Advertising campaign management engine

Abstract

This invention provides a method, system, and computer storage medium for providing an advertising campaign management engine in an artificial intelligence system. [Solution] First, performance data and user preferences are analyzed through high-conversion design modeling and mining to identify design elements such as image styles and text layouts that promote engagement. Next, design requirements generation transforms the ad optimization profile into structured prompts that guide the refinement of the ad design. Finally, design fine-tuning leverages an advanced generative model to adjust ad elements such as images and layouts based on the generated requirements. The ad campaign management engine supports an iterative process that enables automatic updates and optimization, allowing ads to be tailored to specific demographic information.

Description

Background Art

[0001] Users can interact with generative artificial intelligence technologies in different types of applications and services to achieve computing tasks. Generative AI refers to a class of AI systems and algorithms designed to generate new data or content that is similar to the data on which it is trained, or in some cases, completely different from the data on which it is trained. Generative AI systems can create and support text generation, image generation, music and audio generation, video generation, and data synthesis. In particular, generative AI systems can support item listing systems in multiple ways to improve operational efficiency, customer engagement, and online shopping. For example, an item listing system can employ a generative AI system for content generation (e.g., product descriptions), personalized shopping experiences (e.g., recommendation engines), product discovery (e.g., visual search), and security management (e.g., fraud detection). An item listing system can utilize generative AI through application programming interfaces (APIs), pre-trained models, and custom AI solutions to enhance item listing system functionality.

Summary of the Invention

[0002] The various aspects of the technology described herein generally relate to systems, methods, and computer storage media for providing advertising campaign management using an advertising campaign management engine in an artificial intelligence system. Advertising campaign management refers to the process of planning, executing, monitoring, and optimizing advertising activities to achieve specific business objectives. Advertising campaign management includes targeting the right audience, selecting the right ad formats, tracking performance metrics, and making data-driven adjustments to maximize engagement, conversions, and return on investment. The advertising campaign management engine of this technological solution integrates both historical and real-time performance data to dynamically refine the ad design.

[0003] The ad campaign management engine comprises three main components. First, high-conversion design modeling and mining analyzes user preferences to identify design elements that drive engagement, such as image style and text layout. Second, design requirements generation transforms user profiles or user group profiles into structured prompts that guide the refinement of ad designs. Finally, design fine-tuning leverages advanced generative models to adjust ad elements such as images and layouts based on the generated requirements. The ad campaign management engine supports an iterative process that enables automated updates and optimization, allowing ads to be tailored to specific demographics. For example, it can generate quirky designs for younger demographics and simplified layouts for older demographics. The refined ads are then delivered and continuously monitored for performance improvements. The benefits of this approach include a personalized and relevant ad experience for users, increased conversion rates through real-time adjustments, automated refinement that reduces manual work and delays, and scalability that can adapt to evolving market needs and user preferences.

[0004] In operation, the ad campaign management engine optimizes ad campaigns by leveraging historical data, user feedback, and advanced generative models. The process begins by collecting ad campaign data such as design-performance metrics (clicks, conversion rates) and / or user feedback (explicit surveys or implicit engagement). This data is used to identify patterns and guide the refinement of ads.

[0005] The offline mining process analyzes ad design and performance data to understand which layouts, text styles, and image representations generate high engagement. User feedback is mined to reveal audience preferences, such as older viewers preferring detailed descriptions and large fonts, while younger users prefer minimalist designs. This data helps train ad optimization models to determine factors influencing ad layout and content, as well as image manipulation preferences across different demographic segments. Using these ad optimization models and ad campaigns, ad optimization profiles can be generated for each ad campaign.

[0006] The conditional text template generator uses these insights (i.e., ad optimization profiles) to recommend adjustments to ad layouts and images. For example, an ad targeting middle-aged professionals might have a bold, dark blue headline, while an ad targeting younger audiences might emphasize a striking image with minimal text. Advertisers provide raw campaign elements such as text and images, and the system applies personalized adjustments. The design requirements generator is used to generate fine-tuning prompts for ad campaigns.

[0007] Generative models are used to create the final ad. For example, a multimodal ad layout generative model ensures a cohesive arrangement of text and visuals, while an ad image manipulation model adjusts visual elements such as brightness and color balance. The refined ad is delivered, and its performance is continuously monitored. If performance is insufficient, the ad campaign management engine iteratively processes the ad by attempting fine-tuning to improve engagement. By automating the ad refinement process, the ad campaign management engine streamlines campaign management while creating personalized and relevant ads and reducing manual work. Its scalability and adaptability ensure its effectiveness in a dynamic advertising environment.

[0008] This summary is provided in a simplified form to introduce the selection of concepts that will be further described in the embodiments for carrying out the following inventions. This summary is not intended to identify any important or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter. [Brief explanation of the drawing]

[0009] The technologies described herein are described in detail below with reference to the accompanying drawings. [Figure 1A] This is a schematic diagram of an artificial intelligence system for providing advertising campaign management in an item listing system, according to an embodiment of the technology described herein. [Figure 1B] This is a schematic diagram of an artificial intelligence system for providing advertising campaign management in an item listing system, according to an embodiment of the technology described herein. [Figure 1C] This is a schematic diagram of an artificial intelligence system for providing advertising campaign management in an item listing system, according to an embodiment of the technology described herein. [Figure 1D]This is a schematic diagram of an artificial intelligence system for providing advertising campaign management in an item listing system, according to an embodiment of the technology described herein. [Figure 2A] This is a block diagram of an artificial intelligence system for providing advertising campaign management in an item listing system, according to an embodiment of the technology described herein. [Figure 2B] This is a flowchart of an artificial intelligence system for providing advertising campaign management in an item listing system, according to the aspects of the technology described herein. [Figure 3] This figure shows a first exemplary method for providing advertising campaign management in an item listing system according to aspects of the technology described herein. [Figure 4] This figure shows a second exemplary method for providing advertising campaign management in an item listing system according to aspects of the technology described herein. [Figure 5] This figure shows a third exemplary method for providing advertising campaign management in an item listing system according to aspects of the technology described herein. [Figure 6] This is a block diagram of an exemplary computing environment for an item listing system, suitable for use in implementing aspects of the technology described herein. [Figure 7] This is a block diagram of an exemplary distributed computing environment suitable for use in implementing aspects of the technology described herein. [Figure 8] This is a block diagram of an exemplary computing environment suitable for use in implementing aspects of the technology described herein. [Modes for carrying out the invention]

[0010] overview Item listing systems and platforms support storing items (products or assets) in an item database, receiving queries, and providing a search system to identify items as search results based on those queries. Items (e.g., physical or digital items) refer to products or assets offered for listing on the item listing platform. The search system supports identifying result items from the item database for received queries. The item database may specifically be for a content platform or item listing platform, such as the eBay Content Platform developed by eBay Inc. (San Jose, California). Item listing systems may also provide generative AI-supported applications ("generative AI applications") that leverage generative AI models (e.g., Large Language Models - "LLM") to create, generate, or produce content, data, or output. LLMs are a specific class of generative AI models primarily focused on generating human-like text. Generative AI models, such as GPT (Generative-Pre-trained Transformer) and its variations, are designed to generate human-like text or other types of data based on input received (e.g., via a prompt interface). These applications use generative AI to perform a variety of tasks across different domains, providing improvements in automation, efficiency, and human-like interaction.

[0011] Digital platforms that support advertising can include online services or websites that enable businesses or individuals to advertise their products, services, or content to targeted audiences. Digital platforms rely on a complex infrastructure of technologies that enable ad delivery, targeting, and measurement. At the core of these platforms are ad servers that store, manage, and deliver ads to users. These servers process requests from websites or apps, match them with appropriate ads based on targeting criteria, and deliver those ads in real time. Ad servers also track key performance metrics such as impressions, clicks, and conversions, which are important for reporting and analysis.

[0012] Digital platforms often collect vast amounts of data about user behavior, demographic information, and interactions through cookies, pixels, or device identifiers. This data is processed to create user profiles, which are then used to segment audiences for more precise targeting. Targeting mechanisms may include contextual targeting based on the content users are viewing, demographic targeting such as age or location, behavioral targeting based on past actions, and retargeting that focuses on users who have previously interacted with advertisements. All of this is designed to ensure that the right advertisement reaches the right user at the right time.

[0013] To measure advertising effectiveness, digital platforms employ advanced tracking and analytics systems. Metrics such as impressions, clicks, conversions, and return on ad spend (ROAS) are obtained in real time, allowing advertisers to evaluate the performance of their campaigns. Machine learning algorithms further optimize these campaigns by analyzing user interactions and adapting ad delivery to improve engagement and conversion rates.

[0014] Furthermore, many platforms offer APIs (Application Programming Interfaces) and SDKs (Software Development Kits), enabling advertisers and third-party tools to integrate with the platform. These integrations allow for the automation of ad creation, management, and reporting, as well as the ability to embed ads within external websites and applications. The sum total of these technological components forms a seamless and dynamic ecosystem that enables advertisers to reach highly targeted audiences, optimize ad performance, and track campaign success in real time.

[0015] Traditionally, digital platforms have not been configured with comprehensive computing logic and infrastructure to effectively fine-tune advertising campaigns based on ad design and performance data. Typically, while advertising campaigns are managed by the digital platform (e.g., search engines, e-commerce sites, or item listing services), campaign design is handled by the seller or a third-party design company. As a result, performance data obtained from advertising campaigns cannot be automatically or seamlessly used to refine the characteristics of the ad design. Advertising campaigns allow sellers to precisely identify their target audience (e.g., specific users or user groups) and connect with that audience. These campaigns include one or more ads targeted at the intended audience. However, even with precisely defined target audiences, the performance of an advertising campaign can be influenced by the characteristics of the ad design. In particular, certain performance data (e.g., conversion rates) can vary significantly among different target audiences.

[0016] As an example, consider an e-commerce platform running an advertising campaign for a seller (advertiser) advertising a new line of shoes. The ad targets two different groups: young adults aged 18-25 who are interested in fashion, and middle-aged adults aged 35-50 who prioritize comfort. Even if the ad perfectly targets each group, the results can differ significantly. For the younger audience, the ad's sophisticated and trendy design, with its bold visuals and strong emphasis on style, may resonate and lead to a high conversion rate. However, the same ad may not perform similarly well for the older group. Older adults prefer designs that emphasize comfort, durability, and functionality, resulting in a lower conversion rate for this group.

[0017] Such performance discrepancies indicate a core challenge of fine-tuning ad designs during campaigns. E-commerce platforms can collect conversion data such as click and purchase numbers from each group, but the process of adjusting designs based on this data is often fragmented. E-commerce platforms play a role in managing campaigns, but ad designs are typically created by sellers or third-party agencies. This division of labor makes it difficult to quickly adapt ad designs based on performance insights. For example, if the conversion rate of the elderly group is low, the platform's data is not necessarily directly linked to the ability of the design team to modify ads in real time, and the optimization process is delayed. As a result, even with accurate targeting, the success of a campaign still depends largely on how well the ads resonate with the specific preferences of each user group and how efficiently the design can be adapted based on performance data. Therefore, a more comprehensive item listing system with an alternative foundation for providing ad management functions can improve the computing processes and interfaces for ad management.

[0018] Description of Technical Solutions Conceptually, an ad campaign management engine enables technical solutions designed to optimize ad campaigns by dynamically refining ad designs based on performance data and user preferences. By leveraging data analysis, machine learning, and generative modeling, the ad campaign management engine functions as an end-to-end solution that improves ad effectiveness in real time and automates and streamlines the entire campaign optimization process.

[0019] In operation, the advertising campaign management engine uses past advertising design-performance data to identify advertising design patterns that focus on user preferences rather than traditional demographic segmentation to promote user engagement and conversion. Through advanced methods such as clustering and feature analysis, the advertising campaign management engine defines distinct user groups and associates the profiles of those user groups with specific design elements such as image style, text layout, and localization. In particular, the advertising optimization model is generated based on modeling the relationships between user groups, preference characteristics, and preference associations. The advertising optimization model enables the personalization of advertising content to match the unique preferences and expectations of each viewer segment.

[0020] To convert these insights into actionable outputs, the advertising campaign management engine processes an advertising optimization profile associated with a user profile or user group profile for an advertising campaign and a design requirement generator for generating structured design prompts in a format such as JSON. These outputs specify the necessary adjustments to advertising elements that are tailored to resonate with distinct user groups, ranging from image style and text placement to other fine-tuned design requirements.

[0021] Refinement is performed through generative models (e.g., design fine-tuning models and multimodal generative models) that adjust both visual and text advertising components. Tools such as image diffusion models ensure style consistency, while layout generators provide accurate placement of text and objects, seamlessly aligning the updated content with the specified preferences. This automation guarantees that advertising design maintains consistency, personalization, and effectiveness. The updated advertising campaign is delivered.

[0022] A continuous feedback loop is a fundamental feature of the ad campaign management engine's operation, enabling iterative improvement. By collecting performance data such as clicks and conversions from delivered ads, the ad campaign management engine identifies optimization opportunities, refines the design accordingly, and re-delivers updated content to maximize engagement and conversion rates.

[0023] In this way, the workflow includes defining user preferences through clustering and analysis, generating design requirements, refining ad content using generative models, delivering tailored ads, and iterative processing based on performance metrics. The ad campaign management engine is built to scale to accommodate large datasets and diverse user groups in production environments. The ad campaign management engine seamlessly integrates with e-commerce platforms and ad networks to automate traditional manual processes, adapt to new design features and evolving user preferences, and leverage real-time performance data. The ad campaign management engine demonstrates a transformative approach to campaign management with its ability to deliver highly personalized, efficient, and impactful ads, enhancing user engagement and driving higher conversion rates.

[0024] As an example, an e-commerce platform can be used to run a campaign promoting a new smartwatch. The advertiser (seller) aims to target three different user groups: young adults aged 18-30, middle-aged customers aged 35-50, and seniors aged 60 and over. Initially, three ad designs are created. Ad A features a sleek, modern design with bold images and minimal text to appeal to young adults. Ad B targets middle-aged users who value fitness tracking, emphasizing the health benefits of the smartwatch. Ad C is designed for seniors, emphasizing simplicity and a large, easy-to-read display.

[0025] As the campaign progresses, performance data reveals disparities in engagement. Ad A achieves a high conversion rate among young adults, while Ad C performs poorly because older adults perceive the ad layout as overly complex. Ad B shows moderate success, but could resonate better with its target group if it were more effectively tailored.

[0026] Ad campaign management engines can be used to dynamically optimize these ads. By analyzing performance metrics and using ad optimization models generated based on user preferences for ad elements such as text style, images, and layout, ad optimization models can be used in conjunction with design requirements generators to identify user preferences and fine-tune ad campaigns. For example, older adults may prefer brighter backgrounds and simplified text, while middle-aged users respond better to detailed health-related content placed in prominent positions.

[0027] Using this data, the ad campaign management engine generates structured prompts for refinement. For example, the ad campaign management engine outputs a JSON file specifying changes such as "increase text size," "use a lighter background," or "focus on fitness effects in the tagline." These specifications are processed by a multimodal generation model. For ad C, the ad campaign management engine updates the layout, placing key features more prominently and replacing complex visuals with clearer, brighter images. For ad B, the ad campaign management engine adjusts the text style to make fitness effects more clearly visible.

[0028] The updated ads are delivered, and this cycle repeats. New performance data shows improved engagement from older users, and ad C is now achieving a higher conversion rate. Similarly, ad B resonates more strongly with middle-aged users after refinement. Meanwhile, ad A remains highly effective, but subtle tweaks ensure it continues to align with the evolving preferences of younger adults.

[0029] This iterative process not only maximizes the effectiveness of each ad but also significantly reduces manual intervention. The ad campaign management engine automates the generation and application of refinements, ensuring that ads remain relevant and impactful throughout the entire campaign period. Advertisers benefit from a streamlined and scalable solution, while users receive personalized and engaging content tailored to their preferences.

[0030] Exemplary systems and resources The embodiments of the technical solutions can be illustrated by referring to Figures 1A to 1D as an example. Figure 1A shows an exemplary workflow related to providing advertising management functionality. As an example, consider an e-commerce platform that runs an online advertising campaign for an advertiser to promote a new smartwatch. In this campaign, the advertiser targets different user groups based on factors such as age, location, gender, and purchase history. The advertiser creates advertising designs that target each specific user group. For example, Ad A could feature a sophisticated and modern design targeting young adults aged 18-30, Ad B could emphasize the health benefits of the smartwatch to appeal to middle-aged customers aged 35-50, and Ad C could emphasize the ease of use and large display of the smartwatch for seniors aged 60 and over.

[0031] However, even with precise targeting, the success of an advertisement depends on its design. For example, ad A may perform very well with young adults and generate a high conversion rate, while ad C may not be as effective with older adults, who may not find the large display features appealing or prefer a simpler design. Similarly, ad B may resonate with middle-aged customers but fail to attract younger audiences. The conversion rate, i.e., the percentage of users who purchase a smartwatch after viewing an ad, will differ significantly between groups due to these differences in preferences. For instance, young adults may click on ad A and make a purchase, while older adults may ignore ad C, resulting in a lower conversion rate for that group.

[0032] Even with a clear understanding of their target user groups, advertisers face the challenge of fine-tuning their ads during a campaign. Ad designers, typically third-party organizations, operate independently of the e-commerce platform. While the platform can collect valuable data on each ad's performance—such as which groups clicked on which ads and how many made purchases—it's difficult to immediately refine ad designs in response to this data. For example, if ad C isn't performing well with older adults, but is still the most relevant ad for that group based on its characteristics, advertisers can't easily adjust the ad design to make it more appealing. This separation between campaign management and ad design often leads to longer delays in making necessary changes, meaning that by the time adjustments are made, the overall success of the campaign may have already been affected.

[0033] An ad campaign management engine can be designed to leverage historical campaign performance data, particularly conversion metrics, to generate dynamic design fine-tuning requirements for specific user groups. These requirements are then used to automatically refine the ads. The ad campaign management engine supports an iterative process in which updated designs are delivered, performance data is collected, and further optimizations are achieved.

[0034] The advertising campaign management engine operates based on three main components: high-conversion design modeling and mining, design requirements generation, and design fine-tuning models. Each component contributes to creating an end-to-end pipeline for dynamic ad refinement.

[0035] High-conversion design modeling and mining focuses on analyzing and modeling design patterns that influence user engagement and conversion rates. This process involves three key steps: user group definition, preference feature mining, and preference association. User group definition is used to categorize users into distinct groups based on their preferences for advertising features. This process differs from traditional user segmentation based on purchase behavior in that it focuses on preferences for advertising design elements. Techniques such as clustering and feature importance estimation are used to create actionable user profiles.

[0036] Preference feature mining involves identifying and classifying important features. These features are empirically selected and validated to ensure they are editable by state-of-the-art deep learning models. Features are divided into two main categories: ad image manipulation factors (e.g., image style: individual, classic; background: light, light; resolution: high resolution) and ad layout and content adjustment factors (e.g., content selection: important text only, detailed text; text position / size / style: centrally aligned slogan, light-colored text; and localization: language-specific options (e.g., Japanese, German)).

[0037] Preference association involves associating user profiles with their corresponding design preferences. A combination of manual labeling and data-driven rule mining ensures personalized ad tailoring for each user group.

[0038] Design requirements can be generated based on users or user groups associated with an advertising campaign. For example, once a user is identified, their profile and associated design preferences are processed by the design requirements generator. This component then translates the preferences into structured prompts suitable for the fine-tuning model.

[0039] for example, Input: User image and product details. Output: Design requirements in JSON format for a specific user group, such as the following:

[0040] [Table 1]

[0041] Design fine-tuning is achieved using advanced multimodal generation models. These models refine advertising elements such as layout, images, and text based on the generated requirements. Image refinement: Using models such as Sega Diffusion, advertising images are adapted to meet style preferences (e.g., unique, bright). Content layout adjustment: Tools such as PosterLLaVa generate advertising layouts with precise placement of text and images.

[0042] for example, Input: Ad image and layout preferences Output: Layout elements in JSON format for rendering below:

[0043] [Table 2]

[0044] Once the fine-tuning is complete, the updated ads will be delivered and displayed to users in a tailored format. Performance data will be continuously collected for further refinement. Example: For young female users viewing product advertisements Preferred image style: unique, bright, minimalist; Content: A centrally placed slogan and concise text. Refine advertising integrates these preferences to maximize user engagement and conversion rates.

[0045] The benefits expected from this approach are significant. First, users perceive ads as more personalized and relevant because they are tailored to their own style preferences. This level of customization can lead to more meaningful engagement with content, as ads better align with individual user preferences and interests. Second, the ability to adjust designs based on real-time performance data improves conversion rates. By continuously refining ad designs according to how they perform for different user groups, ads become more effective in driving actions such as purchases or sign-ups. Third, the ad refinement process is automated, reducing the need for manual intervention by sellers or third-party organizations. This streamlines the campaign management process, enabling faster adjustments and minimizing delays. Finally, the ad campaign management engine is scalable, with an extensible interface that can adapt to new design features and evolving user preferences. This flexibility ensures that the platform can adapt to the growing demands of product-scale deployment, while supporting the continuous development of advertising strategies as user behavior and market conditions change.

[0046] As shown in Figure 1A, the process begins with advertising campaign data 102A (e.g., performance data for detailed ad design and user feedback). This data serves as a basis for identifying patterns in ad effectiveness and user preferences.

[0047] Offline Mining 104A describes an offline mining process for extracting actionable insights from advertising campaign data. Two main data streams are analyzed: mining performance data for ad design and mining user feedback data. Mining performance data for ad design involves correlating performance metrics (e.g., clicks, conversions) with ad design features such as layout, text, and style. Mining user feedback data involves examining explicit feedback (e.g., surveys) and implicit feedback (e.g., engagement metrics) to understand user preferences.

[0048] The main elements mined include title, text, content, and style 106A. Examples of insights generated include that older users prefer more detailed text, larger fonts, and red elements; younger users engage better with minimalist, text-free ads; and female users are drawn to ads with pink-themed text. Image style 108A: Examples of style preferences include that female viewers prefer bright and vibrant images, while male viewers, especially younger ones, prefer darker, more melancholic visuals. The mined data is used to train an ad optimization model. This model identifies ad layout and content adjustment factors 110A (factors that influence how text and visual elements should be positioned) and ad image manipulation factors 112A (style adjustments to image features).

[0049] Referring to Figure 1B, Figure 1B illustrates an alternative embodiment of high-conversion design modeling and mining. Typical design patterns or features that can be modeled or edited by recent deep learning models need to be identified along with user preferences for these editable design patterns. This can be achieved by offline data mining, which can be broken down into three main steps. First, in user group definition 110B, users need to be classified into different groups, each group representing a clear intent or preference for advertising features. This process is similar to conventional user portrait mining for advertising, although the target is slightly different. While conventional user portrait mining focuses on grouping users based on their purchasing preferences, this work focuses on user preferences in advertising material design. Techniques such as user clustering or feature importance estimation are commonly used to achieve this.

[0050] Secondly, in preference feature mining 120B, ad features need to be identified because these features must be editable by the AI ​​model. Instead of relying on a data-driven approach, these features are manually classified through empirical methods and then validated online for their importance. Since ads consist primarily of images and text descriptions of items for sale, preference features are classified into two main types: ad image manipulation factors and ad layout and content adjustment factors. In the ad campaign management engine, ad image manipulation factors include image style (individualistic or classic image), background specifications (light or pastel background), and image details (high resolution, etc.). Regarding ad layout and content adjustment factors, these include content selection (important text only or as much ad text as possible), text position / size / style (centered slogan or pastel text), and language / country (Japan, Germany, or Italy, etc.). The advertising campaign management engine has the ability to expand its functionality by supporting these preference features and providing an interface for adding new features, enabling continuous upgrades to meet product requirements.

[0051] Finally, in preference association 130B, the outputs from the user group definition step and the preference feature mining step can be associated in such a way that when a user accesses the site, the user's profile is matched with preference features, enabling automatically personalized ad fine-tuning. This association process is performed through a combination of manual labeling and data-driven rule mining. In this way, preference association in the ad campaign management engine is the process of linking user profiles or group characteristics identified through user group definition with specific design preferences extracted through preference feature mining. This linking enables automated and personalized ad adjustments, ensuring that ads resonate with the preferences of the target audience and increase engagement and conversion rates. This process relies on a combination of data-driven rule mining and manual labeling to maintain both accuracy and flexibility.

[0052] User profiles are created for specific groups based on demographic data, behavioral patterns, and estimated preferences. For example, young professionals may be categorized as preferring sophisticated, minimalist designs, while older individuals tend to prefer advertisements with larger fonts and brighter backgrounds. Simultaneously, design preferences can be categorized into features such as text style, image brightness, and layout structure. Some users may prefer bold fonts and centered text, while others may respond better to soft color palettes and off-center designs.

[0053] Next, preference association links these user profiles or groups to relevant design features. Data-driven rule mining analyzes historical data to identify correlations between user attributes and ad performance metrics. For example, patterns in the data may show that middle-aged professionals engage more with ads featuring dark blue headlines and detailed descriptions. If data-driven methods are insufficient, manual labeling ensures that unique or complex preferences are accurately captured and associated with user profiles.

[0054] In an advertising campaign for fitness trackers, the ad campaign management engine can define three user groups: young professionals aged 25-35 who are tech-savvy and time-conscious; middle-aged fitness enthusiasts aged 40-55 who focus on health and wellness; and seniors aged 65 and over who prioritize ease of use and large display. Design preferences for these groups can reveal that young professionals prefer bold images and minimal text, middle-aged enthusiasts prefer ads that balance health information and visuals, and seniors respond best to simple layouts, large fonts, and high-contrast visuals.

[0055] Through preference association, these groups are linked to specific design features. For young professionals, advertisements may feature centrally placed product images, concise taglines, and a modern color palette such as black and metallic tones. For middle-aged fitness enthusiasts, advertisements can use a two-column layout with detailed health information on one side and product visuals on the other. For seniors, advertisements may include a high-contrast background, large sans-serif fonts, and a simplified single-column layout.

[0056] This association process allows the system to automatically adjust ad design based on identified profiles. A fitness tracker ad tailored for seniors might feature large, easy-to-read text like "Stay Active at Any Age," complemented by a bright background and a concise call to action. In contrast, an ad for younger professionals might highlight a tagline like "Track Your Peak Performance" in bold text overlaid on a minimalist product image.

[0057] Automating this process ensures real-time ad personalization, significantly increasing their relevance and effectiveness across diverse user groups. This capability not only improves engagement but also supports higher conversion rates by aligning ads with the specific needs and preferences of their intended audience.

[0058] Referring to Figure 1A, conditional text template generators (114A and 116A) are provided. Using insights from the ad optimization model, the ad campaign management engine develops conditional text templates for ad layout and content adjustments, including adjusting poster layouts such as titles, subtitles, and object placement, and for ad image adjustments, refining image stylistic elements to suit user preferences.

[0059] The delivery of an advertising campaign includes the advertiser (seller, 120A) providing an initial campaign that includes advertising text 122A consisting of content for a title, slogan, or description, and advertising images (124A) including visual elements such as product photographs or graphics.

[0060] The conditional text template generator 114A adjusts the advertising poster layout 130A by applying layout-specific refinements and aligns text elements to maximize engagement. Similarly, the conditional text template generator 116A allows for fine-tuning of the advertising image 132A, ensuring visual consistency with specified preferences.

[0061] In the ad refinement and generation stage, the refined text and image elements are integrated to generate the updated ad 140A. This process uses two advanced generation models: Multimodal ad layout generation model 134A: This model generates a cohesive layout by harmonizing the text and visual placement. Ad image manipulation model 136A: This model applies stylistic adjustments to images, such as brightness, color balance, or theme effects.

[0062] As shown in Figure 1C, a user profile, including a description 150A and attributes 160A such as demographic information, behavioral data, and aesthetic preferences, is used by a conditional text template generator 116A to fine-tune the ad image. This adjustment process adapts the visual elements of the ad, such as color scheme, image style, and spatial composition, to align with the preferences and expectations indicated by the user profile. For example, if the user profile specifies a preference for vibrant, high-contrast images and minimalist designs, the conditional text template generator 116A adjusts the ad image to reflect these characteristics by emphasizing brightness, sharpening key visual elements, and removing unnecessary background clutter.

[0063] Similarly, as shown in Figure 1D, the same user profile, including description 150A and attribute 160A, is further utilized by the conditional text template generator 114A in combination with an advanced generative model to create updated advertisements. This integration of profiles enables a comprehensive approach to ad refinement, where both text and visual elements are dynamically aligned to the user's preferences. The conditional text template generator 114A generates structured prompts that guide the advanced generative model to effectively synthesize and lay out the ad content. For example, if the user profile indicates a preference for a concise, centrally placed slogan with accompanying visual elements, the generator ensures that the text placement and font style emphasize clarity and impact, while the generative model refines the overall layout and aesthetic consistency of the ad.

[0064] These iterative adjustments are driven by user-centric data and executed with advanced computational tools, resulting in highly personalized and visually appealing ads designed to maximize user engagement and conversion rates.

[0065] The final output is an optimized ad tailored to the preferences of a specific user group, guaranteeing higher engagement and conversion rates. For example, a refined ad for an older audience might feature a red title, large font, detailed description, and a bright background. In contrast, an ad for a younger audience might emphasize a simpler design with minimal text and darker, more trendy images.

[0066] This iterative cycle of ad design optimization, delivery, and feedback ensures that the ad campaign management engine remains dynamic and responsive to changing user behavior and preferences. The scalability and adaptability of the ad campaign management engine enable seamless integration with large-scale advertising platforms and continuous improvement in ad campaign performance.

[0067] Referring to Figure 2A, Figure 2A shows an item listing system 100 that includes an artificial intelligence system 100A, an advertising campaign management engine 110, a mining and modeling engine 112, an advertising optimization model 114, an advertising optimization profile, a design requirements generator 118, advertising design-performance data 120, an advertising campaign 130, a design fine-tuning model 140, an updated advertising campaign 150, an advertising campaign management engine client 160A (administrator), an advertising campaign management engine client 160B (seller), an advertising campaign management engine client 160C (buyer), and iterative refinement 170.

[0068] The artificial intelligence system 100A establishes a dynamic framework for optimizing advertising campaigns by iteratively refining advertisements to align with user preferences and behavior. At its core, the advertising campaign management engine 110 coordinates an integrated pipeline for data mining, ad design optimization, and the generation of fine-tuned advertisements.

[0069] The mining and modeling engine 112 enables initial steps to extract actionable insights through three core activities: user group definition, preference mining, and preference association. User group definition classifies audiences based on attributes such as demographic information, behavioral patterns, and aesthetic preferences. For example, users may be segmented into groups such as young professionals who prefer minimalist layouts, or families who are drawn to vibrant and colorful designs. Preference mining identifies design features that resonate with these groups, such as font size, color palette, and image style. For example, older users may prefer large, clear fonts with detailed text, while younger audiences may prefer sleek, text-free designs. Preference association combines these mined insights with user profiles to establish personalized advertising design strategies, such as associating tech-savvy users with bold, high-contrast visuals that highlight technological innovation.

[0070] Insights derived from the mining and modeling processes are communicated to an ad optimization model 114 that generates a tailored ad optimization profile 116 for a user group or individual. The functionality of the ad optimization profile 116 is associated with user descriptions and user attributes that can be used to generate a blueprint that specifies the design adjustments needed to improve engagement and conversion. For example, an ad optimization profile may include a conditional text template that stipulates that environmentally conscious users should be highlighted with earth tones, images that reflect sustainability, and concise text that promotes environmental benefits.

[0071] The design requirements generator 118 converts these ad optimization profiles 116 into structured specifications for ad components. These specifications may include instructions for adjusting font sizes, rearranging text elements, or changing the visual tone of images. For example, a campaign targeting minimalism might specify a clear layout with a centrally placed product image, minimal text, and ample white space. The specifications are informed by a combination of historical ad design-performance data 120 and real-time campaign analysis to ensure relevance and accuracy.

[0072] Ad campaign 130 corresponds to the ad campaign provided by the advertiser. Ad campaign 130 includes raw text and visual elements that serve as the basis for refinement. Design fine-tuning model 140 implements the design requirements for generating the improved ad. Using an advanced generative model, the fine-tuning process may include increasing font sharpness, adjusting image brightness, or rearranging layout elements to maximize visibility and user engagement. The refined output results in an updated ad campaign 150, which is delivered to a targeted audience. This updated campaign 150 follows optimization profiles and design specifications to meet the preferences of specific users. For example, a campaign designed for seniors may include larger text with a light background, while a campaign for young adults may emphasize sophisticated images with concise and modern text placement.

[0073] Iterative refinement 170 supports a feedback-driven process that continuously improves ad design. When an updated campaign is delivered, performance metrics such as click-through rate, conversion rate, and engagement duration are collected and analyzed. These metrics are communicated to the mining and modeling engines, enabling refinement through subsequent iterations. For example, if an ad is not performing well due to overly complex visuals, the ad campaign management engine 110 simplifies the design in the next iteration while maintaining the core messaging elements.

[0074] The advertising campaign management engine 110 works with advertising campaign management engine clients such as administrators (advertising campaign management client 160A) who monitor system operation and metrics, sellers (advertising campaign management client 160B) who supply initial advertising content and track campaign success, and buyers (advertising campaign management client 160C) who provide performance data for further refinement through interaction and feedback. These stakeholders collaborate within the ecosystem to achieve optimized advertising results.

[0075] The integration of advanced AI, machine learning, and iterative adjustments ensures that each ad campaign is progressively optimized for better engagement and conversions. The ad campaign management engine 100 delivers measurable results while providing advertisers with a scalable and adaptable solution that meets the diverse needs of users, ultimately transforming the landscape of dynamic advertising.

[0076] For illustrative purposes, the ad campaign management engine 110 is designed to generate dynamic design refinements for specific user groups by utilizing historical campaign performance data (i.e., ad design-performance data 120), particularly conversion metrics. These refinements are used to automatically adjust ads, and the ad campaign management engine 110 supports an iterative process of continuously collecting performance data, optimizing designs, and delivering updated content to increase user engagement and conversion rates.

[0077] The advertising campaign management engine 110 operates through a series of interrelated components that work together to create an end-to-end pipeline for dynamically refining ads. The first key component (i.e., the mining and modeling engine 112) focuses on analyzing design patterns that influence user engagement and conversion. This process begins by defining user groups based on user preferences for ad features, in contrast to traditional demographic segmentation that typically focuses on purchasing behavior. Using advanced techniques such as clustering and feature importance estimation, actionable user profiles are created that highlight ad features that resonate with specific groups.

[0078] The next step in this process is preference feature mining, which involves identifying and classifying the design elements that have the greatest impact on user engagement. These features are validated and selected based on their ability to be tuned by state-of-the-art deep learning models. Features are divided into categories such as ad image manipulation (e.g., style, background, resolution) and content adjustment (e.g., text style, placement, and language preferences). The final step in this stage is preference association, in which user profiles are linked to the design features that best match their preferences. This is achieved through a combination of manual labeling and data-driven rule mining, ensuring that ads are tailored to meet the specific needs of each user group.

[0079] The second key component of the advertising campaign management engine 110 is design requirements generation (e.g., design requirements generator 118), which processes user profiles and design preferences to create structured design prompts. These prompts specify the necessary adjustments for various advertising components, including image styles, text placement, and localization, and are used by the design fine-tuning model of the advertising campaign management engine 110. The generated prompts provide input for the next stage of the process, which focuses on fine-tuning the advertising design.

[0080] Design fine-tuning is performed using advanced generative models (e.g., design fine-tuning model 140) that adjust both the visual and text elements of the ad. Multimodal models, such as image diffusion models, adapt images to align with style preferences, and layout generators precisely position text and other elements within the ad to meet design requirements. The result is a refined ad design that aligns with the specific preferences of the target user group. Once the ad designs are fine-tuned, they are delivered, and performance data, including user interactions and conversion rates, is collected to guide further optimization.

[0081] A continuous feedback loop (i.e., iterative refinement) ensures that ads are constantly improved based on real-time performance metrics. Ads can be regularly updated and rearranged to maximize engagement, and each iteration becomes more effective in achieving the desired results. The process of refining ads through this engine enables significant customization, ensuring that ads are more relevant and personalized to users, and ultimately driving higher engagement and conversion rates.

[0082] For clarity and efficient reference, a glossary of key terms and concepts related to technical solutions is provided below. The advertising campaign management engine 110 also possesses the ability to handle large datasets and evolving user preferences, providing scalability. The automation of the design refinement process reduces the need for manual intervention, streamlining the campaign management workflow and enabling faster and more responsive adjustments. Furthermore, the advertising campaign management engine 110 is adaptable, supporting new design features and user preferences as they emerge, ensuring it remains effective in changing market conditions. This scalability and flexibility makes the advertising campaign management engine well-suited for product-scale delivery where rapid iteration and continuous optimization are essential for the success of advertising strategies.

[0083] Ad Design - Performance Data 120 refers to historical data collected from ad campaigns, including performance metrics such as click-through rates and conversion rates, as well as user feedback obtained through explicit surveys or implicit engagement behaviors. This data forms the basis for modeling the relationships between user groups, preference characteristics, and association mechanisms, which are essential for optimizing ad design and targeting.

[0084] User group definition is the process of segmenting users into different categories based on attributes such as demographic information, behavioral patterns, and estimated preferences. This segmentation identifies meaningful groups that share common characteristics, such as a preference for minimalist design or detailed text.

[0085] Preference features (or factors) are editable design characteristics of advertisements and are categorized into ad image manipulation factors and ad layout and content adjustment factors. Image manipulation factors include properties such as style, brightness, and resolution, while content adjustment factors cover text placement, font size, and language localization. These features form the basis for generating conditional text templates and are essential for fine-tuning ad designs to match user preferences.

[0086] Preference association is a hybrid framework that connects user profiles to preference features. This process combines data-driven rule mining or machine learning techniques to ensure accurate and personalized ad tailoring for individual users or groups.

[0087] The ad optimization model 114 is a predictive framework created using insights derived from ad design-performance data. The ad optimization model 114 maps relationships between user groups, user group preferences, and how these preferences relate to specific design elements. It generates conditional text templates and optimization profiles, enabling customization of ad layouts, content, and image factors to effectively fit targeted campaigns.

[0088] An ad optimization profile (e.g., ad optimization profile 116) represents a structured dataset that can be used to specify the refinements needed for ad design, including layout adjustments, content adjustments, and image manipulation, to align with the needs and expectations of the target audience. An ad optimization profile is associated with user descriptions and user attributes for a specific user or user group.

[0089] Conditional text templates are structured representations of design requirements generated by advertising optimization models. These templates specify layout and content adjustment factors as well as image manipulation factors, providing clear instructions for refining ad layouts and visuals to better align with user preferences.

[0090] The design requirements generator 118 is often implemented as a Large Language Model (LLM) and creates fine-tuning prompts based on ad optimization profiles and campaign specifications. By using a multimodal layout fine-tuning template, the generator produces structured prompts tailored for use by the design fine-tuning model.

[0091] The design fine-tuning model 140 is a generative AI model or multimodal layout generator that adjusts the ad layout and visuals according to fine-tuning prompts. These models refine the ad elements to create a cohesive and visually appealing design that aligns with the preferences of the target audience. Specifically, the multimodal layout generator can automatically create designs by combining different types of content, such as text and images, and organizing them into a cohesive layout. The generator uses algorithms or AI models to understand and arrange these elements, often based on context and predetermined design principles, to enhance visual appeal and usefulness.

[0092] A multimodal layout generator may be associated with a multimodal layout tweaking prompt template, which is a structured input designed to guide the fine-tuning of a model for generating or optimizing layouts that integrate various types of content (e.g., text and images). This template typically includes instructions on how to organize and position different modalities, taking into account factors (i.e., ad image manipulation factors and ad layout and content adjustment factors). The LLM may use the multimodal layout tweaking prompt template associated with the ad optimization profile to generate tweaking prompts for designing the tweaking model.

[0093] The advertising campaign management engine 110 integrates the advertising optimization model 114, the design requirements generator 118, and the design fine-tuning model 140. To support iterative refinement cycles, the advertising campaign management engine 110 manages real-time performance metrics, design optimization, and auto-adjustments to ensure continuous improvement of advertising campaigns.

[0094] Accessing historical ad design-performance data 120 involves retrieving and analyzing metrics and feedback from previous campaigns to identify patterns, user preferences, and success-driving factors. Historical ad design-performance data 120 is used to generate ad optimization models, which map relationships between user groups, preferences, and association mechanisms and generate conditional text templates to guide layout and image adjustments.

[0095] Advertisers provide raw campaign materials such as text and images, which are then processed through an optimization pipeline. The ad campaign management engine 110 generates ad optimization profiles that can be used to associate user attributes with specific refinement requirements, such as preferred font size or image brightness. These ad optimization profiles can be used to generate conditional text templates that notify the design requirements generator, which then converts the conditional text templates into structured fine-tuning prompts.

[0096] The design fine-tuning model 140 uses these prompts to adjust ad elements, refining text placement, color schemes, and visuals. The result is a tailored ad campaign designed to resonate with specific user preferences. Once delivered, these refined campaigns integrate user preferences into the design and maximize engagement.

[0097] Performance metrics from delivered ads are fed back into the ad optimization model, enabling the generation of subsequent optimized profiles and updated ads. This iterative refinement and feedback loop ensures that each cycle improves engagement and conversion rates, continuously enhancing the effectiveness of the ad campaign.

[0098] Referring to Figure 2B, Figure 2B shows a schematic diagram 200B relating to providing an advertising campaign management engine according to the embodiments described herein. The technical solution of the advertising campaign management engine can be illustrated by steps and an example advertising campaign.

[0099] Step 201B: Collecting Historical Ad Data. This process begins by collecting historical data from previous ad campaigns to understand performance trends. Historical ad data includes images used in campaigns, such as sophisticated photographs of minimalist technology products, and text layouts, such as bold fonts centered around calls to action like "Buy Now!". Metrics such as click-through rate (CTR) and conversion rate are also obtained. For example, a particular ad may have had a 15% CTR and an 8% conversion rate. For instance, an ad targeting "young professionals" achieved a 20% CTR by utilizing a simple design with minimal text, demonstrating a preference for minimalism in this segment.

[0100] Step 202B: User Group Definition. Next, users are segmented into distinct clusters based on demographic and behavioral data. These clusters may include groups such as "tech enthusiasts," "environmentally conscious buyers," and "luxury-oriented consumers." Each cluster reflects unique preferences. For example, "environmentally conscious buyers" tend to prefer visuals with green tones and messages emphasizing sustainability, such as "Eco-friendly. Planet first." This segmentation ensures that advertisements can be tailored to resonate with the specific preferences of each group.

[0101] Step 203B: Preference Feature Mining and Preference Association. After user segments are defined, key features of successful ads are extracted and analyzed to determine their impact on performance. This includes visual elements such as whether images are minimal, monochrome, or vibrant, as well as textual elements such as font size, content length, and placement. For example, "tech enthusiasts" show higher engagement rates with ads featuring bold, centered fonts and minimalist product images. By identifying these preferences, the system can fine-tune ad elements for better alignment with user expectations.

[0102] Step 204B: Training the ad design optimization model. The ad campaign management engine then trains an AI model to associate user preferences with high-performing ad features. Convolutional neural networks (CNNs) may be used to analyze image preferences, while transformers handle text optimization. For example, the model might predict that an ad targeting "luxury enthusiasts" will achieve a 25% CTR if it incorporates high-resolution images of luxury goods, gold-toned fonts, and a clear, minimalist layout. These predictions guide ad design to maximize engagement.

[0103] Step 205B: Design Fine-tuning. Once the ad design optimization model is in place, the ad campaign management engine can update ads based on the ad optimization profile, ad optimization model, and design fine-tuning model. In some embodiments, it is assumed that ads may be dynamically updated in real time as users access the platform. For example, a user identified as a "tech enthusiast" might be shown a sophisticated black-and-white ad with bold text emphasizing innovation. On the other hand, an "environmentally conscious buyer" might encounter an ad with a green color scheme and a message about sustainability. This real-time adjustment ensures that the content resonates immediately with the audience and enhances the potential for interaction.

[0104] Step 206B: User Interaction and Feedback. After delivering optimized and / or personalized ads, the ad campaign management engine tracks user interactions, including metrics such as CTR, conversion rate, and time spent engaging with the ad. For example, a newly designed ad targeting "Young Professionals" might achieve a 30% CTR, indicating a successful adjustment. This feedback loop provides useful data for evaluating the effectiveness of different ad elements.

[0105] Step 207B: Iterative refinement using performance data. Finally, the ad campaign management engine periodically updates its AI model with the latest data to ensure continued accuracy and relevance. For example, new data might reveal that "tech enthusiasts" tend to prefer dark mode-themed ads, prompting the system to adjust its recommendations. By retraining the model with new insights, the ad campaign management engine continues to adapt to evolving user preferences and market trends.

[0106] For example, historical data analysis reveals that younger users are more likely to engage with minimalist ad designs featuring bold images. This insight is used to inform the creation of conditional text templates that prioritize reducing text and increasing the prominence of visual elements in future campaigns. By aligning with these preferences, the ad campaign management engine ensures that ad designs resonate effectively with this demographic information.

[0107] The process of creating ad optimization profiles associates specific user groups with their unique preferences. For example, middle-aged professionals are identified as preferring detailed text and subdued color schemes. An ad optimization profile is generated for this user group, and this profile can be used to design adjustments such as a dark blue, centrally aligned headline with supporting text placed below it, ensuring clarity and alignment with their expectations.

[0108] Using these ad optimization profiles, the design requirements generator creates structured prompts to guide the refinement process. For environmentally conscious campaigns, the design requirements generator generates fine-tuning prompts that instruct the design fine-tuning model to incorporate nature elements into images and apply green tones to text layouts. These prompts ensure that the resulting ads reach their intended audience effectively.

[0109] The fine-tuning process itself utilizes advanced generative models to refine advertising elements. For example, an advertisement for a smartwatch undergoes adjustments to its image, applying increased brightness and sharper contrast to highlight key product features. Simultaneously, the accompanying text is repositioned to the upper right to improve readability and draw attention to important details.

[0110] If a delivered campaign performs poorly, the ad campaign management engine performs iterative improvements. In scenarios targeting seniors, ads with overly complex visuals and small font sizes are redesigned. The layout is simplified, the font size is increased, and the background is changed to a lighter color tone. These adjustments result in increased engagement and demonstrate the system's ability to dynamically adapt to feedback. By combining advanced analytics, machine learning, and generative design tools, this technical solution automates ad refinement, ensuring that ads remain scalable, adaptable, and tailored to user needs. Each iteration delivers more personalized and effective campaigns, improving engagement and achieving measurable results.

[0111] Embodiments of the technical solution may be described, for example, with reference to Figures 1A to 1D, Figure 2A, and Figure 2B. Figure 2A is a block diagram of an exemplary technical solution environment, based on the exemplary environment described with reference to Figures 6, 7, and 8, used when implementing embodiments of the technical solution. Generally, the technical solution environment includes a technical solution system suitable for providing an exemplary item listing system 600 in which the method of the present disclosure may be employed. Specifically, Figure 2A shows a high-level architecture of an item listing system 100 in an implementation of the present disclosure. Together with other engines, managers, generators, selectors, or components (collectively referred to herein as “components”) not shown, the item listing system 100 of Figure 2A supports the functions described in Figures 1A to 1D.

[0112] Exemplary Method Referring to Figures 3, 4, and 5, flowcharts illustrating a method for providing an advertising campaign management engine in an artificial intelligence system are shown. The method may be performed using the artificial intelligence system described herein. In embodiments, there are computer-executable or computer-usable instructions embodied on one or more computer storage media, which, when executed by one or more processors, cause one or more processors to execute a method (e.g., a computer implementation method) in an artificial intelligence system (e.g., a computerized system or computer system).

[0113] Referring to Figure 3, a flowchart illustrating method 300 for providing an advertising campaign management engine in an artificial intelligence system is provided. In block 302, historical ad design-performance data associated with multiple historical ad campaigns is accessed. In block 304, an ad optimization model is generated. In block 306, the ad campaigns are accessed. In block 308, an ad optimization profile for the ad campaigns is generated. In block 310, fine-tuning prompts are generated based on the ad optimization profile for the ad campaigns. In block 312, an updated ad campaign is generated. In block 314, the updated ad campaign is delivered.

[0114] Referring to Figure 4, a flowchart illustrating method 400 for providing an advertising campaign management engine in an artificial intelligence system is provided. In block 402, the advertising campaign is accessed. In block 404, an advertising optimization profile is generated for the advertising campaign, and the advertising optimization profile is associated with a user. In block 406, a fine-tuning prompt is generated based on the advertising optimization profile and the advertising campaign. In block 408, an updated campaign is generated. In block 410, the updated campaign is delivered to the user associated with the advertising optimization profile.

[0115] Referring to Figure 5, a flowchart illustrating method 500 for providing an advertising campaign management engine in an artificial intelligence system is provided. In block 502, historical ad design-performance data associated with multiple historical ad campaigns is accessed. The historical ad design-performance data includes ad performance metrics and user feedback data. In block 504, an ad optimization model is generated. Generating the ad optimization model is based on offline processing, including data analysis for user segmentation and preference insights associated with the ad design-performance data. In block 506, the ad optimization model is deployed.

[0116] technical improvements Embodiments of the present invention are described with reference to several inventive features (e.g., operations, systems, engines, and components) relating to an item listing system. The inventive features described include, with respect to an advertising campaign management engine associated with an artificial intelligence system, operations, interfaces, data structures, and arrangement of computing resources relating to providing the functions described herein.

[0117] Embodiments of the present invention relate to the field of computing, and more particularly to the field of artificial intelligence systems. The exemplary embodiments described below provide, among other things, systems, methods, and program products for performing the operation of a generative AI security engine that provides advertising campaign management. Thus, these embodiments improve the fields of artificial intelligence technology and item listing platform technology, enhancing the efficiency and effectiveness of advertising management through advanced data-driven techniques and automation. For example, by leveraging AI to analyze historical ad design-performance data, user preferences, and feedback, these embodiments enable the dynamic and precise optimization of ads tailored to specific user groups or individuals. This integration of AI into the advertising campaign management process reduces computational redundancy by automating iterative adjustments and eliminates the need for iterative manual refinement. In addition, the use of advanced generative models and structured prompts ensures that ad layouts and designs are consistently aligned with user preferences in real time, significantly improving engagement rates and conversion results.

[0118] In an item listing platform, this embodiment improves the relevance of ads displayed to users, resulting in a better overall user experience. This personalization directly benefits advertisers while optimizing platform resource usage by increasing conversion potential and prioritizing high-performance ad placement. The combination of scalability, adaptability, and real-time optimization positions this embodiment as a significant advancement in both artificial intelligence and item listing platform technology.

[0119] The functionality of embodiments of the present invention will be further illustrated by implementation examples and case studies, which demonstrate the operation of providing advertising campaign management using an advertising campaign management engine in an artificial intelligence system to improve computing processing in an artificial intelligence system as a means of solving specific problems in artificial intelligence technology. Overall, these improvements result in reduced CPU computation, reduced memory requirements, and increased flexibility in the artificial intelligence system compared to the operation of previous conventional artificial intelligence systems performed for similar functionality.

[0120] Additional explanations to supplement the detailed description of the invention Exemplary Item Listing System Environment Referring now to Figure 6, Figure 6 illustrates a computing environment of an exemplary item listing system 600 in which embodiments of the present disclosure may be employed. In particular, Figure 6 shows a high-level architecture of an exemplary item listing platform 610 that may host the environment of the technical solution or a portion thereof. It should be understood that this and other configurations described herein are described as examples. For example, as stated above, many of the elements described herein may be implemented as individual or distributed components, or together with other components, in any suitable combination and arrangement. Other configurations and elements (e.g., machines, interfaces, functions, sequences, and groupings of functions) may be used in addition to or instead of those illustrated.

[0121] The item listing system 600 may be a cloud computing environment that provides computing resources for functions related to the item listing platform 610. For example, the item listing system 600 supports the provision of computing components and services, including servers, storage, databases, networking, applications, and machine learning, related to the item listing platform 610 and client devices 620. Multiple client devices (e.g., client devices 620) include hardware or software for accessing resources on the item listing system 600. Client devices 620 may include applications (e.g., client application 622) and interface data (e.g., client application interface data 624) that support client-side functions related to the item listing system. Multiple client devices can access the computing components of the item listing system 600 via a network (e.g., network 630) to perform computing processes.

[0122] The Item Listing Platform 610 is responsible for providing a computing environment or architecture that includes infrastructure to support the provision of Item Listing Platform functions (e.g., e-commerce functions). The Item Listing Platform supports storing items in an item database, receiving queries, and providing a search system to identify search results based on those queries. The Item Listing Platform may also provide a computing environment with features for managing, selling, buying, and recommending different types of items. Specifically, the Item Listing Platform 610 may be a content platform or e-commerce platform, such as the eBay Content Platform developed by eBay Inc., located in San Jose, California.

[0123] The item listing platform 610 may provide item listing processing 630 and item listing interfaces 640. Item listing processing 630 may include service processing, communication processing, resource management processing, security processing, and fault tolerance processing to support specific tasks or functions in the item listing platform 610. Item listing interfaces 640 may include service interfaces, communication interfaces, resource interfaces, security interfaces, and management and monitoring interfaces to support functions between item listing platform components. Item listing processing 630 and item listing interfaces 640 enable communication, coordination, and seamless functionality of the item listing system 600.

[0124] For example, features associated with the Item Listing Platform 610 may include shopping processing (e.g., product search and browsing, product selection and shopping cart, checkout and payment, and order tracking), user account processing (e.g., user registration and authentication, and user profiles), seller and product management processing (e.g., seller registration, product listing, and inventory management), payment and financial processing (e.g., payment processing, refunds, and returns), order fulfillment processing (e.g., order processing and fulfillment, and inventory management), customer support and communication interfaces (e.g., customer support chat / email and notifications), security and privacy interfaces (e.g., authentication and authorization, payment security), recommendation and personalization interfaces (e.g., product recommendations, customer reviews, and ratings), analytics and reporting interfaces (e.g., sales and inventory reporting, and user behavior analysis), and API and integration interfaces (e.g., APIs for third-party integrations).

[0125] The item listing platform 610 may provide an item listing platform database (e.g., item listing platform database 650) for efficiently managing and storing different types of data. The item listing platform database 650 may include relational databases, NoSQL databases, search databases, cache databases, content management systems, analytics databases, payment gateway databases, customer relationship management databases, log and error databases, inventory and supply chain databases, and multichannel databases, which, when used in combination, efficiently manage data and provide users with an e-commerce experience.

[0126] The item listing platform 610 supports applications (e.g., application 660) which are computer programs, software components, or services that provide specific functions or sets of functions to meet specific item listing platform requirements or user requirements. Applications can be client-side (user-facing) and server-side (backend). Applications may also include applications without AI support (e.g., application 662), applications supported by traditional AI models (e.g., application 664), and applications supported by generative AI models (e.g., application 666). Examples of applications may include online point-of-sale applications, mobile shopping apps, management and operations consoles, payment gateway integrations, user account and authentication applications, search and recommendation engines, inventory and inventory management applications, order processing and fulfillment applications, customer support and communication tools, content management systems, analytics and reporting applications, marketing and promotion applications, multi-channel integration applications, log and error tracking applications, customer relationship management (CRM) applications, security applications, and APIs and web services, which are used in combination to efficiently provide users with an e-commerce experience.

[0127] The item listing platform 610 may include a machine learning engine (e.g., machine learning engine 670). Machine learning engine 670 refers to a machine learning framework or platform that provides the infrastructure and tools for designing, training, evaluating, and deploying machine learning models. Machine learning engine 670 can function as a foundation for developing and deploying machine learning applications and solutions. Machine learning engine 670 may also provide tools for visualizing data and model results, and for interpreting model decisions to gain insights into how the models are making predictions.

[0128] The Machine Learning Engine 670 can provide the libraries, algorithms, and utilities necessary to perform various tasks within a machine learning workflow. This workflow may include data processing, model selection, model training, model evaluation, hyperparameter tuning, scalability, model deployment, inference, integration, customization, and data visualization. The Machine Learning Engine 670 includes pre-trained models for various tasks, thereby simplifying the development process. In this way, the Machine Learning Engine 670 streamlines the entire machine learning process, from data preparation and model training to deployment and inference, making it accessible and efficient for different types of users working on a wide range of machine learning applications (e.g., customers, data scientists, machine learning engineers, and developers).

[0129] The machine learning engine 670 may be implemented in the item listing system 600 as a component that enhances various aspects of the functionality of the item listing system by leveraging machine learning algorithms and techniques (e.g., machine learning algorithm 672). The machine learning engine 670 may provide a selection of machine learning algorithms and techniques used to train computers to learn from data and make predictions or decisions without explicit programming. These techniques are widely used in various applications across different industries and may include supervised learning (e.g., linear regression: classification, support vector machines (SVM)), unsupervised learning (e.g., clustering, principal component analysis (PCA), correlation rules (e.g., a priori)), reinforcement learning (e.g., Q-learning, deep Q networks (DQN)), and deep learning (e.g., neural networks, convolutional neural networks (CNN), and recurrent neural networks (RNN)), and ensemble learning random forests.

[0130] The machine learning training data 120 supports the process of building, training, and fine-tuning a machine learning model. The machine learning training data 120 consists of labeled datasets used to train a machine learning model to recognize patterns, make predictions, or perform specific tasks. Training data typically includes two main components: input features (X) and labels or target values ​​(Y). Input features may include variables, attributes, or characteristics used as input to the machine learning model. Depending on the nature of the problem, input features (X) can be numerical, categorical, or textual. For example, in a model for predicting house prices, input features might include the number of bedrooms, floor area, and neighborhood. Labels or target values ​​(Y) contain the values ​​that the model is intended to predict or classify. Labels represent the desired output or ground truth for each corresponding set of input features. For example, in a spam email classifier, labels indicate whether each email is spam or not (i.e., binary classification). The training process involves presenting the model with training data, and the model learns to make predictions or decisions by identifying patterns and relationships between input features (X) and target values ​​(Y). The machine learning algorithm tunes its internal parameters during training to minimize the difference between the actual labels in the training data and their predictions. The machine learning engine 670 can use historical and real-time data to train models, make predictions, and continuously improve performance and user experience.

[0131] The machine learning engine 670 may include machine learning models (e.g., machine learning model 676) generated using the machine learning engine workflow. Machine learning model 676 may include generative AI models and conventional AI models, both of which may be used in the item listing system 600. Generative AI models are designed to generate new data, often in the form of text, images, or other media, based on patterns and knowledge learned from existing data. Generative AI models can be used in a variety of ways, including content generation, product image generation, personalized product recommendations, natural language chatbots, and content summarization. Conventional AI models encompass a wide range of algorithms and techniques and can be used in a variety of ways, including recommendation systems, predictive analytics, search algorithms, fraud detection, customer segmentation, image classification, natural language processing (NLP), and A / B testing and optimization. Often, a combination of both generative and conventional AI models can be used to deliver a fully comprehensive and effective e-commerce experience that combines data-driven insights and creativity.

[0132] The machine learning engine 670 can be used to analyze data, make predictions, and automate processes to provide users with a more personalized and efficient purchasing experience. Examples include product recommendations, search, and filtering; price optimization; inventory and inventory management; customer segmentation; customer churn prediction and retention; fraud detection; sentiment analysis; customer support and chatbots; image and video analysis; and advertising targeting and marketing. Specific applications of machine learning in the item listing platform 610 may vary depending on specific goals, available data, and resources.

[0133] Exemplary Distributed Computing System Environment Referring here to Figure 7, which shows an exemplary distributed computing environment 700 in which an implementation of the present disclosure may be adopted. In particular, Figure 7 shows a high-level architecture of an exemplary cloud computing platform 710 that can host an environment or part thereof of a technical solution (e.g., a data trusty environment). It should be understood that this and other configurations described herein are described only as examples. For example, as stated above, many of the elements described herein may be implemented as separate or distributed components or in combination with other components, and in any suitable combination and arrangement. Other configurations and elements (e.g., machines, interfaces, functions, sequences, and groupings of functions) may be used in addition to or instead of those illustrated.

[0134] A data center may support a distributed computing environment 700, which includes a cloud computing platform 710, racks 720, and nodes 730 (e.g., computing devices, processing units, or blades) within the racks 720. The environment of the technical solution may be implemented using a cloud computing platform 710 that runs cloud services across different data centers and geographical areas. The cloud computing platform 710 may implement components of a fabric controller 740 for provisioning and managing the allocation, deployment, upgrade, and management of resources for cloud services. Typically, the cloud computing platform 710 functions to store data or run service applications in a distributed manner. The cloud computing infrastructure 710 in the data center may be configured to host and support the operation of endpoints for a particular service application. The cloud computing infrastructure 710 may be a public cloud, a private cloud, or a dedicated cloud.

[0135] Node 730 may be provisioned to have a host 750 (e.g., an operating system or runtime environment) running a software stack defined on Node 730. Node 730 may also be configured to run specialized functions (e.g., compute nodes or storage nodes) within the cloud computing platform 710. Node 730 is assigned to run one or more parts of a tenant's service application. A tenant can refer to a customer that utilizes the resources of the cloud computing platform 710. The components of the cloud computing platform 710's service application that support a particular tenant can be referred to as a multi-tenant infrastructure or tenancy. The terms service application, application, or service are used interchangeably herein and broadly refer to any software or part of software that operates on or accesses the location of storage and compute devices within a data center.

[0136] If two or more separate service applications are supported by node 730, node 730 may be divided into virtual machines (e.g., virtual machines 752 and 754). Physical machines may also run separate service applications concurrently. Virtual machines or physical machines may be configured as individualized computing environments supported by resources 760 (e.g., hardware resources and software resources) within the cloud computing platform 710. Resources are intended to be configured for specific service applications. Furthermore, each service application may be divided into functional parts so that each functional part can run on a separate virtual machine. In the cloud computing platform 710, multiple servers may be used to run service applications and perform data storage processing in a cluster. In particular, these servers may perform data processing independently but are provided externally as a single device called a cluster. Each server in the cluster may be implemented as a node.

[0137] The client device 780 may be linked to a service application within the cloud computing platform 710. The client device 780 may be any type of computing device that corresponds to the computing device 700 described with reference to Figure 7, for example, the client device 780 may be configured to issue commands to the cloud computing platform 710. In embodiments, the client device 780 may communicate with the service application through virtual internet protocol (IP) and load balancers, or other means that direct communication requests to designated endpoints within the cloud computing platform 710. Components of the cloud computing platform 710 may communicate with each other via networks (not shown) which may include, but are not limited to, one or more local area networks (LANs) and / or wide area networks (WANs).

[0138] Exemplary computing environment Having briefly outlined the embodiments of the present invention, exemplary operating environments in which embodiments of the present invention can be carried out are described below to provide a general context for various aspects of the present invention. Referring first, in particular to Figure 8, an exemplary operating environment for carrying out embodiments of the present invention is shown and collectively designated as computing device 800. Computing device 800 is merely an example of a suitable computing environment and does not imply any limitation on the scope of use or functionality of the present invention. Furthermore, computing device 800 should not be construed as having any dependence or requirements on any one or combination of the illustrated components.

[0139] The present invention can be described in the general context of computer code or machine-usable instructions, including computer-executable instructions such as program modules, which are executed by computers or other machines, such as portable information terminals or other handheld devices. Generally, a program module, which includes routines, programs, objects, components, data structures, etc., refers to code that performs a task or implements a particular abstract data type. The present invention can be implemented in a variety of system configurations, including handheld devices, consumer electronics, general-purpose computers, and more specialized computing devices. The present invention can also be implemented in a distributed computing environment in which tasks are performed by remote processing devices linked via a communication network.

[0140] Referring to Figure 8, the computing device 800 includes a bus 810 that directly or indirectly connects memory 812, one or more processors 814, one or more presentation components 816, input / output ports 818, input / output component 820, and an exemplary power supply 822. Bus 810 represents what may be one or more buses (such as an address bus, a data bus, or a combination thereof). The various blocks in Figure 8 are shown with lines to clarify the concepts and also intend other configurations of the components and / or functions of the components described. For example, a presentation component such as a display device may be considered an I / O component. Also, a processor has memory. Recognizing that the nature of the art is such, it is reiterated that the figures in Figure 8 merely illustrate an exemplary computing device that may be used in connection with one or more embodiments of the present invention. Categories such as “workstation,” “server,” “laptop,” and “handheld device” are all within the scope of Figure 8 and are considered to be references to “computing devices,” so no distinction is made between such categories.

[0141] The computing device 800 typically includes various computer-readable media. Computer-readable media can be any available media that can be accessed by the computing device 800, and include both volatile and non-volatile media, and removable and non-removable media. Examples, but not limited to, computer-readable media may include computer storage media and communication media.

[0142] Computer storage media include volatile and non-volatile, removable and non-removable media implemented in any way or technique for storing information such as computer-readable instructions, data structures, program modules, or other data. Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disk (DVD) or other optical disk storage devices, magnetic cassettes, magnetic tapes, magnetic disk storage devices or other magnetic storage devices, or any other media that can be used to store desired information and are accessible by the computing device 800. Computer storage media exclude signals themselves.

[0143] Communication media typically include any information distribution medium that carries computer-readable instructions, data structures, program modules, or other data in a modulated data signal, such as a carrier wave or other transport mechanism. The term “modulated data signal” means a signal in which one or more characteristics are set or modified to encode information within the signal. By example, but not limited to, communication media include wired media such as wired networks or direct wired connections, as well as wireless media such as acoustic, RF, infrared, and other wireless media. Any combination of the above should also be included within the scope of computer-readable media.

[0144] Memory 812 includes computer storage media in the form of volatile memory and / or non-volatile memory. Memory may be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid-state memory, hard drives, optical disc drives, etc. Computing device 800 includes one or more processors that read data from various entities such as memory 812 or I / O component 820. Presentation component 816 presents data instructions to a user or other device. Exemplary presentation components include display devices, speakers, printing components, vibration components, etc.

[0145] I / O port 818 allows computing device 800 to be logically connected to other devices, including I / O components 820, some of which may be built-in. Exemplary components include microphones, joysticks, gamepads, satellite antennas, scanners, printers, and wireless devices.

[0146] Additional structural and functional features of the embodiment of the technical solution While various components used in this specification have been identified, it should be understood that any number of components and configurations may be employed to achieve the desired functionality within the scope of this disclosure. For example, components of embodiments shown in the drawings are indicated by lines for conceptual clarity. Other arrangements of these and other components may also be implemented. For example, while some components are shown as single components, many of the elements described herein may be implemented as individual components, distributed components, or in combination with other components, and in any suitable combination and arrangement. Some elements may be omitted entirely. Furthermore, various functions described herein as being performed by one or more entities may be performed by hardware, firmware, and / or software, as described below. For example, various functions may be performed by a processor that executes instructions stored in memory. Thus, other configurations and elements (e.g., machines, interfaces, functions, sequences, and groupings of functions) may be used in addition to or instead of those illustrated.

[0147] The embodiments described in the following paragraphs may be combined with one or more alternative embodiments described in detail. In particular, the claimed embodiments may alternatively include references to two or more other embodiments. The claimed embodiments may specify further limitations of the claimed subject matter.

[0148] The subject matter of embodiments of the present invention is specifically described herein in order to satisfy legal requirements. However, the description itself is not intended to limit the scope of this patent. Rather, the inventors intend that the claimed subject matter may also be embodied in other ways, in combination with other current or future technologies, to include different steps or combinations of similar steps than those described herein. Furthermore, the terms “step” and / or “block” may be used herein to indicate different elements of the method employed, but the terms should not be construed as implying any particular order among the various steps disclosed herein, unless the order of the individual steps is explicitly described.

[0149] For the purposes of this disclosure, the word “including” has the same broad meaning as the word “comprising,” and the word “accessing” encompasses “receiving,” “referencing,” or “retrieving.” Furthermore, the word “communicating” has the same broad meaning as the words “receiving” or “transmitting,” which are made possible by software or hardware-based buses, receivers, or transmitters using the communication medium described herein. In addition, words such as “a” and “an” include both singular and plural forms unless otherwise indicated. Thus, for example, the constraint “a feature” is satisfied if there is one or more features. Also, the term “or” includes conjunctive, disjunctive, and both (thus a or b includes either a or b, and both a and b).

[0150] For the purposes of the above detailed description, embodiments of the present invention will be described with reference to a distributed computing environment, but the distributed computing environment described herein is merely illustrative. Components may be configured to perform novel aspects of the embodiments, and the term “configured to ~” may mean “programmed to ~” to perform a particular task or to implement a particular abstract data type using code. Furthermore, embodiments of the present invention may generally refer to the environments and schematic diagrams of the technical solutions described herein, but it should be understood that the described techniques may be extended to other implementations.

[0151] Embodiments of the present invention have been described in relation to specific embodiments, which are intended to be illustrative rather than restrictive in any respect. Alternative embodiments will become apparent to those skilled in the art to which the invention pertains without departing from the scope of the invention.

[0152] From the above description, it will be understood that the present invention is suitable and clear for achieving all the above-mentioned objectives and effects, and that it has other advantages inherent in its structure.

[0153] It will be understood that certain features and partial combinations are useful and can be used without reference to other features or partial combinations. This is intended and within the scope of the claims.

Claims

1. A computerized system, One or more computer processors, The system includes a computer memory that stores computer-enabled instructions that cause the one or more computer processors to perform an operation when used by the one or more computer processors, and the operation is as follows: Historical ad design associated with multiple historical ad campaigns – accessing performance data, The process involves generating an advertising optimization model based on the aforementioned historical advertising design-performance data, wherein the advertising optimization model is generated based on modeling the relationships between user groups, preference characteristics, and preference associations. Accessing advertising campaigns, Using the advertising optimization model and the advertising campaign, generate an advertising optimization profile for the advertising campaign. Using a design requirements generator, generate fine-tuning prompts based on the ad optimization profile and the ad campaign. Using the design fine-tuning model and the fine-tuning prompts, generate updated ad campaigns. A system that includes delivering the aforementioned updated advertising campaign.

2. The system according to claim 1, wherein the historical ad design-performance data includes ad performance metrics and user feedback data, and generating the ad optimization model is based on offline processing including data analysis for user segmentation and preference insights associated with the historical ad design-performance data.

3. The system according to claim 1, wherein the preference features include an advertising image manipulation factor and an advertising layout and content adjustment factor.

4. The preference association system according to claim 1, which is based on a hybrid association framework involving manual labeling and data-driven rule mining or machine learning.

5. The system according to claim 1, wherein the ad optimization model operates to generate conditional text templates associated with ad layout and content adjustment factors or ad image manipulation factors, which are associated with ad optimization profiles corresponding to ad campaigns.

6. The system according to claim 1, wherein the design requirements generator is a large-scale language model that generates tweak prompts for a design tweak model using a multimodal layout tweak prompt template associated with an advertising optimization profile, the multimodal layout tweak prompt template is a structured data format that defines key predetermined elements for generating tweak prompts.

7. The system according to claim 1, wherein the design fine-tuning model is a multimodal layout generator that generates the advertising campaign according to the fine-tuning prompt.

8. The system according to claim 1, wherein generating the updated advertising campaign is further based on user-selected preferences for updating the advertising campaign.

9. The aforementioned operation is, Access to ad design and performance data for the aforementioned updated ad campaign, Using the ad design-performance data, ad optimization model, and updated ad campaign related to the updated ad campaign, generate a subsequent ad optimization profile for the updated ad campaign. Using the design requirements generator, generate subsequent fine-tuning prompts based on the subsequent ad optimization profile and the updated ad campaign. Using the aforementioned design fine-tuning model and the subsequent fine-tuning prompts, generate subsequent updated ad campaigns. The system according to claim 1, further comprising delivering the subsequent updated advertising campaign.

10. The system according to claim 1, wherein the advertising optimization model, the design requirements generator, and the design fine-tuning model define an advertising campaign management engine related to real-time performance metrics, design optimization, and auto-adjustment.

11. A computer storage medium in which a computer executable instruction causing an operation to be performed by a computing system having a processor and memory is embodied, wherein the operation is: Accessing advertising campaigns, Using an ad optimization model and the ad campaign, generate an ad optimization profile for the ad campaign, wherein the ad optimization profile is associated with a user. Using a design requirements generator, generate fine-tuning prompts based on the ad optimization profile and the ad campaign. Using the design fine-tuning model and the fine-tuning prompts, generate updated ad campaigns. A medium that includes delivering the updated advertising campaign to present to the user associated with the advertising optimization profile.

12. The medium according to claim 11, wherein the ad design-performance data includes ad performance metrics and user feedback data, and the generation of the ad optimization model is based on offline processing including data analysis for user segmentation and preference insights associated with the ad design-performance data.

13. The media according to claim 11, wherein the ad optimization profile is associated with a user profile and the user's attributes, and the user profile and the user's attributes are used to adapt ad layout and content adjustment factors and ad image manipulation factors to generate the updated ad campaign.

14. The medium according to claim 11, wherein the preference features include an advertising image manipulation factor and an advertising layout and content adjustment factor.

15. The medium according to claim 11, wherein the ad optimization model operates to generate conditional text templates related to ad layouts and content adjustment factors or ad image manipulation factors associated with ad optimization profiles corresponding to ad campaigns.

16. A method by which a computer performs an action. Steps to access an advertising campaign, The steps include generating an ad optimization profile for the ad campaign using the ad optimization model and the ad campaign, A step of using a design requirements generator to generate fine-tuning prompts based on the ad optimization profile and the ad campaign, The steps include generating an updated ad campaign using the design fine-tuning model and the fine-tuning prompt, A method including the steps of delivering the aforementioned updated advertising campaign.

17. The method according to claim 16, wherein the ad design-performance data includes ad performance metrics and user feedback data, and the ad optimization model is generated based on offline processing including data analysis for user segmentation and preference insights associated with the ad design-performance data.

18. The method according to claim 16, wherein the ad optimization profile is associated with a user profile of a user or a target user group, and the user profile is used to adapt ad layout and content adjustment factors and ad image manipulation factors to generate the updated ad campaign.

19. The method according to claim 16, wherein the preference features include an ad image manipulation factor and an ad layout and content adjustment factor.

20. The method according to claim 16, wherein the ad optimization model operates to generate conditional text templates related to ad layouts and content adjustment factors or ad image manipulation factors associated with ad optimization profiles corresponding to ad campaigns.

Images