising creatives to be reviewed and their target delivery area information;

[0194] Cultural taboo features are extracted from the initial advertising materials to form a multidimensional taboo vector;

[0195] Based on a preset cultural dynamic rule base, the multidimensional taboo vector is identified to obtain multiple benchmark cultural taboo contents and their cultural conflict indices;

[0196] The benchmark cultural taboo content that exceeds the preset threshold in the cultural conflict index is recorded as the target cultural taboo content.

[0197] The cross-modal adaptation engine is invoked to generate compliant alternatives for the target culturally taboo content;

[0198] The target cultural taboo content in the initial ad creative is replaced with the compliant alternative to obtain the optimized target ad creative.

[0199] By converting advertising materials into multi-dimensional taboo vectors for quantitative analysis, the system overcomes the shortcomings of traditional static keyword libraries in analyzing cultural metaphors, achieving accurate identification of complex cultural taboos. Then, based on the objective evaluation standard of cultural conflict index, it can automatically filter out high-risk content, avoiding the subjectivity and inefficiency of manual screening. Finally, through a cross-modal adaptation engine, it generates and executes compliant alternatives, achieving collaborative correction of text and image content. This completely changes the outdated operating method of requiring separate processing and manual adjustment of text and image content, significantly improving the accuracy of review and realizing the full automation and high efficiency of the advertising compliance review process, greatly reducing the company's labor costs and compliance risks.

[0200] In one embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, causes the processor to perform the following steps:

[0201] Obtain the initial advertising creatives to be reviewed and their target delivery area information;

[0202] Cultural taboo features are extracted from the initial advertising materials to form a multidimensional taboo vector;

[0203] Based on a preset cultural dynamic rule base, the multidimensional taboo vector is identified to obtain multiple benchmark cultural taboo contents and their cultural conflict indices;

[0204] The benchmark cultural taboo content that exceeds the preset threshold in the cultural conflict index is recorded as the target cultural taboo content.

[0205] The cross-modal adaptation engine is invoked to generate compliant alternatives for the target culturally taboo content;

[0206] The target cultural taboo content in the initial ad creative is replaced with the compliant alternative to obtain the optimized target ad creative.

[0207] By converting advertising materials into multi-dimensional taboo vectors for quantitative analysis, the system overcomes the shortcomings of traditional static keyword libraries in analyzing cultural metaphors, achieving accurate identification of complex cultural taboos. Then, based on the objective evaluation standard of cultural conflict index, it can automatically filter out high-risk content, avoiding the subjectivity and inefficiency of manual screening. Finally, through a cross-modal adaptation engine, it generates and executes compliant alternatives, achieving collaborative correction of text and image content. This completely changes the outdated operating method of requiring separate processing and manual adjustment of text and image content, significantly improving the accuracy of review and realizing the full automation and high efficiency of the advertising compliance review process, greatly reducing the company's labor costs and compliance risks.

[0208] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0209] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0210] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. An advertising optimization method based on cultural compliance, characterized in that, The method includes: Obtain the initial advertising creatives to be reviewed and their target delivery area information; Cultural taboo features are extracted from the initial advertising materials to form a multidimensional taboo vector; Based on a preset cultural dynamic rule base, the multidimensional taboo vector is identified to obtain multiple benchmark cultural taboo contents and their cultural conflict indices; The benchmark cultural taboo content that exceeds the preset threshold in the cultural conflict index is recorded as the target cultural taboo content. The cross-modal adaptation engine is invoked to generate compliant alternatives for the target culturally taboo content; The target cultural taboo content in the initial ad creative is replaced with the compliant alternative to obtain the optimized target ad creative. The step of invoking the cross-modal adaptation engine to generate compliant alternatives for the target culturally taboo content includes: Based on the contrastive learning algorithm, multiple candidate alternatives are generated for the target cultural taboo content in the cross-modal semantic space; Obtain the detailed cultural rules corresponding to the regional information; Based on the detailed cultural rules, a regional suitability score is calculated for each of the candidate alternatives; Based on the regional suitability score, the optimal compliant alternative is dynamically selected from the multiple candidate alternatives. The step of obtaining the subdivided cultural rules corresponding to the regional information includes: Collect comment text data from social media platforms in the target delivery area within a preset time window; Based on a pre-defined cultural dimension dictionary, the comment words in the comment text data are divided into corresponding cultural dimensions; The number of comments under each cultural dimension is counted, and the fluctuation value of the number of comments under each cultural dimension within the preset time window is calculated. Based on the number of comments and the fluctuation value of the number of comments, a corresponding weight value is assigned to each cultural dimension to form a dynamically updated subdivided cultural rule for calculating the regional suitability score; wherein, the weight value = α·number of comments + β·fluctuation value of the number of comments, where α reflects general attention and β reflects recent sensitivity, and a time decay factor or adaptive learning coefficient is introduced to the weight value to achieve incremental updates. The weight value of each cultural dimension is written into the subdivided cultural rule entries to form a formal rule set, which can be updated periodically for use in the regional suitability score.

2. The advertising optimization method based on cultural compliance according to claim 1, characterized in that, After the step of replacing the corresponding target cultural taboo content in the initial advertising creative based on the compliant alternative to obtain the optimized target advertising creative, the method further includes: The target advertising creative is input into a pre-built digital twin pre-screening system; wherein the digital twin pre-screening system simulates the target user group in the target advertising area; Obtain simulated emotional feedback data output by the digital twin pre-screening system for the target advertising material; Based on the simulated emotional feedback data, a compliance risk assessment result for the target advertising creative is generated; Determine whether the compliance risk assessment results have achieved the expected assessment results; If the expected evaluation results are achieved, the target advertising material will be delivered to the target delivery area.

3. The advertising optimization method based on cultural compliance according to claim 2, characterized in that, Following the step of placing the target advertising material in the target placement area if the expected evaluation result is achieved, the method further includes: The system continuously scans comment data from social media platforms corresponding to the target advertising area using a pre-defined multilingual BERT model. Monitor the frequency of occurrence of preset cultural taboo keywords from the comment data; Monitor whether the growth rate of the occurrence frequency per unit time is greater than a preset growth rate; When the rate of increase of the frequency of occurrence per unit time is found to be greater than the preset rate of increase, an online early warning mechanism is triggered.

4. The advertising optimization method based on cultural compliance according to claim 1, characterized in that, The step of identifying the multidimensional taboo vector based on a preset cultural dynamic rule base to obtain multiple benchmark cultural taboo contents and their cultural conflict indices includes: From the cultural dynamic rule base, multiple cultural review rules associated with the target delivery area information are retrieved; each cultural review rule corresponds to a baseline cultural taboo content and a corresponding basic severity score. The dimension value in the multidimensional taboo vector corresponding to each of the benchmark cultural taboo contents is used as the confidence score for the existence of the benchmark cultural taboo content. The initial conflict score is obtained by multiplying the confidence score by the corresponding baseline severity score. Based on the target delivery area information, each triggered cultural review rule is assigned a regional weight coefficient, and the initial conflict score is multiplied by the regional weight coefficient to obtain the adjusted target conflict score. The system detects whether there are cross-modal collaboration violations in the initial advertising materials and determines the modal superposition coefficient based on the detection results. The target conflict scores are aggregated, and the aggregation result is multiplied by the modal superposition coefficient to obtain the cultural conflict index of the initial advertising material.

5. The advertising optimization method based on cultural compliance according to claim 1, characterized in that, Before the step of identifying the multidimensional taboo vector based on a preset cultural dynamic rule base to obtain multiple benchmark cultural taboo contents and their cultural conflict indices, the method further includes: Retrieve revised regulatory data from the target deployment area at preset intervals; The legal data was analyzed using a natural language processing model to identify newly added, modified, or repealed cultural taboo clauses. The identified taboo clauses are transformed into structured, machine-readable rules, and version tags and effective dates are added to them; Based on the version tag and effective time, the machine-readable rules are incorporated into the cultural dynamic rule base in an incremental update manner.

6. An advertising optimization device based on cultural compliance, characterized in that, The device includes: The acquisition module is used to acquire the initial advertising materials to be reviewed and their target delivery area information; The extraction module is used to extract cultural taboo features from the initial advertising material to form a multi-dimensional taboo vector; The identification module is used to identify the multidimensional taboo vector based on a preset cultural dynamic rule base, and obtain multiple benchmark cultural taboo contents and their cultural conflict indices; The marking module is used to mark the benchmark cultural taboo content that exceeds the preset threshold as the target cultural taboo content; The calling module is used to invoke the cross-modal adaptation engine to generate compliant alternatives for the target culturally taboo content; The replacement module is used to replace the target cultural taboo content in the initial advertising material based on the compliant alternative solution, so as to obtain the optimized target advertising material; The calling module includes: The candidate alternative generation submodule is used to generate multiple candidate alternatives for the target cultural taboo content in a cross-modal semantic space based on a contrastive learning algorithm. The sub-module for obtaining detailed cultural rules is used to obtain detailed cultural rules corresponding to the regional information. The regional suitability score calculation submodule is used to calculate a regional suitability score for each of the candidate alternatives based on the subdivided cultural rules. The compliance alternative selection submodule is used to dynamically select the optimal compliance alternative from the multiple candidate alternatives based on the regional adaptability score. The sub-module for obtaining detailed cultural rules includes: The data acquisition unit is used to collect comment text data from social media platforms in the target delivery area within a preset time window; The segmentation unit is used to segment the comment words in the comment text data into the corresponding cultural dimensions based on a preset cultural dimension dictionary; The statistics unit is used to count the number of comments under each cultural dimension and calculate the fluctuation value of the number of comments under each cultural dimension within the preset time window; A setting unit is used to set a corresponding weight value for each cultural dimension based on the number of comments and the fluctuation value of the number of comments, so as to form a dynamically updated subdivided cultural rule for calculating the regional suitability score; wherein, the weight value = α·number of comments + β·fluctuation value of the number of comments, where α reflects general attention and β reflects recent sensitivity, and a time decay factor or adaptive learning coefficient is introduced to the weight value to achieve incremental updates. The weight value of each cultural dimension is written into the subdivided cultural rule entries to form a formal rule set, which can be updated periodically for use in regional suitability scoring.

7. A computer-readable storage medium, characterized in that, The system contains a computer program that, when executed by a processor, causes the processor to perform the steps of the cultural compliance-based advertising optimization method as described in any one of claims 1 to 5.

8. An electronic device, characterized in that, The device includes a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the cultural compliance-based advertising optimization method as described in any one of claims 1 to 5.

Images