An advertisement material production method and system based on artificial intelligence

Abstract

The application relates to the technical field of advertisement material production, in particular to an advertisement material production method and system based on artificial intelligence. The advertisement material production method based on artificial intelligence comprises the following steps: S1: collecting advertisement data watched by a target user in the past, dividing the advertisements into a first preset number of advertisement subareas, obtaining historical advertisement data in each advertisement subarea, and obtaining various labels in each subarea after statistical analysis of the historical advertisement data; S2: according to a second preset number of historical advertisement data in each advertisement subarea, analyzing the historical advertisement data to obtain a first advertisement result or a second advertisement result; and S3: taking a region with a corresponding weight greater than a first preset threshold weight as a first anchor region, comparing and analyzing the first anchor region with a preset advertisement element size to obtain a first analysis result or a second analysis result. The application sets the interest advertisement elements and positions of a target user.

Description

Technical Field

[0001] This invention relates to the field of advertising material production technology, and in particular to an advertising material production method and system based on artificial intelligence. Background Technology

[0002] With the popularization of digital media and the rapid development of network technology, the advertising industry has undergone a major transformation from traditional media to digital media. Digital advertising can be presented in various forms such as images, text, and video, and can also be precisely targeted based on user interests and behaviors. Therefore, the technology for creating advertising creatives has become a key factor in the effectiveness of digital advertising. However, while existing advertising creative production methods adapt to rapid market changes and immediate user needs based on user interests, this approach neglects user preferences in different regions and fails to promptly identify advertising placement areas that match user needs. Summary of the Invention

[0003] To overcome the drawback of advertising material settings not conforming to user preferences, this invention provides an artificial intelligence-based advertising material production method and system.

[0004] The technical solution of this invention is: a method for creating advertising materials based on artificial intelligence, comprising the following steps:

[0005] S1: Collect historical ad viewing data of target users, divide the ads into a first preset number of ad sub-regions, obtain historical ad data in each ad sub-region, and obtain various tags in each sub-region after statistical analysis;

[0006] S2: Based on the second preset number of historical advertising data in each advertising sub-region, analyze the historical advertising data to obtain a first advertising result or a second advertising result;

[0007] S3: The region whose corresponding weight is greater than the first preset threshold weight is taken as the first anchor region. The first anchor region is compared and analyzed with the preset advertising element size to obtain the first analysis result or the second analysis result.

[0008] S4: Combine elements based on the target user's interests in the ad settings area to obtain ad creatives.

[0009] Preferably, the step of collecting historical ad viewing data of target users, dividing the ads into a first preset number of ad sub-regions, obtaining historical ad data in each ad sub-region, and statistically analyzing the data to obtain various tags in each sub-region includes: based on the historical ad viewing data of target users, dividing the ad regions into a first preset number of ad sub-regions according to the proportion of the size and number of recent ad regions of target users, and counting the number of tags in each sub-region.

[0010] Preferably, the step of analyzing the historical advertising data based on a second preset number of historical advertising data within each advertising sub-region to obtain a first advertising result or a second advertising result includes:

[0011] If it is the first advertising result, then obtain the advertising repetition area and the corresponding number of times the advertising is covered based on the first advertising result, and obtain the corresponding weight of each area based on the advertising repetition area and the corresponding number of times the advertising is covered.

[0012] If it is the second advertising result, then the remaining area of ​​each advertising sub-region is filled according to the advertising elements that the user is interested in, where the remaining area is the difference between each advertising sub-region and the first anchoring region.

[0013] Preferably, the step of using the region whose corresponding weight is greater than the first preset threshold weight as the first anchor region, and comparing and analyzing the first anchor region with the preset ad element size to obtain a first analysis result or a second analysis result includes:

[0014] If it is the first analysis result, then the first anchoring area is trimmed according to the preset advertising element size to obtain the advertising setting area;

[0015] If the result is the second analysis result, the first anchoring area is expanded and connected according to the preset ad element size to obtain the ad setting area.

[0016] Preferably, if the result is the first analysis result, then trimming the first anchoring area according to the preset size of the advertising element to obtain the advertising setting area includes: adjusting the first anchoring area according to the weight of the first anchoring area to obtain the advertising setting area.

[0017] Preferably, if the result is a second analysis result, then the first anchoring region is expanded and connected according to the preset advertising element size to obtain the advertising setting region, including: obtaining the target user's interest elements within a first preset time period, obtaining the expansion tendency region according to the target user's interest elements, and obtaining the advertising setting region according to the expansion tendency region and the first anchoring region.

[0018] Preferably, the step of obtaining target user interest elements within a first preset time period and obtaining amplification tendency regions based on the target user interest elements includes: using a stack to collect a third preset number of historical interest elements of the target user, using the LFU algorithm to obtain a sequence of candidate user interest elements, and selecting the overlapping regions where the corresponding interest elements have a weight greater than a second preset threshold as amplification tendency regions based on the candidate user interest element sequence.

[0019] Preferably, obtaining the advertising setting area based on the amplification tendency region and the first anchoring region includes: selecting adjacent regions based on the first anchoring region and the amplification tendency region to obtain a first-level adjacent region set;

[0020] When the first-level adjacent region set does not have an expansion tendency region, the first anchor region is expanded proportionally to obtain the advertising setting region;

[0021] When there are regions with a tendency to expand in the first-level adjacent region set, expansion is performed based on the orientation of the regions with a tendency to expand, and adjacent regions are selected according to the total expanded region to obtain the second-level adjacent location set.

[0022] Repeat the analysis and expansion. When there is no expansion-prone area in the N-level adjacent area set or the total area after expansion is greater than or equal to the preset advertising element size, the advertising setting area is obtained after pruning according to the weight size.

[0023] Preferably, the step of combining target user interest elements in the ad setting area to obtain ad creatives includes: based on the amplification tendency area corresponding to each interest element in the N-level adjacent area set, adjusting the amplification tendency area larger than a first preset range to obtain an interest element setting area, and setting it using the corresponding interest element creatives; using an interest prediction model to obtain target user ad interest elements; constructing ad creatives based on the target user ad interest elements and the main element area, wherein the main element area is the difference set area between the ad setting area and the interest element setting area; using a Transformer model to construct an interest prediction model; and using the evaluation results of the F1 score to adjust the model parameters and iteratively optimize the model.

[0024] Preferably, an artificial intelligence-based advertising creative production system includes:

[0025] The advertising area division module is used to divide the advertising area to obtain a first preset number of advertising sub-areas;

[0026] The advertising analysis and judgment module is used to determine whether a historical advertisement is a duplicate area or a remaining area of ​​each advertisement sub-area;

[0027] An anchoring region segmentation module is used to define regions whose corresponding weights are greater than a first preset threshold weight as first anchoring regions;

[0028] The ad setting area trimming module is used to adjust the first anchor area based on the weight of the first anchor area to obtain the ad setting area;

[0029] The ad setting area expansion module is used to expand and connect the first anchor area according to the preset ad element size to obtain the ad setting area;

[0030] The interest element setting area module modifies the expansion tendency area that is larger than the first preset range to obtain the interest element setting area.

[0031] The beneficial effects are: This invention analyzes the location, elements, and structure of advertisements that the target user was interested in in the past, obtains the regional division when making the advertisement material based on the structure, trims or expands the area based on the analysis of the overlapping areas of all advertisements, obtains the location that the target user is most interested in, and plans the advertisement material at that location based on the elements that the target user was recently interested in, so as to obtain complete advertisement material that meets the user's needs. Attached Figure Description

[0032] Figure 1 This is a flowchart of an artificial intelligence-based advertising material production method according to the present invention. Detailed Implementation

[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] Example 1: A method for creating advertising creatives based on artificial intelligence, such as... Figure 1 As shown, it includes the following steps:

[0035] S1: Collect historical ad viewing data of target users, divide the ads into a first preset number of ad sub-regions, obtain historical ad data in each ad sub-region, and obtain various tags in each sub-region after statistical analysis;

[0036] Based on the historical ad viewing data of target users, the ad area is divided into a first preset number of ad sub-regions according to the proportion of the size and number of recent ad areas of target users, and the number of tags in each sub-region is counted.

[0037] It should be noted that by obtaining and analyzing the historical ad viewing data of the target users, the ad structure that the target users are interested in can be determined. For example, if the target users prefer an ad structure where the text part is located below the image part, then the structure of the ad is confirmed by dividing it into regions. In this process, tags are used to obtain the labels of whether each region contains text or an image part.

[0038] S2: Based on the second preset number of historical advertising data in each advertising sub-region, analyze the historical advertising data to obtain the first advertising result or the second advertising result;

[0039] If it is the first advertising result, then obtain the advertising repetition area and the corresponding number of times the advertising is covered based on the first advertising result, and obtain the corresponding weight of each area based on the advertising repetition area and the corresponding number of times the advertising is covered.

[0040] If it is the second advertising result, then the remaining area of ​​each advertising sub-region is filled according to the advertising elements that the user is interested in, where the remaining area is the difference between each advertising sub-region and the first anchoring region.

[0041] It should be noted that the system differentiates between specific images and background fillers for different ad elements based on the target user's historical ad viewing data. When the ad element is a specific image, the system finds the common coverage area of ​​all specific images based on the target user's historical ad viewing data and assigns a weight based on the frequency of the common coverage area. When the ad element is a background filler, the system fills the remaining area of ​​each ad sub-region based on the ad elements that the user is interested in.

[0042] S3: Take the region whose corresponding weight is greater than the first preset threshold weight as the first anchor region, compare and analyze the first anchor region with the preset ad element size, and obtain the first analysis result or the second analysis result;

[0043] If it is the first analysis result, then the first anchoring area is trimmed according to the preset advertising element size to obtain the advertising setting area;

[0044] If the result is the second analysis result, the first anchoring area is expanded and connected according to the preset ad element size to obtain the ad setting area.

[0045] Based on the weights within the first anchor area, the advertising setting area is obtained by adjusting the first anchor area according to the weight magnitude.

[0046] Obtain the target user's interest elements within the first preset time period, obtain the expansion tendency area based on the target user's interest elements, and obtain the advertising setting area based on the expansion tendency area and the first anchoring area.

[0047] The stack is used to collect a third preset number of historical interest elements of the target user. The LFU algorithm is used to obtain the interest element sequence of the candidate user. Based on the interest element sequence of the candidate user, the overlapping area of ​​the corresponding interest element with a weight greater than the second preset threshold is selected as the expansion tendency area.

[0048] Based on the first anchored region and the expansion tendency region, neighboring regions are selected to obtain a first-level neighboring region set;

[0049] When there is no expansion-prone area in the first-level adjacent area set, the first anchored area is expanded proportionally to obtain the advertising setting area;

[0050] When there are regions with an amplification tendency in the first-level adjacent region set, amplification is performed based on the orientation of the amplification tendency region, and adjacent regions are selected according to the total region after amplification to obtain the second-level adjacent location set.

[0051] Repeat the analysis and expansion. When there is no expansion-prone area in the N-level adjacent area set or the total area after expansion is greater than or equal to the preset advertising element size, the advertising setting area is obtained after pruning according to the weight size.

[0052] It should be noted that within the image coverage area with different weights, the area with a weight greater than the first preset threshold is used as the first anchoring area. When the first anchoring area exceeds the preset ad element size, it is pruned according to the weight to obtain the preset ad element size. When the first anchoring area is smaller than the preset ad element size, the target user's interest elements within the first preset time period are obtained through the LFU algorithm. The overlapping area with a weight greater than the second preset threshold that the target user has recently been interested in is selected as the expansion tendency area. The first anchoring area is expanded in a directed manner. When there is no expansion tendency area, the image is expanded proportionally.

[0053] S4: Combine elements based on the target user's interests in the ad settings area to obtain ad creatives.

[0054] Based on the amplification tendency regions corresponding to each interest element within the N-level adjacent region set, the amplification tendency regions larger than the first preset range are trimmed to obtain the interest element setting region, and the corresponding interest element material is used for setting. The target user's advertising interest elements are obtained using the interest prediction model. The advertising material is constructed based on the target user's advertising interest elements and the main element region, where the main element region is the difference set region between the advertising setting region and the interest element setting region. The interest prediction model is constructed using the Transformer model, and the model parameters are adjusted using the evaluation results of the F1 score, and the model is iteratively optimized.

[0055] Example 2: Based on Example 1, an artificial intelligence-based advertising material production system includes:

[0056] The advertising area division module is used to divide the advertising area to obtain a first preset number of advertising sub-areas;

[0057] The advertising analysis and judgment module is used to determine whether a historical advertisement is a duplicate area or a remaining area of ​​each advertisement sub-area;

[0058] Anchor region segmentation module is used to define regions whose corresponding weights are greater than a first preset threshold weight as first anchor regions;

[0059] The ad setting area trimming module is used to adjust the first anchor area based on the weight of the first anchor area to obtain the ad setting area.

[0060] The ad setting area expansion module is used to expand and connect the first anchor area according to the preset ad element size to obtain the ad setting area;

[0061] The interest element setting area module modifies the expansion tendency area that is larger than the first preset range to obtain the interest element setting area.

[0062] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An artificial intelligence-based advertisement material production method, characterized by, The method comprises the following steps: S1: collecting historical advertisement data watched by a target user, dividing the advertisements into a first preset number of advertisement sub-regions, obtaining historical advertisement data in each advertisement sub-region, and obtaining various labels in each sub-region after statistical analysis of the historical advertisement data; S2: analyzing the historical advertisement data according to a second preset number of historical advertisement data in each advertisement sub-region to obtain a first advertisement result or a second advertisement result, including: if the first advertisement result, obtaining an advertisement repeated region and a corresponding advertisement coverage number according to the first advertisement result, and obtaining a corresponding weight of each region according to the advertisement repeated region and the corresponding advertisement coverage number; if the second advertisement result, filling the remaining regions of each advertisement sub-region according to the advertisement elements of interest to the user, wherein the remaining regions are the difference set of each advertisement sub-region and the first anchor region; S3: taking the region with a corresponding weight greater than a first preset threshold weight as a first anchor region, comparing and analyzing the first anchor region with a preset advertisement element size to obtain a first analysis result or a second analysis result, including: if the first analysis result, pruning the first anchor region according to the preset advertisement element size to obtain an advertisement setting region; if the second analysis result, expanding and connecting the first anchor region according to the preset advertisement element size to obtain an advertisement setting region; S4: combining the advertisement setting region according to the target user interest elements to obtain advertisement materials.

2. The AI-based advertisement material production method of claim 1, wherein the AI-based advertisement material production method is characterized by, The collecting of historical advertisement data watched by a target user, the dividing of the advertisements into a first preset number of advertisement sub-regions, the obtaining of historical advertisement data in each advertisement sub-region, and the obtaining of various labels in each sub-region after statistical analysis of the historical advertisement data, comprise: based on the historical advertisement data watched by the target user, dividing the advertisement region according to the proportion of the size and number of the recent advertisement region of the target user to obtain a first preset number of advertisement sub-regions, and counting the number of labels in each sub-region.

3. The AI-based advertisement material production method of claim 1, wherein the AI-based advertisement material production method is characterized by, If the first analysis result, the pruning of the first anchor region according to the preset advertisement element size to obtain an advertisement setting region, comprises: based on the weight in the first anchor region, adjusting the first anchor region according to the weight size to obtain an advertisement setting region.

4. The AI-based advertisement material production method of claim 1, wherein the AI-based advertisement material production method is characterized by, If the second analysis result, the expanding and connecting of the first anchor region according to the preset advertisement element size to obtain an advertisement setting region, comprises: obtaining target user interest elements in a first preset time period, obtaining an expansion tendency region according to the interest elements of the target user, and obtaining an advertisement setting region according to the expansion tendency region and the first anchor region.

5. The AI-based advertisement material production method of claim 4, wherein the AI-based advertisement material production method is characterized by, The obtaining of target user interest elements in a first preset time period and the obtaining of an expansion tendency region according to the interest elements of the target user, comprise: collecting a third preset number of historical interest elements of a target user using a stack, obtaining a candidate user interest element sequence using an LFU algorithm, and selecting an overlapping region with a corresponding interest element greater than a second preset threshold weight as an expansion tendency region according to the candidate user interest element sequence.

6. The AI-based advertisement material production method of claim 4, wherein the AI-based advertisement material production method is characterized by, The advertisement setting area is obtained according to the expansion tendency area and the first anchor area, and the method comprises the following steps: based on the first anchor area and the expansion tendency area, a first-level adjacent area set is selected to obtain a first-level adjacent area set; When the first-level adjacent area set does not have an expansion tendency area, the first anchor area is expanded to obtain an advertisement setting area; When the first-level adjacent area set has an expansion tendency area, the expansion tendency area is expanded based on the position of the expansion tendency area, and a second-level adjacent area set is obtained based on the total area after expansion; When the N-level adjacent area set does not have an expansion tendency area or the total area after expansion is greater than or equal to the preset advertisement element size, the advertisement setting area is obtained after pruning based on the weight size.

7. The artificial intelligence-based advertisement material production method of claim 6, wherein the advertisement material is produced based on the advertisement material production plan. The advertisement material is obtained by combining the target user interest elements in the advertisement setting area, and the method comprises the following steps: based on the expansion tendency area corresponding to each interest element in the N-level adjacent area set, the expansion tendency area greater than the first preset range is trimmed to obtain an interest element setting area, and the corresponding interest element material is set, the target user advertisement interest element is obtained by using an interest prediction model, the target user advertisement interest element and a main element area are used to construct an advertisement material, the main element area is a difference set area of the advertisement setting area and the interest element setting area, the interest prediction model is constructed by using a Transformer model, the model parameters are adjusted by using the evaluation results of F1 scores, and the model is iteratively optimized. ​ 8. An artificial intelligence-based advertisement material production system for implementing an artificial intelligence-based advertisement material production method according to any one of claims 1 to 7, characterized by, It comprises: An advertisement area division module is used to divide the advertisement area to obtain a first preset number of advertisement sub-areas; An advertisement analysis and judgment module is used to judge whether the historical advertisement is an advertisement repeated area or a remaining area of each advertisement sub-area; An anchor area division module is used to take the area with a weight greater than a first preset threshold weight as a first anchor area; An advertisement setting area pruning module is used to adjust the first anchor area based on the weight in the first anchor area to obtain an advertisement setting area; An advertisement setting area expansion module is used to expand and connect the first anchor area based on a preset advertisement element size to obtain an advertisement setting area; An interest element setting area module is used to trim the expansion tendency area greater than the first preset range to obtain an interest element setting area.

Images