An advertisement material combination optimization method

Abstract

The application relates to an advertisement material combination optimization method, which comprises the following steps: obtaining display constraint information corresponding to multiple advertisement positions based on advertisement task information; performing feature analysis processing on each advertisement material unit to generate corresponding material feature parameters; performing cross-advertisement position combination generation processing on the advertisement material unit to generate multiple advertisement material combinations; determining combination level evaluation parameters of each advertisement material combination respectively; comparing and sorting the multiple advertisement material combinations to screen out an advertisement material combination with the optimal evaluation result as a target material combination; and performing configuration output on the multiple advertisement positions and executing advertisement delivery. The advertisement delivery process is changed from the traditional single-point material optimization to the combination optimization process of multiple advertisement position linkage, which can not only ensure that the content expression between different advertisement positions is consistent and coherent, but also can make the material keep good adaptation effect under different display environments, and significantly improve the overall quality and conversion effect of advertisement delivery.

Description

Technical Field

[0001] This invention relates to the technical field of advertising creative combination optimization, and in particular to a method for advertising creative combination optimization. Background Technology

[0002] Currently, in actual advertising campaigns, the same advertising task usually needs to be displayed simultaneously in multiple different types of ad placements, such as homepage banners, feed ads, and recommendation ads. Different ad placements differ in display size, content structure, and user interaction methods. At the same time, each ad placement needs to maintain content consistency and synergistic campaign performance.

[0003] However, existing technologies typically select and optimize materials for a single ad placement, lacking overall modeling of the material combination relationship between multiple ad placements, making it difficult to guarantee content consistency and synergistic effects across ad placements. At the same time, existing solutions often screen materials based on a single metric, failing to uniformly evaluate and compare the overall effect of different material combinations, making it difficult to determine the optimal cross-ad placement material combination solution. Summary of the Invention

[0004] To address the problem that existing technologies cannot achieve overall optimization of ad creative combinations in scenarios involving multiple ad placements, this application provides an ad creative combination optimization method.

[0005] An advertising creative combination optimization method, comprising: Obtain advertising task information and candidate ad creative pool. The candidate ad creative pool contains multiple ad creative units. Based on the advertising task information, obtain the display constraint information for each ad slot. Each advertising creative unit undergoes feature parsing to generate corresponding creative feature parameters. The creative feature parameters include at least content semantic feature parameters, presentation feature parameters, and historical effect feature parameters. Based on the material feature parameters and display constraint information, the advertising material unit is processed to generate multiple advertising material combinations across advertising positions. Determine the combination-level evaluation parameters for each ad creative combination. The combination-level evaluation parameters should include at least cross-ad placement consistency parameters, display adaptation parameters, and cross-ad placement collaboration parameters. Based on combined-level evaluation parameters, multiple ad creative combinations are compared and ranked to select the ad creative combination with the best evaluation results as the target ad creative combination; Based on the target creative combination, configure and output across multiple ad placements, and execute ad delivery.

[0006] By adopting the above technical solution, advertising task information, material feature parameters, and display constraint information are uniformly integrated and combined for generation and comprehensive evaluation at the cross-ad placement level. This makes advertising materials no longer a single-point optimization, but rather a collaborative configuration relationship between multiple ad placements. As a result, while ensuring the adaptation of individual ad placements, the overall display consistency and the effectiveness of the campaign can be improved.

[0007] Preferably, the step of generating multiple ad creative combinations by performing cross-ad slot combination processing on ad creative units based on creative feature parameters and display constraint information includes: Based on the material feature parameters, each advertising material unit is classified to generate multiple material category sets; Based on the display constraint information, calculate the adaptation result value of each material category set corresponding to each ad slot; Based on the adaptation results, a selection process is performed to generate material combination units corresponding to each ad slot; Based on the material combination units of each ad position, cross-ad position combination construction processing is performed to generate multiple ad material combinations, where each ad material combination includes at least the material combination units corresponding to each ad position.

[0008] By adopting the above technical solution, and by introducing a classification and adaptation result value calculation mechanism based on material feature parameters, different types of materials can be precisely matched with the display constraints of each ad slot. Based on the matching, corresponding material combination units are constructed, thereby improving the matching accuracy between materials and ad slots and providing high-quality input for subsequent cross-ad slot combinations.

[0009] The preferred step of determining the combination-level evaluation parameters for each ad creative combination includes: Based on the content correspondence between each material combination unit, consistency calculation is performed to generate corresponding cross-ad slot consistency parameters; Based on the display constraint information, the values ​​of the adaptation results are mapped to generate corresponding display adaptation parameters; Based on the combination and distribution relationship of each material combination unit in each ad position, collaborative calculation processing is performed to generate corresponding cross-position collaborative parameters; Based on the integration of cross-ad slot consistency parameters, display adaptation parameters, and cross-slot collaboration parameters, corresponding combined-level evaluation parameters are generated.

[0010] By adopting the above technical solution, and by jointly modeling the content correspondence, display adaptation and cross-position distribution relationship between material combination units, a comprehensive quantitative evaluation of the consistency, adaptability and synergy of advertising material combinations can be achieved, thereby enabling more accurate selection of material combinations with the best overall effect.

[0011] Preferably, the step of generating corresponding cross-position collaborative parameters by performing collaborative calculations based on the combination distribution relationship of each material combination unit in each ad position includes: Determine the corresponding unit arrangement sequence for each ad slot based on the preset ad slot order, and construct the corresponding cross-position distribution sequence based on the unit arrangement sequence; Based on the cross-position distribution sequence, the combination distribution relationship of material combination units between adjacent ad positions is statistically analyzed. The combination distribution relationship includes at least the number of corresponding categories, the number of category jumps, and the number of category repetitions. The corresponding consistent distribution value is determined based on the ratio of the number of corresponding categories to the total number of preset adjacent position pairs; the corresponding jump distribution value is determined based on the ratio of the number of category jumps to the total number of preset adjacent position pairs; and the corresponding repeat distribution value is determined based on the ratio of the number of category repetitions to the length of the cross-position distribution sequence. Collaborative scoring is calculated based on consistent distribution values, jump distribution values, and repeated distribution values ​​to generate corresponding cross-position collaborative scores. The cross-position collaboration score is normalized to generate the corresponding cross-position collaboration parameters.

[0012] By adopting the above technical solution, and by constructing a cross-position distribution sequence and introducing a statistical analysis mechanism based on distribution characteristics such as category correspondence, jumps, and repetitions, the collaborative relationship of cross-ad slot combinations can be characterized in a structured way. A unified collaborative index is formed through scoring calculation, thereby effectively improving the overall coherence and rationality of the combination in multi-ad slot scenarios.

[0013] Preferably, the step of constructing cross-advertisement creative combinations based on the creative combination units of each ad placement to generate multiple ad creative combinations, wherein each ad creative combination includes at least the creative combination units corresponding to each ad placement, includes: Based on advertising task information, obtain the correlation features between different advertising positions; Based on the association features, determine the combination constraints between the various ad slots; Based on the combination constraints, the various material combination units are combined by minimizing the conditions to generate multiple corresponding advertising material combinations.

[0014] By adopting the above technical solution, and by introducing cross-ad placement association features and establishing combination constraint relationships based on these features, the process of generating ad creative combinations is transformed from simple splicing into a constrained structured construction process. This allows for control of the expression relationships between different ad placements during the generation stage, thereby improving the overall rationality and synergy of the combination.

[0015] Preferably, the step of constructing multiple advertising creative combinations by minimizing the conditions of each creative combination unit according to the combination constraint relationship includes: Based on the combination constraint relationship, determine the target expression role of each ad position in the cross ad position combination, as well as the visual distribution relationship and historical effect inheritance relationship. The target expression role includes at least the main expression role, the inheriting expression role and the conversion expression role. Based on the correspondence between each material combination unit and each target expression role, calculate the role deviation value of each material combination unit in each ad position; Based on visual distribution relationships, calculate the visual progression deviation value of each material combination unit in each ad position; Based on the historical performance continuity relationship, calculate the performance continuity deviation value of each material combination unit in each ad position; Based on the deviation values ​​of the character, the visual progression, and the effect succession, conditional aggregation is performed to generate corresponding combined conditional deviation values. The combination construction process is performed based on the deviation value of the combination conditions to minimize the combination construction value and generate multiple ad creative combinations.

[0016] By adopting the above technical solution, and by introducing target expression roles, visual distribution and effect acceptance constraints during the combination construction process, and by performing condition minimization processing based on deviation values, the functional division of materials in each ad position is made clearer, while ensuring the coherence of visual presentation and effect delivery, thereby significantly improving the overall expression effect of cross-ad position combinations.

[0017] Preferably, the step of classifying each advertising creative unit based on creative feature parameters to generate multiple creative category sets includes: Based on the material feature parameters, adaptive weights, content semantic feature parameters, presentation feature parameters, and historical effect feature parameters are determined. Based on the semantic feature parameters of the content and the adaptive weights determined according to the feature parameters of the creative materials, semantic classification processing is performed on each advertising creative unit to generate the corresponding initial category affiliation; Based on the performance feature parameters, the initial category assignment is corrected to generate the corresponding intermediate category assignment. Based on historical performance feature parameters, the intermediate category assignments are further refined to generate the corresponding final category assignments. Based on the final category assignment of each advertising creative unit, a corresponding set of creative categories is generated.

[0018] By adopting the above technical solution, a multi-stage classification mechanism is constructed and adaptive weights are introduced to dynamically adjust semantic features. At the same time, multiple rounds of correction are carried out by combining performance features and historical effect features, so that the material classification results can take into account semantic expression, visual presentation and historical effect performance, thereby improving classification accuracy and subsequent combination quality.

[0019] Preferably, the step of performing feature parsing processing on each advertising creative unit to generate corresponding creative feature parameters includes: Perform content parsing on each advertising creative unit to extract text and image information from each unit; The text information is segmented into words to generate corresponding word sequences; Determine the degree value of the term sequence and generate corresponding adaptive weights; The word sequence is processed by word frequency statistics and semantic vector mapping to generate corresponding content semantic feature parameters, and the adaptive weights and content semantic feature parameters are associated and bound together. Visual feature extraction and processing are performed based on image information to generate corresponding representational feature parameters; Obtain historical delivery data for each ad creative unit, and perform statistical analysis based on the historical delivery data to generate corresponding historical performance characteristic parameters; Based on the content semantic feature parameters, the presentation feature parameters, and the historical effect feature parameters, the corresponding material feature parameters are generated by parameter integration.

[0020] By adopting the above technical solution, multi-dimensional material feature parameters are generated by jointly analyzing text and image information and combining word sequence, adaptive weight and semantic vector mapping, so that the semantic, visual and effect information of the material can be fully characterized, thus providing a reliable data foundation for subsequent classification, combination and evaluation.

[0021] Preferably, the step of obtaining display constraint information for multiple ad slots based on ad task information includes: Based on the advertising task information, multiple ad slots are identified, along with the identification information for each ad slot; Based on the identification information, determine the corresponding display specifications, which include display size parameters, layout structure parameters, and content length limit parameters. Based on the identification information, the corresponding environmental location information is determined. The environmental location information includes terminal device type parameters, geographic location parameters, and network environment parameters. The display specifications and environmental location information are correlated and mapped to generate corresponding constraint parameters. These constraint parameters are then integrated to generate corresponding display constraint information.

[0022] By adopting the above technical solution, and by parsing the advertising space identification information and combining it with the display specification parameters and environmental location information to form a unified model, the display requirements of the advertising space can be expressed in a structured form, thereby improving the accuracy of the description of the adaptation relationship between the material and the advertising space.

[0023] Preferably, the step of associating and mapping the displayed specification parameters and environmental location information to generate corresponding constraint parameters includes: Based on the terminal device type parameter, the display size parameter is adapted to the terminal to generate the terminal-adapted display size parameter. Based on geographic location parameters, the layout structure parameters are adapted to the region to generate the region-adapted layout structure parameters. Based on network environment parameters, the content length limit parameters are adapted to network conditions to generate network-adapted content length limit parameters. Constraint parameters are generated by combining and processing the display size parameters after terminal adaptation, the layout structure parameters after regional adaptation, and the content length limit parameters after network adaptation.

[0024] By adopting the above technical solutions and introducing terminal adaptation, regional adaptation and network adaptation mechanisms in the process of generating display constraints, the display needs of different ad positions in different terminals, different regions and different network environments can be effectively matched, thereby improving the adaptability and stability of advertising materials in the actual delivery environment.

[0025] In summary, this application includes at least one of the following beneficial technical effects: The process of selecting creative materials, originally focused on a single ad placement, has been restructured into a holistic optimization process centered on cross-ad placement combinations. By modeling the multi-dimensional features of ad creatives and employing a constraint-driven combination generation mechanism, creative selection has shifted from isolated decision-making to collaborative decision-making. Specifically, firstly, by introducing ad task information and constructing display constraint information for multiple ad placements, the differences in size, structure, and environmental conditions among different ad placements are transformed into a unified constraint expression, providing clear boundary conditions for subsequent combinations. Based on this, candidate ad creatives undergo multi-dimensional feature analysis of content semantics, presentation style, and historical performance. Feature integration forms a unified creative feature expression, ensuring that each creative not only reflects its content attributes but also its visual performance capabilities and historical campaign performance, providing comprehensive data support for subsequent combination decisions. Furthermore, in the combination generation stage, instead of simply splicing creatives together, cross-ad placement creative combinations are constructed based on the matching relationship between the aforementioned creative features and display constraints. This allows creatives on different ad placements to form a structured combination relationship while meeting their respective display requirements, thus avoiding the problem of fragmented creative expression between different ad placements. Building upon this foundation, a combined evaluation mechanism is introduced. This mechanism evaluates each creative combination from a holistic perspective, considering multiple ad placements as a whole. This evaluation not only considers the content relevance between different ad placements but also the compatibility between the creatives and ad placements, as well as the synergistic performance of the combination within the overall campaign. By comparing and ranking multiple combinations, the optimal creative combination is determined. This transforms the advertising process from traditional single-point creative optimization to a multi-ad placement coordinated optimization process. It ensures consistency and coherence in content expression across different ad placements, maintains good adaptability of creatives in various display environments, and enhances the overall performance of the combined strategy through holistic collaborative evaluation, thereby significantly improving the overall quality and conversion rate of advertising campaigns. Attached Figure Description

[0026] Figure 1 This is a flowchart of an advertising material combination optimization method according to an embodiment of this application. Detailed Implementation

[0027] The present application will be further described in detail below with reference to the accompanying drawings.

[0028] In one embodiment, such as Figure 1 As shown, this application discloses a method for optimizing advertising creative combinations. The method includes: S10. Obtain advertising task information and candidate advertising material pool. The candidate advertising material pool contains multiple advertising material units. Obtain the display constraint information of the corresponding multiple advertising positions based on the advertising task information. S20. Perform feature parsing processing on each advertising material unit to generate corresponding material feature parameters. The material feature parameters shall include at least content semantic feature parameters, presentation feature parameters and historical effect feature parameters. S30. Based on the material feature parameters and display constraint information, perform cross-ad slot combination generation processing on the advertising material unit to generate multiple advertising material combinations; S40. Determine the combination-level evaluation parameters for each ad creative combination. The combination-level evaluation parameters shall include at least cross-ad placement consistency parameters, display adaptation parameters, and cross-ad placement collaboration parameters. S50. Based on the combined-level evaluation parameters, compare and rank multiple ad creative combinations to select the ad creative combination with the best evaluation results as the target ad creative combination. S60: Based on the target material combination, configure and output across multiple ad placements, and execute ad delivery.

[0029] In this embodiment, advertising task information is used to characterize the overall goals and constraints of the current advertising campaign. It can be stored and transmitted in the system as structured data, specifically including the target type, target user range, campaign time interval, and budget control parameters, serving as the input basis for subsequent processing steps. The candidate advertising creative pool stores all available advertising creative resources. In engineering implementation, it typically corresponds to a creative database or cache set. Each advertising creative unit exists as the smallest processing unit, consisting of image content, text description, and corresponding creative identifier, and can be individually invoked and combined. The advertising slot represents the specific location of the advertisement display. In the system, it is usually distinguished by advertising slot identifier information and associated with attributes such as display size, layout structure, and terminal environment. Display constraint information is the result of a unified modeling of the advertising slot display requirements, which can consist of display size range, content structure limitations, and environmental adaptation conditions, and participates in subsequent calculations in a parameterized form.

[0030] Creative feature parameters are used to uniformly represent the features of advertising creative units. In engineering implementation, they can be represented by multi-dimensional vectors or parameter sets. Among them, content semantic feature parameters are used to characterize the semantic information of the text content, such as generating semantic expressions through word segmentation and vector mapping; presentation feature parameters are used to describe the visual presentation features of the creative, such as image size, color distribution, and structural layout; historical performance feature parameters are used to reflect the performance of the creative in historical campaigns, such as statistical indicators like click-through rate and conversion rate. By integrating the above multi-dimensional features, a unified creative feature expression can be formed. Advertising creative combinations represent the result of combining multiple advertising creative units according to different ad placements. In the system, it can be represented as a structured object composed of multiple creative combination units, each corresponding to an ad placement and its selected creative content. Combination-level evaluation parameters are used to quantitatively evaluate the overall effectiveness of ad creative combinations. These include cross-ad placement consistency parameters, display adaptation parameters, and cross-ad placement collaboration parameters. Cross-ad placement consistency parameters characterize the degree of consistency in content expression among creatives in different ad placements. Display adaptation parameters characterize the degree of matching between creatives and the constraints of their corresponding ad placements. Cross-ad placement collaboration parameters characterize the collaborative relationship between creatives in different ad placements within the combined structure. The target creative combination is the optimal combination obtained after sorting and filtering multiple ad creative combinations; it serves as the final output in the actual ad delivery process.

[0031] For example, in an e-commerce promotion scenario, the system receives an advertising task for skincare products, which includes the promotional goals and target audience. The candidate ad creative pool stores multiple ad creative units containing product images and text. The system generates corresponding display constraints based on the display requirements of the homepage banner, feed ad, and recommendation slots, and extracts semantic features, visual features, and historical performance features for each creative unit. Based on this, multiple ad creative combinations are generated, each corresponding to a different ad slot's creative configuration scheme. By calculating the evaluation parameters of each combination in terms of content consistency, display adaptability, and cross-slot collaboration, the multiple combinations are ranked, and finally, the combination with the best overall performance is selected as the target creative combination.

[0032] In actual operation, the system first determines the delivery target and ad placement set based on the advertising task information, and transforms the display requirements of the ad placements into unified display constraint information. Then, it performs feature parsing processing on each ad creative unit in the candidate ad creative pool, converting the original creatives into calculable feature parameter representations. Based on this, it generates multiple cross-ad placement creative combinations according to the matching relationship between the creative features and display constraints, and performs quantitative analysis on each combination through combination-level evaluation parameters, thereby completing the sorting and optimization among multiple combinations. Finally, the selected target creative combinations are output to each ad placement for execution, thus realizing the optimization of ad creative combinations in multi-ad placement linkage scenarios.

[0033] Furthermore, the step of obtaining display constraint information for multiple ad slots based on ad task information includes: S101. Based on the advertising task information, determine multiple advertising slots and the identification information of each advertising slot; S102. Based on the identification information, determine the corresponding display specification parameters, including display size parameters, layout structure parameters, and content length limit parameters. S103. Based on the identification information, determine the corresponding environmental location information, which includes terminal device type parameters, geographic location parameters, and network environment parameters. S104. The display specifications and environmental location information are associated and mapped to generate corresponding constraint parameters. The various constraint parameters are then integrated to generate corresponding display constraint information.

[0034] Furthermore, the step of associating and mapping specification parameters and environmental location information to generate corresponding constraint parameters includes: S1041. Based on the terminal device type parameter, perform terminal adaptation processing on the display size parameter to generate the terminal-adapted display size parameter. S1042. Based on the geographic location parameters, perform regional adaptation processing on the layout structure parameters to generate regionally adapted layout structure parameters. S1043. Based on network environment parameters, perform network adaptation processing on the content length limit parameters to generate network-adapted content length limit parameters. S1045. Constraint parameters are generated by combining and processing the display size parameters after terminal adaptation, the layout structure parameters after regional adaptation, and the content length limit parameters after network adaptation.

[0035] In this embodiment, taking the example of an e-commerce platform launching a multi-ad placement campaign for a skincare product, the process of obtaining display constraint information for multiple ad placements based on the advertising task information is explained. First, the system obtains advertising task information, which includes advertising target parameters and corresponding target audience information. Based on the advertising task information, it determines the identification information for multiple ad placements, including at least homepage banner ad placements, feed ad placements, and recommendation ad placements. Then, based on each ad placement identification information, it obtains the display specification parameters corresponding to each ad placement. These display specification parameters include at least display size parameters, layout structure parameters, and content length limit parameters. For example, the display size parameter for the homepage banner ad placement is a large horizontal image size, the feed ad placement corresponds to a mixed text and image structure, and the recommendation ad placement corresponds to a small-sized image and price information display structure.

[0036] Based on the obtained display specification parameters, the environmental location information corresponding to each advertising space is further obtained according to the identification information of each advertising space. The environmental location information includes at least terminal device type parameters, geographical location parameters, and network environment parameters. For example, the terminal device type is obtained by user terminal identification as a mobile terminal or a personal computer, the geographical location is obtained as a different city area by location information, and the network environment is obtained as a wireless network or a mobile data network by network status detection. Subsequently, based on the environmental location information, the display specification parameters are corrected. Specifically, based on the terminal device type parameter, the display size parameter is adapted to the terminal; for example, when the terminal device type is a mobile terminal, the display size parameter is adjusted to a portrait aspect ratio, and when the terminal device type is a personal computer, the display size parameter is adjusted to a landscape aspect ratio. Based on the geographical location parameter, the layout structure parameter is adapted to the region; for example, a brand-oriented layout structure is generated in first-tier city areas, and a promotion-oriented layout structure is generated in non-first-tier city areas. Based on the network environment parameter, the content length limit parameter is adapted to the network; for example, video or dynamic content is allowed in a wireless network environment, while static images and concise text are restricted in a mobile data network environment.

[0037] After completing the aforementioned terminal adaptation, region adaptation, and network adaptation processes, the display size parameters adapted to the terminal, the layout structure parameters adapted to the region, and the content length limit parameters adapted to the network are combined and processed to generate constraint parameters corresponding to each ad slot. Based on these constraint parameters, the various ad slots are then constrained and integrated to generate display constraint information for the multiple ad slots. Through this process, the display constraint information not only reflects the display requirements of the ad slot itself but also incorporates the influence of terminal devices, geographical location, and network environment on the display format, thus providing a unified and executable constraint basis for subsequent ad creative combination generation and combination-level evaluation.

[0038] Specifically, taking a skincare essence product advertised across multiple ad placements on an e-commerce platform as an example, the process of modifying the display specifications based on the environmental location information is explained. The system first determines the corresponding ad placement identifiers based on the advertising task information, including the homepage banner ad placement, the feed ad placement, and the recommendation ad placement. It then obtains the display specifications parameters for each ad placement. Specifically, the initial display size of the homepage banner ad placement is set to 1920×1080 pixels, the layout structure is an overlay of text and images, and the content length is limited to no more than 30 characters. The initial display size of the feed ad placement is set to 1080×1080 pixels, the layout structure is a side-by-side text and images, and the content length is limited to no more than 40 characters. The initial display size of the recommendation ad placement is set to 600×600 pixels, the layout structure is an image + price tag structure, and the content length is limited to no more than 15 characters.

[0039] Subsequently, the system obtains the corresponding environmental location information based on the ad placement identification information. The terminal device type parameter identifies it as a mobile terminal, the geographical location parameter identifies it as a third-tier city area, and the network environment parameter detects it as a 4G mobile data network with an average bandwidth of 5Mbps and a network latency of 80ms. Based on this, the system performs terminal adaptation processing on the display size parameters according to the terminal device type parameters. The display size of the homepage banner ad placement is adjusted from 1920×1080 pixels to a vertical screen size of 1080×1920 pixels; the display size of the feed ad placement is adjusted from 1080×1080 pixels to 720×720 pixels; and the display size of the recommended ad placement is adjusted from 600×600 pixels to 400×400 pixels, in order to reduce the terminal rendering load and adapt to the mobile display ratio.

[0040] Furthermore, based on the geographic location parameters, the system performs regional adaptation processing on the layout structure parameters. The layout structure parameters of the homepage banner ad slot are adjusted from an image and text overlay structure to a promotional structure that highlights price information. The layout structure parameters of the information flow ad slot are adjusted from an image and text side-by-side structure to an image-dominated + price tag structure. The layout structure parameters of the recommended ad slot remain an image + price tag structure but a discount indicator area is added, thereby matching the layout structure with the price-sensitive characteristics of users in third-tier cities.

[0041] Based on this, the system performs network adaptation processing on the content length limit parameters according to the network environment parameters. The content length limit parameter for the homepage banner ad slot is adjusted from no more than 30 characters to no more than 15 characters, the content length limit parameter for the information flow ad slot is adjusted from no more than 40 characters to no more than 20 characters, and the content length limit parameter for the recommendation ad slot is adjusted from no more than 15 characters to no more than 10 characters. At the same time, the material type is restricted from video material to static image material to ensure that the page loading time is controlled within 2 seconds under the condition of 5Mbps bandwidth.

[0042] After completing the terminal adaptation, region adaptation, and network adaptation processes, the system combines the terminal-adapted display size parameters, the region-adapted layout structure parameters, and the network-adapted content length limit parameters to generate constraint parameters corresponding to the homepage banner ad slot, the news feed ad slot, and the recommendation ad slot, respectively. The system then integrates these constraint parameters to generate display constraint information for the multiple ad slots. This display constraint information is used to limit the display format and content structure of each ad creative unit in different ad slots during the subsequent ad creative combination generation process, thereby achieving adaptive adjustment of display constraints based on environmental location information.

[0043] Furthermore, the step of performing feature parsing processing on each advertising creative unit to generate corresponding creative feature parameters includes: S201. Perform content parsing processing on each advertising material unit to extract text and image information from each advertising material unit; S202. Perform semantic parsing processing based on text information to generate corresponding content semantic feature parameters; S203. Perform visual feature extraction processing based on image information to generate corresponding representation feature parameters; S204. Obtain the historical delivery data corresponding to each advertising material unit, and perform statistical analysis and processing based on the historical delivery data to generate the corresponding historical performance feature parameters. S205. Based on the content semantic feature parameters, the presentation feature parameters, and the historical effect feature parameters, parameter integration processing is performed to generate the corresponding material feature parameters.

[0044] Furthermore, the step of performing semantic parsing processing based on text information to generate corresponding content semantic feature parameters includes: S2021. Perform word segmentation on the text information to generate the corresponding word sequence; S2022. Determine the degree value of the term sequence and generate corresponding adaptive weights; S2023. Perform word frequency statistical processing based on the word sequence to generate corresponding word frequency feature parameters; S2024. Perform semantic vector mapping processing based on the term sequence to generate corresponding semantic vector feature parameters; S2025. Based on the fusion processing of word frequency feature parameters and semantic vector feature parameters, corresponding content semantic feature parameters are generated, and adaptive weights are associated and bound with the content semantic feature parameters.

[0045] Based on the aforementioned embodiment of multi-slot coordinated advertising of skincare essences on e-commerce platforms, the process of performing feature analysis on each of the aforementioned advertising material units to generate corresponding material feature parameters is described. The system first obtains multiple advertising material units from a candidate advertising material pool. Each advertising material unit contains at least one set of image information and corresponding text information. For example, one advertising material unit may contain a product display image with a resolution of 1080×1920 pixels and the corresponding text information: "High-purity niacinamide essence, brightens skin tone, buy one get one free." Subsequently, content analysis is performed on each of the advertising material units to extract text and image information. The text information is parsed using character encoding, and the image information is read in pixel matrix form.

[0046] After acquiring the text information, the system performs semantic parsing processing based on the text information. First, the text information is segmented into words. For example, "high-purity niacinamide essence, brightening skin tone, buy one get one free" is decomposed into the word sequence "high-purity niacinamide essence brightening skin tone buy one get one free". Based on the word sequence, word frequency statistics are performed to obtain word frequency feature parameters, where the word frequency values ​​of keywords such as "niacinamide brightening" are 1. Simultaneously, weight mapping is performed on the word items based on a preset keyword dictionary. For example, the promotion weight corresponding to "buy one get one free" is set to 0.9, and the quality weight corresponding to "high purity" is set to 0.8. Further, semantic vector mapping processing is performed based on the word sequence, mapping each word item to a fixed-dimensional semantic vector. For example, each word item is mapped to a 128-dimensional semantic vector. The semantic vector feature parameters corresponding to the text information are generated through vector superposition. In a specific implementation, a 128-dimensional semantic vector can be obtained, where each dimension takes the value of a floating-point number between 0 and 1. Subsequently, the word frequency feature parameters and the semantic vector feature parameters are fused together. For example, a weighted fusion method is used, where the word frequency feature parameters are normalized with a weight coefficient of 0.4, and the semantic vector feature parameters are fused with a weight coefficient of 0.6, thereby generating the content semantic feature parameters.

[0047] After completing the semantic parsing of the text information, the system performs visual feature extraction processing based on the image information. For example, it performs color distribution analysis and texture feature extraction processing on the image information. In a specific implementation, by statistically analyzing the RGB channels of the image, the average color values ​​are obtained as R=180, G=160, and B=150. At the same time, a 256-dimensional image feature vector is generated through convolutional feature extraction, which serves as the feature parameter of the presentation form to characterize the visual style and presentation form of the advertising material unit.

[0048] Furthermore, the system acquires historical delivery data corresponding to each of the aforementioned ad creative units. For example, the ad creative unit has 100,000 impressions, 8,000 clicks, and 1,200 conversions in the past 7 days. Based on the historical delivery data, the system performs statistical analysis and processing to calculate a click-through rate (CTR) of 8% and a conversion rate (CVR) of 15%. The system then normalizes the above indicators to generate corresponding historical performance characteristic parameters.

[0049] After obtaining the content semantic feature parameters, the presentation feature parameters, and the historical effect feature parameters, the system performs parameter integration processing on the above parameters. For example, it splices and weights the 128-dimensional semantic vector, the 256-dimensional visual feature vector, and the normalized effect index parameters to generate a material feature parameter vector with a comprehensive dimension of 385. This vector is used to characterize the comprehensive characteristics of the advertising material unit in terms of semantic content, visual presentation, and historical effect, thereby providing a unified data foundation for subsequent cross-ad slot combination generation processing and combination-level evaluation.

[0050] Furthermore, based on the material feature parameters and display constraint information, the process of generating multiple ad creative combinations across ad placements includes: S301. Based on the material feature parameters, classify each advertising material unit to generate multiple material category sets; S302. Based on the display constraint information, filter and process the collection of each material category to generate material combination units corresponding to each ad slot; S303. Based on the material combination units of each ad position, perform cross-ad position combination construction processing to generate multiple ad material combinations, wherein each ad material combination includes at least the material combination units corresponding to each ad position.

[0051] Furthermore, the step of classifying each advertising creative unit based on the creative feature parameters to generate multiple creative category sets includes: S3011. Based on the content semantic feature parameters and adaptive weights, perform semantic classification processing on each advertising material unit to generate the corresponding initial category affiliation; S3012. Based on the performance feature parameters, the initial category assignment is corrected to generate the corresponding intermediate category assignment; S3013. Based on historical effect feature parameters, the intermediate category assignment is further corrected to generate the corresponding final category assignment. S3014. Based on the final category affiliation of each advertising material unit, generate the corresponding material category set.

[0052] Furthermore, the step of filtering and processing each set of material categories based on the display constraint information to generate material combination units corresponding to each ad slot includes: S3021. Based on the display constraint information, calculate the adaptation result value of each material category set corresponding to each ad slot; S3022. Based on the adaptation result value, perform filtering and processing to generate material combination units corresponding to each ad position.

[0053] In this embodiment, the material category set represents the grouping result after the advertising material units have been classified. In engineering implementation, it can be organized using a category index structure. Each material category set contains multiple advertising material units with the same or similar category affiliation, used for subsequent filtering and combination by category. The initial category affiliation represents the category result obtained from the first classification of advertising material units based on content semantic feature parameters and adaptive weights. In implementation, it can be obtained by weighting the semantic features and matching them with a preset category model, reflecting the main expressive direction of the material at the semantic level. The intermediate category affiliation is the classification result after adjusting the initial category affiliation by introducing performance feature parameters. It corrects the classification result by matching parameters such as the visual structure and performance style of the material, ensuring consistency between content expression and visual presentation. The final category affiliation is the result after further correction based on the intermediate category affiliation by incorporating historical performance feature parameters. It further optimizes the classification result by introducing statistical indicators such as historical click-through rate and conversion rate, so that the classification not only reflects content and form but also reflects actual performance.

[0054] The material combination unit represents a set of materials or material instances selected after screening for a specific ad placement. In engineering implementation, it typically corresponds one-to-one with an ad placement and serves as a basic building block for constructing cross-ad placement combinations. The adaptation result value measures the degree of adaptation of a material category set to a specific ad placement. It is calculated by matching the material characteristics in the material category set with the display constraints corresponding to the ad placement. For example, a quantitative score is generated by comprehensively calculating the degree of size matching, structural conformity, and environmental adaptability. This score reflects the usability and adaptation level of the material category in that ad placement. The cross-ad placement combination construction process arranges and combines material combination units corresponding to different ad placements according to preset combination rules to generate a complete ad material combination. In implementation, this can be achieved by traversing the material combination units of different ad placements and combining them to generate multiple candidate combination schemes.

[0055] For example, in a campaign scenario including homepage banner ads, feed ads, and recommendation ads, the system first categorizes candidate ad creatives, resulting in categories such as efficacy, promotion, and brand. Each creative is initially categorized based on semantic features, then refined according to visual style, and finally further refined using historical click-through rates to arrive at its final category. Subsequently, based on the display constraints of each ad placement, the system calculates the fit of each creative category set for each ad placement. For instance, efficacy-related creatives show higher fit in feed ads but lower fit in banner ads. Therefore, the system selects efficacy-related creative combinations for feed ads, promotion-related creative combinations for banner ads, and brand-related creative combinations for recommendation ads. Finally, the system combines the creative combinations corresponding to each ad placement to generate multiple complete ad creative combination schemes.

[0056] In practice, the system first classifies advertising material units step by step through a multi-stage classification mechanism, so that the material categories can comprehensively reflect semantic, presentation and historical effect characteristics. Then, based on display constraint information, the system performs adaptation calculations on different categories of materials, matches the classification results with the needs of the ad placement, and selects material combination units suitable for each ad placement. On this basis, the system combines material units from different ad placements through cross-ad placement combination construction processing, and finally forms multiple candidate advertising material combinations, providing input for subsequent combination-level evaluation and ranking.

[0057] Specifically, the purpose of setting adaptive weights is to address the imbalance of information contribution in the expression of text semantic features. Since text information consists of multiple terms, the importance of different terms in semantic expression varies significantly. If each term is treated equally during semantic modeling, key expressions may be weakened or non-key expressions may be amplified, thus affecting the accuracy of semantic classification and subsequent decisions. Therefore, by judging the degree value of terms and generating adaptive weights, the semantic features can be dynamically adjusted according to the importance of terms during the construction process, thereby improving the accuracy of semantic expression. The performance feature parameters mainly come from image structure and visual attributes. They have already formed stable numerical expressions through a unified feature extraction process and do not have the weight imbalance problem caused by the difference in term granularity in text. The historical performance feature parameters are directly generated based on statistical data. Their values ​​themselves reflect the strength of the actual delivery performance and have natural weight attributes. Therefore, there is no need to introduce additional adaptive weights for adjustment, thus ensuring that the overall feature system achieves a balance between complexity and stability.

[0058] More specifically, the adaptation result value is essentially a quantitative result of the degree of matching between the set of material categories and the target ad placement. In engineering implementation, it is usually calculated by aligning the material-side features with the ad placement-side constraints. In the specific implementation process, the display constraint information is first decomposed into a set of computable constraint parameters, such as display size range, layout structure type, and content length limit. At the same time, the presentation feature parameters and content semantic feature parameters corresponding to each advertising material unit in the material category set are parameterized. Then, item-by-item matching calculations are performed within a unified parameter space. For size constraints, the matching degree can be calculated by size ratio difference or whether it falls within the allowable range. For layout structure, matching can be performed by structural type consistency or structural feature similarity. For content length limit, it can be calculated by the deviation between the text length and the limit threshold. After obtaining each matching result, each matching result is normalized to convert it into a matching score under the same dimension. Then, the matching scores are weighted and fused according to the preset weight coefficient to obtain the comprehensive matching score of the material category set in the corresponding ad position. This comprehensive matching score is used as the adaptation result value to reflect the overall adaptation degree of the material category in the ad position and serves as the basis for subsequent screening and combination generation.

[0059] Furthermore, the process of constructing cross-advertisement creative combinations based on the creative combination units of each ad placement to generate multiple ad creative combinations, wherein each ad creative combination includes at least the creative combination units corresponding to each ad placement, includes: S3031. Based on advertising task information, obtain the correlation features between cross-ad slots; S3032. Based on the aforementioned association features, determine the combination constraint relationships between each ad slot; S3033. Based on the combination constraint relationship, perform condition minimization combination construction on each of the material combination units to generate multiple corresponding advertising material combinations.

[0060] Furthermore, the step of constructing multiple corresponding advertising material combinations by performing condition minimization combinations on each of the material combination units according to the aforementioned combination constraint relationship includes: S30331. Based on the combination constraint relationship, determine the target expression role of each ad position in the cross-ad position combination, as well as the visual distribution relationship and historical effect inheritance relationship. The target expression role includes at least the main expression role, the inheriting expression role and the conversion expression role. S30332. Based on the correspondence between each material combination unit and each target expression role, calculate the role deviation value of each material combination unit in each ad position; S30333. Based on the visual distribution relationship, calculate the visual progressive deviation value of each of the material combination units in each advertisement position; S30334. Based on the historical effect continuity relationship, calculate the effect continuity deviation value of each of the material combination units in each ad position; S30335. Based on the character deviation value, the visual progression deviation value, and the effect continuation deviation value, perform condition aggregation processing to generate a corresponding combined conditional deviation value. S30336. Based on the deviation value of the combination conditions, perform a minimum combination construction process to generate the multiple advertising material combinations.

[0061] In this embodiment, advertising task information is used to characterize the overall goals and constraints of the current advertising campaign. In engineering implementation, it typically exists in the form of structured parameters, including target user group characteristics, conversion target types, campaign strategies, and display order requirements among multiple ad placements. This information is input into the composite construction module within the system through a unified data interface, serving as the basis for subsequent cross-ad placement relationship modeling. The correlation features between cross-ad placements describe the collaborative relationship among multiple ad placements under the same campaign task. It is not a single indicator but a comprehensive feature formed by analyzing the user path, exposure rhythm, and conversion logic involved in the advertising task information. In implementation, these correlation features can manifest as sequential dependencies, information continuity relationships, and expression intensity distribution relationships between ad placements, used to characterize the interaction between different ad placements in the overall campaign chain.

[0062] Combination constraint relationships are constraint models further abstracted from associated features. They are used to structurally restrict the process of combining ad placements. In engineering implementation, they typically exist as a multi-dimensional parameter set, which includes at least the primary and secondary relationships, visual distribution relationships, and historical effect inheritance relationships. The target expression role describes the functional positioning of each ad placement in the overall ad expression. In implementation, it can be obtained by parsing associated features and assigning roles. For example, ad placements at the beginning of the user exposure path are identified as those responsible for attracting user attention, middle ad placements are identified as those responsible for receiving and expanding information, and back-end ad placements are identified as those responsible for conversion guidance, thus forming a clear division of expression. Visual distribution relationships characterize the overall distribution of different ad placements in visual presentation. They originate from the expressive feature parameters in the material combination unit and are obtained after analysis and processing in conjunction with the ad placement order. For example, by comparing and analyzing the color intensity, compositional complexity, and visual focus of different ad placement materials, a visual layout pattern of strong at the beginning and weak at the end or evenly distributed can be formed. Historical performance continuity is used to characterize the connection between the historical performance of different ad placements. It is obtained by correlation analysis of the historical performance characteristic parameters of the creative combination unit. For example, by analyzing the changing trends of different ad placement creatives in indicators such as click-through rate and conversion rate, a performance continuity model is formed that guides clicks in the preceding stage and promotes conversion in the subsequent stage.

[0063] The material combination unit represents a collection of materials or candidate material instances obtained after adaptation and screening for a single ad slot. In engineering implementation, it typically exists as an index structure or data object and is associated with a corresponding ad slot. The role deviation value measures the degree of matching of the material combination unit when it assumes the target expressive role in a specified ad slot. It is calculated by comparing the semantic feature parameters of the material's content with the semantic requirements of the target expressive role; a smaller deviation value indicates a higher degree of matching. The visual progression deviation value measures whether the visual presentation of the material combination unit in different ad slots conforms to the preset visual distribution relationship. It is calculated by differentiating the distribution of the material combination unit's presentation feature parameters in the ad slot order, reflecting whether there are abrupt changes or discontinuities in the visual presentation. The effect continuity deviation value measures whether the historical effect performance of the material combination unit in different ad slots conforms to the preset continuity relationship. It is calculated by analyzing the changing trend of the material combination unit's historical effect feature parameters in the ad slot order, reflecting whether the effect can form an effective connection between ad slots.

[0064] The combined condition deviation value is a unified evaluation index obtained by comprehensively processing the above-mentioned deviation values. In engineering implementation, it is usually obtained by weighted fusion or normalization of role deviation value, visual progression deviation value, and effect continuity deviation value. It is used to characterize the comprehensive deviation degree of a cross-ad slot creative combination under overall constraints. The condition minimization combination construction process is used to select the combination result with the smallest combined condition deviation value among multiple candidate creative combinations. In implementation, it can calculate and compare all candidate combinations through traversal or optimization algorithms to obtain the ad creative combination that satisfies the combination constraint relationship and has the smallest overall deviation degree.

[0065] In its practical operation, the system first parses the exposure paths and expression requirements between multiple ad placements based on the advertising task information, thereby generating cross-ad placement correlation features. Based on this, it constructs combination constraint relationships, uniformly modeling the expression role, visual distribution, and effect continuity of each ad placement. During the combination generation stage, for each candidate material combination unit, the system calculates its semantic matching degree in fulfilling the target expression role, the rationality of its visual presentation distribution, and the degree of continuity in historical effects across each ad placement. These calculation results are converted into corresponding deviation values, and a unified aggregation mechanism is used to generate combination condition deviation values. Subsequently, the system selects the combination with the smallest deviation value from all candidate combinations as the final output. This ensures that the generated ad material combination not only meets the adaptation requirements of a single ad placement but also forms a continuous expression link and synergistic effect across multiple ad placements, achieving a shift from single-point optimization to overall optimization.

[0066] Furthermore, the step of determining the combination-level evaluation parameters for each ad creative combination includes: S401. Based on the content correspondence between each material combination unit, perform consistency calculation processing to generate corresponding cross-ad slot consistency parameters; S402. Based on the display constraint information, map the adaptation value results to generate corresponding display adaptation parameters; S403. Based on the combination distribution relationship of each material combination unit in each ad position, perform collaborative calculation processing to generate corresponding cross-position collaborative parameters; S404 integrates cross-ad slot consistency parameters, display adaptation parameters, and cross-slot collaboration parameters to generate corresponding combined-level evaluation parameters.

[0067] Furthermore, the step of generating corresponding cross-position collaborative parameters by performing collaborative calculations based on the combination distribution relationship of each material combination unit in each ad position includes: S4031. Determine the corresponding unit arrangement sequence for each ad slot according to the preset ad slot order, and construct the corresponding cross-position distribution sequence based on the unit arrangement sequence; S4032. Based on the cross-position distribution sequence, statistically analyze the combination distribution relationship of the material combination units between adjacent ad positions. The combination distribution relationship includes at least the number of corresponding categories, the number of category jumps, and the number of category repetitions. S4033. Based on the ratio of the number of corresponding categories to the total number of preset adjacent position pairs, determine the corresponding consistent distribution value; based on the ratio of the number of category jumps to the total number of preset adjacent position pairs, determine the corresponding jump distribution value; based on the ratio of the number of category repetitions to the length of the cross-position distribution sequence, determine the corresponding repetition distribution value. S4034. Calculate collaborative scores based on consistent distribution values, jump distribution values, and repeated distribution values ​​to generate corresponding cross-position collaborative scores. S4035. Normalize the cross-positional collaboration score to generate the corresponding cross-positional collaboration parameters.

[0068] In this embodiment, the content correspondence between material combination units is used to characterize the degree of association between materials in different ad slots at the content expression level. In engineering implementation, it can be obtained by aligning the semantic expression results of each material combination unit, for example, by calculating the similarity between semantic feature vectors or determining the consistency of category affiliation, thus providing a basis for consistency calculation. The cross-ad slot consistency parameter is used to quantify the degree of unity of materials in expressing themes between different ad slots. It is usually generated based on the content correspondence through statistical or similarity calculations to reflect whether the materials in the combination are organized around the same expression direction. The display adaptation parameter is used to describe the display matching of material combinations in different ad slots. It is obtained by further mapping the adaptation result value to a unified evaluation space. This mapping process can be achieved through linear normalization or interval transformation to make the adaptation results of different ad slots comparable. The combination distribution relationship is used to describe the arrangement and category distribution of material combination units in multiple ad slots. In implementation, it is usually represented in sequence form as the arrangement of material categories in different ad slots, thus providing a structural basis for subsequent collaborative calculation.

[0069] The unit arrangement sequence represents the sequence of material combination units arranged according to a preset ad slot order. In engineering implementation, it can be generated by indexing the material combination units corresponding to each ad slot in a fixed order, such as arranging them in the order of Banner, feed, and recommendation slots to form a sequence structure. The cross-position distribution sequence is a sequence representation formed by extracting the category information based on the unit arrangement sequence. It only retains the category affiliation information of each material combination unit and is used to describe the distribution state of the combination at the category level. The category correspondence count represents the number of times material combination units in adjacent ad slots belong to the same category. It is obtained by traversing adjacent positions in the cross-position distribution sequence and comparing categories. The category jump count represents the number of times the material category changes in adjacent ad slots, reflecting the degree of expression switching between different ad slots. The category repetition count represents the situation where the same category appears consecutively in the cross-position distribution sequence, which is used to measure the concentration of the combination in a certain category. The consistency distribution value represents the proportion of categories that maintain consistency between adjacent ad slots, calculated as the ratio of the number of times a category exhibits consistency to the total number of adjacent slot pairs. The jump distribution value represents the proportion of categories that exhibit changes, calculated as the ratio of the number of times a category jumps to the total number of adjacent slot pairs. The repetition distribution value represents the degree of continuous repetition of a category within a sequence, calculated as the ratio of the number of times a category repeats to the sequence length. The cross-slot collaboration score is used to comprehensively evaluate the above distribution characteristics. It is generated by weighting the consistency, jump, and repetition distribution values, for example, by positively reinforcing the consistency distribution and suppressing jumps and repetitions, thus obtaining a numerical result reflecting the overall degree of collaboration. The cross-slot collaboration parameter is the result of normalizing the cross-slot collaboration score to ensure it falls within a uniform numerical range, facilitating integration with other evaluation parameters. The combined evaluation parameter is used to uniformly and quantitatively evaluate the entire ad creative combination. It is obtained by fusing the cross-ad slot consistency parameter, display adaptation parameter, and cross-slot collaboration parameter, thus reflecting the combination's comprehensive performance in terms of content consistency, display matching, and structural collaboration.

[0070] In actual operation, the system first constructs a corresponding permutation sequence based on the material combination units of each ad slot, and extracts category information to form a cross-position distribution sequence. Then, by traversing the sequence, it calculates the correspondence, jumps and repetitions between categories, converts the above statistical results into proportional distribution values, and then calculates the collaborative score through weighted calculation. After further normalization, it obtains the cross-position collaborative parameters. Finally, it integrates these parameters with other evaluation parameters to complete the overall collaborative evaluation of the ad material combination, providing a basis for subsequent sorting and screening.

[0071] Assume an ad creative combination corresponds to three ad slots: a homepage banner ad slot, a news feed ad slot, and a featured ad slot. The corresponding creative combination unit categories for the three ad slots are, in order, efficacy, efficacy, and promotion. First, construct the cross-slot distribution sequence as [Efficacy, Efficacy, Promotion]. For adjacent ad slot pairs, there are two groups: Banner-News Feed and News Feed-Feedback. Banner-News Feed pairs have the same category, so the category count is recorded as 1. News Feed-Feedback pairs have different categories, so the category jump count is recorded as 1. Since the efficacy category appears consecutively once in the cross-slot distribution sequence, the category repetition count is recorded as 1. Then, using the preset total number of adjacent slot pairs (2) as the denominator, calculate the consistent distribution value as 1 / 2 and the jump distribution value as 1 / 2. Finally, using the cross-slot distribution sequence length (3) as the denominator, calculate the repetition distribution value as 1 / 3. Then, according to the preset calculation relationship, for example, cross-position collaboration score = 0.5 × consistent distribution value - 0.3 × jump distribution value - 0.2 × repeat distribution value, we get cross-position collaboration score = 0.5 × 0.5 - 0.3 × 0.5 - 0.2 × 0.333, which is approximately 0.0334. Finally, the cross-position collaboration score is mapped to the interval between 0 and 1 to generate the corresponding cross-position collaboration parameters. In this way, the cross-position collaboration parameters are not an abstract concept, but a parameter result directly calculated from the arrangement and distribution relationship of the material combination unit on multiple ad positions.

[0072] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for optimizing the combination of advertising creatives, characterized in that, The method for optimizing advertising creative combinations includes: Obtain advertising task information and a candidate advertising creative pool, wherein the candidate advertising creative pool contains multiple advertising creative units, and obtain the display constraint information corresponding to each advertising position based on the advertising task information; Each of the advertising material units is subjected to feature parsing processing to generate corresponding material feature parameters. The material feature parameters include at least content semantic feature parameters, presentation feature parameters, and historical effect feature parameters. Based on the material feature parameters and the display constraint information, the advertising material unit is processed to generate multiple advertising material combinations across advertising slots. Determine the combination-level evaluation parameters for each of the ad creative combinations, wherein the combination-level evaluation parameters include at least cross-ad placement consistency parameters, display adaptation parameters, and cross-ad placement collaboration parameters; Based on the aforementioned combination-level evaluation parameters, multiple ad creative combinations are compared and ranked to select the ad creative combination with the best evaluation results as the target ad creative combination. Based on the target material combination, configurations are made and outputs are made in multiple ad slots, and ad delivery is executed.

2. The advertising material combination optimization method according to claim 1, characterized in that, The step of generating multiple ad creative combinations by performing cross-ad slot combination processing on the ad creative unit based on the material feature parameters and the display constraint information includes: Based on the material feature parameters, each of the advertising material units is classified to generate multiple material category sets; Based on the display constraint information, calculate the adaptation result value of each of the material category sets corresponding to each ad slot; Based on the adaptation results, a filtering process is performed to generate material combination units corresponding to each ad slot; Based on the material combination units of each ad position, cross-ad position combination construction processing is performed to generate multiple ad material combinations, wherein each ad material combination includes at least the material combination unit corresponding to each ad position.

3. The advertising material combination optimization method according to claim 2, characterized in that, The step of determining the combination-level evaluation parameters for each of the advertising creative combinations includes: Based on the content correspondence between each material combination unit, consistency calculation is performed to generate corresponding cross-ad slot consistency parameters; Based on the display constraint information, the adaptation result value is mapped to generate corresponding display adaptation parameters; Based on the combination distribution relationship of each material combination unit in each ad position, collaborative calculation processing is performed to generate corresponding cross-position collaborative parameters; Based on the cross-ad slot consistency parameters, the display adaptation parameters, and the cross-slot collaboration parameters, corresponding combined-level evaluation parameters are generated.

4. The advertising material combination optimization method according to claim 3, characterized in that, The step of generating corresponding cross-position collaborative parameters by performing collaborative calculations based on the combination distribution relationship of each of the material combination units in each ad position includes: Determine the unit arrangement sequence corresponding to each ad slot according to the preset ad slot order, and construct the corresponding cross-position distribution sequence based on the unit arrangement sequence; Based on the cross-position distribution sequence, the combination distribution relationship of the material combination units between adjacent ad positions is statistically analyzed. The combination distribution relationship includes at least the number of times the categories correspond, the number of times the categories change, and the number of times the categories repeat. Based on the ratio of the number of times the category corresponds to the total number of preset adjacent position pairs, the corresponding consistent distribution value is determined; based on the ratio of the number of times the category jumps to the total number of preset adjacent position pairs, the corresponding jump distribution value is determined; based on the ratio of the number of times the category repeats to the length of the cross-position distribution sequence, the corresponding repeat distribution value is determined. Based on the consistent distribution value, the jump distribution value, and the repeating distribution value, a collaborative score is calculated to generate a corresponding cross-position collaborative score. The cross-position collaboration score is normalized to generate the corresponding cross-position collaboration parameters.

5. The advertising material combination optimization method according to claim 2, characterized in that, The step of constructing cross-advertisement combination units based on each ad placement to generate multiple ad material combinations, wherein each ad material combination includes at least one material combination unit corresponding to each ad placement, includes: Based on advertising task information, obtain the correlation features between different advertising positions; Based on the aforementioned association features, the combination constraint relationships between the various ad slots are determined; Based on the aforementioned combination constraints, each of the material combination units is constructed by condition minimization to generate multiple corresponding advertising material combinations.

6. The advertising material combination optimization method according to claim 5, characterized in that, The step of constructing multiple advertising material combinations by performing condition minimization combinations on each of the material combination units according to the combination constraint relationship includes: Based on the aforementioned combination constraints, the target expression role of each ad placement in the cross-ad placement combination is determined, as well as the visual distribution relationship and historical effect inheritance relationship. The target expression role includes at least the main expression role, the inheriting expression role, and the conversion expression role. Based on the correspondence between each material combination unit and each target expression role, calculate the role deviation value of each material combination unit in each ad position; Based on the visual distribution relationship, the visual progressive deviation value of each material combination unit in each advertisement position is calculated. Based on the historical effect continuity relationship, calculate the effect continuity deviation value of each of the aforementioned material combination units in each ad position; Based on the character deviation value, the visual progression deviation value, and the effect continuation deviation value, condition aggregation processing is performed to generate corresponding combined conditional deviation values; Based on the deviation value of the combined conditions, a minimum combination construction process is performed to generate the multiple advertising creative combinations.

7. The advertising material combination optimization method according to claim 2, characterized in that, The step of classifying each advertising material unit based on the material feature parameters to generate multiple material category sets includes: Based on the material feature parameters, adaptive weights, content semantic feature parameters, presentation feature parameters, and historical effect feature parameters are determined. Based on the content semantic feature parameters and the adaptive weights determined according to the material feature parameters, semantic classification processing is performed on each of the advertising material units to generate the corresponding initial category affiliation; Based on the aforementioned performance feature parameters, the initial category assignment is corrected to generate the corresponding intermediate category assignment. Based on the historical effect feature parameters, the intermediate category assignment is further corrected to generate the corresponding final category assignment; Based on the final category assignment of each advertising creative unit, a corresponding creative category set is generated.

8. The method for optimizing advertising material combinations according to claim 1, characterized in that, The step of performing feature parsing processing on each of the advertising material units to generate corresponding material feature parameters includes: Each of the aforementioned advertising material units undergoes content parsing processing to extract text and image information from each of the aforementioned advertising material units; The text information is segmented to generate a corresponding word sequence; The degree value of the term sequence is determined, and corresponding adaptive weights are generated; The word sequence is subjected to word frequency statistics and semantic vector mapping to generate corresponding content semantic feature parameters, and the adaptive weights and the content semantic feature parameters are associated and bound together. Visual feature extraction processing is performed based on the image information to generate corresponding representation feature parameters; Obtain historical delivery data corresponding to each of the aforementioned advertising material units, and perform statistical analysis and processing based on the historical delivery data to generate corresponding historical performance feature parameters; Based on the content semantic feature parameters, the presentation feature parameters, and the historical effect feature parameters, parameter integration processing is performed to generate corresponding material feature parameters.

9. The method for optimizing advertising material combinations according to claim 1, characterized in that, The step of obtaining display constraint information for multiple ad slots based on the ad task information includes: Based on the advertising task information, multiple advertising slots are determined, along with the identification information for each advertising slot; Based on the identification information, the corresponding display specification parameters are determined, including display size parameters, layout structure parameters, and content length limit parameters; Based on the identification information, the corresponding environmental location information is determined, including terminal device type parameters, geographic location parameters, and network environment parameters; The display specification parameters and environmental location information are associated and mapped to generate corresponding constraint parameters. The various constraint parameters are then integrated to generate corresponding display constraint information.

10. The method for optimizing advertising material combinations according to claim 9, characterized in that, The step of associating and mapping the display specification parameters and environmental location information to generate corresponding constraint parameters includes: Based on the terminal device type parameter, the display size parameter is adapted to the terminal to generate the terminal-adapted display size parameter. Based on the geographic location parameters, the layout structure parameters are adapted to the region to generate the adapted layout structure parameters. Based on the network environment parameters, the content length limit parameters are subjected to network adaptation processing to generate network-adapted content length limit parameters; The constraint parameters are generated by combining and processing the display size parameters adapted to the terminal, the layout structure parameters adapted to the region, and the content length limit parameters adapted to the network.

Images