system

Abstract

A system is provided.SOLUTION: A system, comprising: means for collecting daily click-through rate (CTR) data; means for analyzing features of historical banner advertisements; means for extracting features of high CTR banners; means for generating banners utilizing a learning model; means for optimizing the generated banners based on landing page information; and means for automatically expanding the generated banners to a plurality of sizes.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The technology of the present disclosure relates to a system. [Background technology]

[0002] Patent document 1 discloses a persona chatbot control method performed by at least one processor, the method including the steps of receiving a user utterance, adding the user utterance to a prompt including an instruction sentence related to a description of the chatbot character, encoding the prompt, and inputting the encoded prompt into a language model to generate a chatbot utterance in response to the user utterance. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-180282 Summary of the Invention [Problem to be solved by the invention]

[0004] The conventional process of creating banner ads involves a lot of manual work, which places a heavy burden on designers, especially when generating banners of different sizes. Furthermore, finding the optimal design elements to maximize click-through rates (CTR) requires extensive trial and error, resulting in inefficiency. The objective of this invention is to provide a system that automatically generates optimal banners that maximize CTR and can accommodate multiple sizes, while reducing the designer's workload. [Means for solving the problem]

[0005] The present invention provides a system that includes a means for collecting daily click-through rate (CTR) data, a means for analyzing the characteristics of past banner ads, a means for extracting the characteristics of high-CTR banners, a means for generating banners using a learning model, a means for optimizing the generated banners based on landing page information, and a means for automatically expanding the generated banners to multiple sizes. This system automatically generates optimal banners that achieve high CTR and can expand them to various sizes, significantly reducing the workload of designers. Furthermore, the present invention includes a means for receiving landing page information and adding a catchy slogan to the banner generated based on that information, thereby generating banners that are more effective for the target audience. Furthermore, the learning model includes a means for optimizing the banner's layout, color, catchy slogan, photographic material, motif, line of sight, and font elements using the characteristics of high-CTR banners, making it easy to find the optimal design elements for maximizing CTR.

[0006] "Click-through rate (CTR)" is the ratio of the number of clicks to the number of times an ad is displayed, and is an important indicator of the effectiveness of an ad.

[0007] A "banner ad" is an advertising format that uses images or animations displayed on web pages or applications, and is intended to lead users to a specific landing page when clicked.

[0008] A "learning model" is a statistical model that learns specific patterns and characteristics based on past data and makes predictions and optimizations for new data.

[0009] A "landing page" is the web page a user first visits when they click on an advertisement, and is a page that provides specific information or encourages action.

[0010] A "catchphrase" is a short phrase or phrase used in advertising and marketing to attract users' attention and succinctly convey the appeal of a product or service.

[0011] "Sizing" is the process of adapting a particular design or format to different viewing environments by adjusting or converting it into different sizes and aspect ratios. [Brief explanation of the drawings]

[0012] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Figure 2] 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. [Figure 3] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Figure 5] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a third embodiment. [Figure 6] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and a headset-type terminal according to a third embodiment. [Figure 7] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and a robot according to a fourth embodiment. [Figure 9] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 10] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 11] FIG. 3 is a sequence diagram showing a processing flow of the data processing system according to the first embodiment. [Figure 12] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1. [Figure 13] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system according to the second embodiment when an emotion engine is combined. [Figure 14] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 when an emotion engine is combined. DETAILED DESCRIPTION OF THE INVENTION

[0013] An example of an embodiment of a system according to the technology of the present disclosure will be described below with reference to the accompanying drawings.

[0014] First, the terms used in the following description will be explained.

[0015] In the following embodiments, a coded processor (hereinafter simply referred to as a "processor") may be a single arithmetic device or a combination of multiple arithmetic devices. Furthermore, a processor may be a single type of arithmetic device or a combination of multiple types of arithmetic devices. Examples of arithmetic devices include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), and an APU (Accelerated Processing Unit).

[0016] In the following embodiments, a coded RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a working memory by a processor.

[0017] In the following embodiments, the coded storage is one or more non-volatile storage devices that store various programs, various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), and magnetic tapes.

[0018] In the following embodiments, a communication I / F (Interface) with a symbol is an interface including a communication processor, an antenna, etc. The communication I / F controls communication between multiple computers. Examples of communication standards applied to the communication I / F include wireless communication standards including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), Bluetooth (registered trademark), etc.

[0019] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed connected by "and / or."

[0020] [First embodiment]

[0021] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0022] 1, a data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0023] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0024] The smart device 14 includes a computer 36, a reception device 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The reception device 38, the output device 40, and the camera 42 are also connected to the bus 52.

[0025] The reception device 38 includes a touch panel 38A, a microphone 38B, and the like, and receives user input. The touch panel 38A detects contact with an indicator (for example, a pen or a finger) to receive user input by the touch of the indicator. The microphone 38B detects the user's voice to receive user input by voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0026] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form of expression that the user 20 can perceive (for example, audio and / or text). The display 40A displays visible information such as text and images in accordance with instructions from the processor 46. The speaker 40B outputs audio in accordance with instructions from the processor 46. The camera 42 is a compact digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0027] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54.

[0028] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0029] 2, in the data processing device 12, a specific process is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific process is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0030] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0031] In the smart device 14, the processor 46 performs the reception output process. The storage 50 stores a reception output program 60. The reception output program 60 is used in conjunction with the specific processing program 56 by the data processing system 10. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0032] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0033] The present invention relates to a system that collects daily click-through rate (CTR) data, analyzes the characteristics of past banner advertisements, extracts the characteristics of high-CTR banners, generates banners using a learning model, optimizes the generated banners based on landing page information, and automatically expands the generated banners into multiple sizes.

[0034] As an embodiment of the present invention, a specific program and its processing content will be described below.

[0035] Data collection and analysis phase

[0036] collection

[0037] The server collects CTR data for banner ads from a database every day. This data includes many characteristic information about past banner ads (layout, color, catchy copy, photo material, motif, gaze, font, etc.).

[0038] analysis

[0039] The server analyzes the collected data and extracts the characteristics of banners with high CTRs, using statistical methods to analyze the data and identify design elements and patterns that indicate high CTRs.

[0040] Feature extraction and training phase

[0041] Feature Extraction

[0042] The server builds a training dataset based on features extracted from high-CTR banners, including detailed information about each element (such as which layout is most effective or which color is most eye-catching).

[0043] study

[0044] This data is then used to train a learning model, using machine learning algorithms such as random forests and neural networks to build a model that predicts optimal design elements.

[0045] Banner generation phase

[0046] Receiving information

[0047] The user provides landing page information, including the landing page's key message and promotion.

[0048] Banner Generation

[0049] The server uses the learned model to generate an optimal banner based on the provided landing page information, including layout, color, tagline, stock photos, motif, line of sight, font, etc.

[0050] Size development phase

[0051] Size range

[0052] The generated banner will be automatically expanded to multiple sizes. The device will handle this process and create banner images of different sizes, for example 300x250, 728x90, 160x600, etc.

[0053] Specific examples

[0054] User operations

[0055] A user sets up a landing page with promotional information for "New Year Sale!" and requests to generate a banner ad.

[0056] Server Processing

[0057] 1. The server collects and analyzes past data and extracts the characteristics of high CTR banners.

[0058] 2. The learning model is trained based on the extracted features so that it can generate optimal banners.

[0059] 3. Based on the landing page information provided by the user, the server generates a suitable banner containing the tagline "New Year Sale!"

[0060] 4. The device will automatically expand the generated banner to various sizes such as 300x250, 728x90, and 160x600.

[0061] The above is a detailed description of the embodiment of the present invention. This system reduces the workload of designers and automatically generates optimal banners that maximize CTR.

[0062] The processing flow will be explained below.

[0063] Step 1:

[0064] The server collects daily banner ad click-through rate (CTR) data. It obtains past banner ad features such as layout, color, catchy copy, stock photos, motifs, gaze, and font, as well as their CTR, from databases and log files.

[0065] Step 2:

[0066] The server analyzes the collected data and extracts the characteristics of banners with high CTR. Specifically, it filters data for banner ads with CTRs above a certain threshold and calculates the average characteristics. During this process, it uses statistical methods to identify which design elements contribute to a high CTR.

[0067] Step 3:

[0068] The server builds a training dataset based on the extracted features. It organizes the design elements of the filtered high-CTR banner ads into a feature vector and formats it to be fed into the training model.

[0069] Step 4:

[0070] The server uses the training dataset to train a model. For example, it uses machine learning algorithms such as random forests and neural networks to build a model for generating optimal banners. This model is responsible for predicting high CTR.

[0071] Step 5:

[0072] The user inputs landing page information, including key messaging and promotional content, and requests the creation of a banner ad.

[0073] Step 6:

[0074] The server receives the provided landing page information and generates the optimal banner based on it. The landing page information and the extracted characteristics of the high CTR banner are input into the learning model to determine the optimal design elements. The generated banner includes layout, color, catchy copy, stock photos, motif, line of sight, font, etc.

[0075] Step 7:

[0076] The device receives the generated banner and automatically expands it to multiple sizes, for example, resizing it to different sizes such as 300x250, 728x90, and 160x600, and then saving it. This allows you to prepare a banner that can be used in multiple ad spaces at once.

[0077] Step 8:

[0078] The generated banners in multiple sizes are provided to the user, who can then use them in actual advertising operations to check the effectiveness of CTR.

[0079] The above is the specific processing flow of the system that automatically generates banner ads to maximize CTR and deploys them in various sizes. This series of processes makes it possible to efficiently generate optimal banners that achieve a high CTR while reducing the workload on designers.

[0080] Example 1

[0081] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0082] Conventional banner ad generation systems lack the functionality to automatically analyze and apply design elements that are expected to improve click-through rates (CTR), and require manual design and adjustment. This places a heavy workload on designers, making it difficult to create effective banner ads. The objective of this invention is to provide a system that optimizes design elements and automatically expands to multiple sizes, efficiently generating effective banner ads.

[0083] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0084] In this invention, the server includes means for collecting daily click-through rate (CTR) data, means for analyzing characteristics of past banner ads, means for extracting characteristics of high CTR banners, means for generating banners using a learning model, means for optimizing the generated banners based on landing page information, means for automatically expanding the generated banners into multiple sizes, and means for saving the optimized banners in cloud storage. This makes it possible to automatically and efficiently generate and save optimal banner ads that are expected to have a high CTR.

[0085] "Click-through rate (CTR)" indicates the percentage of clicks on a displayed ad, and is an indicator used to measure the effectiveness of an ad.

[0086] A "banner ad" is advertising content in the form of an image or video displayed on a web page, and is created with the purpose of encouraging users to click on it.

[0087] "Features" refers to identifiable characteristics of a banner ad, such as specific design elements, layout, color, catchy copy, photographic material, motif, line of sight, font, etc.

[0088] A "high CTR banner" refers to a banner ad that has a relatively high click-through rate and contains an effective design and message.

[0089] A "learning model" is a computational model that uses machine learning algorithms to learn patterns and features from data and make predictions and classifications based on future data.

[0090] A "landing page" is a specific web page accessed from a web advertisement or search engine result, and is the first page a user visits.

[0091] "Optimization" refers to the process of adjusting conditions or factors to most effectively achieve a particular objective.

[0092] "Cloud storage" refers to services that allow data to be stored remotely and made accessible over the internet.

[0093] "Sizing" refers to the process of automatically adapting a particular design to different dimensions to accommodate multiple ad slots.

[0094] This invention provides a system for optimizing the click-through rate (CTR) of banner ads. This system collects CTR data, analyzes and extracts the characteristics of banner ads that show high CTR, generates optimal banners using a learning model, optimizes the generated banners based on landing page information, and deploys them in multiple sizes.

[0095] Data collection and analysis

[0096] The server first connects to a database to collect daily banner ad CTR data. This database stores information about past banner ad CTRs, layouts, colors, taglines, stock photos, motifs, gaze patterns, fonts, etc. The server extracts this information using SQL queries and performs data cleaning, such as removing missing and outlier values ​​and normalizing the data.

[0097] The server then analyzes the collected data using statistical methods, such as calculating average CTR, variance, and correlation coefficients, to identify design elements and patterns that indicate high CTR.

[0098] Feature extraction and model training

[0099] The server extracts the features of high-CTR banners and builds a training dataset based on a multidimensional feature vector. This dataset contains the features of each banner and the corresponding CTR.

[0100] Next, the server selects a machine learning algorithm (e.g., random forest, neural network, etc.) and trains the model using the training dataset. This training model is then used to generate banners with optimal design elements to predict high CTR. Specifically, the random forest model can be trained using Scikit-learn, an educational Python library.

[0101] Banner generation and size development

[0102] The user enters the landing page information, which includes the main message and promotional content. This information is sent to the server via the terminal. For example, the promotional information entered may be "New Year Sale!"

[0103] The server takes this input information and uses a trained model to generate an optimal banner, including layout, color, tagline, stock image, motif, line of sight, font, etc. The actual banner image is generated using the Pillow library and other image processing tools.

[0104] The device then converts the generated banner to different sizes, using OpenCV or other image conversion libraries to automatically convert it to different ad sizes such as 300x250, 728x90, and 160x600.

[0105] Final generation and saving

[0106] The user can check the generated banner and request corrections if necessary. The server then saves the optimized banner to cloud storage, often using Amazon S3 for example. The saved banner is then provided to the user as a URL.

[0107] For example, the prompt is:

[0108] Generate the optimal banner based on the promotion information "New Year Sale!"

[0109] Using this system will significantly reduce the workload of designers and enable the automatic generation, deployment, and storage of banner ads that are expected to have a high CTR.

[0110] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0111] Step 1:

[0112] Data collection

[0113] The server collects daily click-through rate (CTR) data from a database. Specifically, it connects to the database and executes SQL queries to extract information related to past banner ads, such as CTR, layout, color, tagline, stock image, motif, gaze, and font. It specifies a date range and ad ID as input, and obtains a dataset containing CTR data as output. For example, it executes the query "SELECT FROM Banners WHERE date >= CURDATE() - INTERVAL 30 DAY;".

[0114] Step 2:

[0115] Data analysis

[0116] The server analyzes the collected data. It uses the CTR data collected in step 1 as input, performs data cleaning, and removes missing and outlier values. It then analyzes the data using statistical methods to identify design elements and patterns that indicate high CTR. Specifically, it calculates the average CTR, variance, correlation coefficient, etc., and obtains a list of elements that indicate high CTR as output.

[0117] Step 3:

[0118] Feature Extraction

[0119] The server extracts features from the high CTR banners identified in step 2. It uses the analysis result dataset as input and builds a feature vector for each banner. For example, it creates a matrix-format dataset containing detailed information such as which layout is most effective or which color is most eye-catching. As output, it obtains a training dataset.

[0120] Step 4:

[0121] Learning Phase

[0122] The server trains the machine learning model. It uses the training dataset constructed in step 3 as input and selects a machine learning algorithm such as a random forest or neural network. Specifically, it uses the Scikit-learn library to run the model's fit method and train the optimal predictive model. The output is the trained learning model.

[0123] Step 5:

[0124] Receiving information

[0125] The user inputs landing page information, specifically key messages and promotional content, through an input form, and the device sends this information to the server. The input is the landing page information, and the output is the data sent to the server.

[0126] Step 6:

[0127] Banner Generation

[0128] The server uses the trained model to generate the optimal banner based on the provided landing page information. Using the landing page information and the trained model as input, it calculates the banner design parameters that predict the highest CTR. Specifically, it determines the tagline, layout, color, stock photos, etc., and generates the actual banner image using the Pillow library. The generated banner image is obtained as output.

[0129] Step 7:

[0130] Size range

[0131] The device will unpack the generated banner into multiple sizes. As input, it uses the banner image generated in step 6 and automatically converts it to different sizes (e.g., 300x250, 728x90, 160x600) using an image conversion library such as OpenCV. As output, it gets the banner image for each size.

[0132] Step 8:

[0133] Final generation and saving

[0134] The server saves the generated banner in cloud storage. As input, it uses banner images of multiple sizes and uploads them to a cloud storage service. Specifically, it saves the banner images using a storage service such as Amazon S3, generates a URL for the saved banner, and provides it to the user. As output, it obtains the URL for the saved banner.

[0135] This allows the system to automatically generate optimal banner ads aiming for high CTR, and efficiently expand and save them in multiple sizes.

[0136] (Application example 1)

[0137] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0138] Conventional banner ad generation systems simply generate and display static banners based on past data. This makes it difficult to provide effective ads that maximize CTR, as it is not possible to optimize in real time or respond to dynamically changing market and user preferences. Furthermore, manual banner size deployment and optimization requires a great deal of effort, hindering efficient advertising operations.

[0139] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0140] In this invention, the server includes means for collecting daily click-through rate (CTR) data, means for analyzing the characteristics of past banner advertisements, means for extracting the characteristics of high-CTR banners, means for generating banners using a learning model, means for optimizing the generated banners based on landing page information, means for automatically expanding the generated banners to multiple sizes, and means for generating and displaying dynamically optimized banner advertisements in real time. This makes it possible to respond to changes in the market and users in real time and generate dynamically optimized banner advertisements. Furthermore, automatic expansion of banner sizes enables efficient advertising management.

[0141] "Click-through rate (CTR)" is an indicator that shows the ratio of the number of clicks to the number of times an ad is displayed.

[0142] A "banner ad" is a graphic advertisement that appears on a web page or within an application.

[0143] "Landing page information" refers to information about the content and structure of the web page to which a user is directed when they click on an advertisement.

[0144] A "learning model" is a machine learning algorithm that learns patterns and rules based on data and makes predictions and classifications.

[0145] "Feature extraction" is the process of extracting useful characteristics or patterns from data.

[0146] "Real-time" refers to the time characteristics in which data is processed and results are provided immediately.

[0147] "Optimization" is the process of adjusting parameters and conditions to achieve the best results in order to achieve a specific objective.

[0148] "Sizing" is the process of adjusting the original design to different sizes to generate multiple sizes of banners.

[0149] "Dynamic optimization" is the process of instantly adjusting the optimum state in response to changing conditions in real time.

[0150] A "prompt" refers to a sentence of instructions or questions that is input into a generative AI model.

[0151] MODE FOR CARRYING OUT THE INVENTION

[0152] To implement the present invention, a system in which a server, a terminal, and a user work in cooperation with each other is required. A specific embodiment of this system will be described below.

[0153] System Overview

[0154] The system collects daily click-through rate (CTR) data, analyzes the characteristics of past banner ads, extracts the characteristics of high-CTR banners, and generates banners using a generative AI model.The generated banners are optimized based on landing page information, automatically expanded into multiple sizes, and dynamically generated and displayed in real time.

[0155] Hardware and software used

[0156] Server: Performs primary data processing and training of learning models. Software used is Python, TensorFlow (Keras), and PIL (Pillow).

[0157] Terminal: Operated by the user, it displays the generated banner and inputs the prompt text. A smartphone or tablet is used as the interface.

[0158] Database: Uses data storage to accumulate CTR data and banner ad feature data.

[0159] Data collection and analysis phase

[0160] The server collects historical and daily banner ad click-through rate (CTR) data from a database. This data includes many characteristic information related to layout, color, tagline, stock photos, motifs, gaze, fonts, etc. The server analyzes this data using statistical methods to identify design elements and patterns that indicate high CTR.

[0161] Feature extraction and training phase

[0162] The server builds a training dataset based on features extracted from high-CTR banners, creating a dataset with detailed information about each element. The training model is then trained using this data to build a model that predicts optimal design elements using machine learning algorithms such as random forests and neural networks.

[0163] Banner generation phase

[0164] The user provides landing page information, including key messaging and promotional content. The server uses the trained model to generate an optimal banner based on the provided landing page information. The generated banner includes layout, color, tagline, stock photos, motif, line of sight, font, etc.

[0165] Size development phase

[0166] The generated banner will be automatically expanded to multiple sizes. The device will process this and create banner images of different sizes. For example, various sizes such as 300x250, 728x90, 160x600 etc. will be generated.

[0167] Dynamic optimization and real-time generation

[0168] The system dynamically optimizes to respond to changes in the market and users in real time: the server instantly adjusts banner ads based on current click-through rate data and trends, ensuring they are displayed optimally to users.

[0169] Specific examples

[0170] A user sets up a landing page with promotional information such as "Summer Sale Begins!" and requests the system to generate a banner ad. The server collects and analyzes past data to extract the characteristics of high-CTR banners. After the model is trained, the server generates an optimal banner containing the tagline "Summer Sale Begins!" based on the landing page information provided by the user. The generated banner is automatically expanded to various sizes, such as 300x250, 728x90, and 160x600, and is dynamically optimized for display depending on the device.

[0171] Prompt Sentence Examples

[0172] "Generate a 'Summer Sale!' banner that has a high click-through rate based on historical data. Resize the generated banner to 300x250, 728x90, and 160x600."

[0173] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0174] Step 1: Data collection

[0175] The server collects daily click-through rate (CTR) data from a database, which includes characteristic information about past banner ads (layout, color, catchy copy, stock photos, motifs, gaze, font, etc.). Input data is obtained from the database, and CTR data is generated as output.

[0176] Step 2: Data analysis

[0177] The server analyzes the collected CTR data using statistical methods to identify banner design elements and patterns that result in high click-through rates. The input data is CTR data, and the output is characteristic data of high CTR banners. This identified data indicates the design elements of high CTR banners.

[0178] Step 3: Feature extraction

[0179] The server builds a training dataset based on the features of high-CTR banners, including detailed information on each element (layout, color, catchy copy, stock photos, motif, gaze, font, etc.). The input data is a dataset containing the features, and the output is a training dataset.

[0180] Step 4: Model training

[0181] The server trains a learning model using a training dataset. Specifically, it builds the model using machine learning algorithms such as random forests and neural networks. The input data is the training dataset, and the output is the trained learning model.

[0182] Step 5: Receive landing page information

[0183] The user provides landing page information that will be the basis for the banner ad they want to generate. The input data is the landing page information entered by the user, and the output is the appropriately formatted landing page information.

[0184] Step 6: Generate a banner

[0185] The server uses the learning model to generate an optimal banner based on the provided landing page information. The generated banner includes layout, color, catchy copy, stock photos, motif, gaze, font, etc. The input data is the landing page information and the learning model, and the output is the generated banner.

[0186] Step 7: Size development

[0187] The server automatically expands the generated banner to multiple sizes, for example, banner images of different sizes such as 300x250, 728x90, 160x600, etc. The input data is the generated banner, and the output is the banner images resized to different sizes.

[0188] Step 8: Dynamic optimization and real-time generation

[0189] The server adjusts banner ads in real time based on current click-through rate data and trends, and displays them optimally to users. The input data is the latest click-through rate data and the generated banner, and the output is a dynamically optimized banner.

[0190] Step 9: Displaying the banner

[0191] Users can view the optimized banner ads on their smartphones or other devices and request modifications or regeneration as necessary. The input data is the dynamically optimized banner, and the output is the displayed banner ad.

[0192] Furthermore, an emotion engine that estimates the user's emotion may be combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion.

[0193] This invention relates to a system that combines click-through rate (CTR) data collection and analysis, banner generation using a learning model, optimization based on landing page information, and an emotion engine. The system aims to further maximize CTR by adjusting banner design elements based on user emotion data.

[0194] Data collection and analysis phase

[0195] collection

[0196] The server collects banner ad CTR data daily, retrieving information about past banner ad features such as layout, color, catchy copy, stock photos, motifs, gaze, and font, as well as their CTR, from databases and log files.

[0197] analysis

[0198] The server analyzes the collected data and extracts the characteristics of banners with high CTR. Specifically, it filters data for banner ads with CTRs above a certain threshold and calculates the average characteristics. In this process, it uses statistical methods to identify which design elements contribute to a high CTR.

[0199] Feature extraction and training phase

[0200] Feature Extraction

[0201] The server builds a training dataset based on the features extracted from high-CTR banners, including detailed information about each element (layout, color, catchy copy, stock photos, motif, gaze, and font).

[0202] study

[0203] This data is then used to train a learning model, using machine learning algorithms such as random forests and neural networks to build a model that predicts optimal design elements.

[0204] Banner generation phase

[0205] Receiving information

[0206] The user inputs landing page information, including key messaging and promotional content, and requests the creation of a banner ad.

[0207] Banner Generation

[0208] The server generates an optimal banner based on the landing page information and the characteristics of high-CTR banners obtained from the learning model, including layout, color, catchy copy, stock photos, motif, line of sight, font, etc.

[0209] Collaboration with emotion engine

[0210] Acquiring emotion data

[0211] The server obtains the user's emotional data from the emotion engine, including whether the user is in a positive, negative, or neutral emotional state.

[0212] Adjusting design elements

[0213] The server adjusts the banner's design elements based on the acquired emotional data, for example, using bright colors when the user is expressing positive emotions and muted colors when expressing negative emotions, and dynamically changes the tagline accordingly.

[0214] Size development phase

[0215] Size range

[0216] The generated banner will be automatically expanded to multiple sizes. The device will handle this process and create banner images of different sizes, for example 300x250, 728x90, 160x600, etc.

[0217] Specific examples

[0218] User operations

[0219] Assume that a user sets up a landing page with promotion information for "New Year Sale!" and requests the generation of a banner ad. Also, assume that the user's sentiment is positive.

[0220] Server Processing

[0221] 1. The server collects and analyzes past data and extracts the characteristics of high CTR banners.

[0222] 2. The learning model is trained based on the extracted features so that it can generate optimal banners.

[0223] 3. Based on the landing page information provided by the user, the server generates a suitable banner containing the tagline "New Year Sale!"

[0224] 4. The server retrieves the user's emotional data from the emotion engine and adjusts the design by using brighter colors if the emotion indicates a positive one.

[0225] 5. The device will automatically expand the generated banner to various sizes such as 300x250, 728x90, and 160x600.

[0226] The above is a detailed description of the embodiment of the present invention. This system reduces the workload of designers, efficiently generates optimal banners that maximize CTR, and also enables design adjustments based on user emotions.

[0227] The processing flow will be explained below.

[0228] Step 1:

[0229] The server collects daily banner ad click-through rate (CTR) data. It obtains past banner ad features such as layout, color, catchy copy, stock photos, motifs, gaze, and font, as well as their CTR, from databases and log files.

[0230] Step 2:

[0231] The server analyzes the collected data and extracts the characteristics of banners with high CTR. Specifically, it filters data for banner ads with CTRs above a certain threshold and calculates the average characteristics. In this process, it uses statistical methods to identify which design elements contribute to a high CTR.

[0232] Step 3:

[0233] The server builds a training dataset based on the features extracted from high-CTR banners, including detailed information about each element (layout, color, catchy copy, stock photos, motif, gaze, font, etc.).

[0234] Step 4:

[0235] The server uses the constructed training dataset to train a model. For example, it uses machine learning algorithms such as random forests and neural networks to build a model for generating optimal banners. This model plays a role in predicting high CTR.

[0236] Step 5:

[0237] The user inputs landing page information, including key messaging and promotional content, and requests the creation of a banner ad.

[0238] Step 6:

[0239] The server receives the provided landing page information and generates the optimal banner based on it. The landing page information and the extracted characteristics of the high CTR banner are input into the learning model to determine the optimal design elements. The generated banner includes layout, color, catchy copy, stock photos, motif, line of sight, font, etc.

[0240] Step 7:

[0241] The server obtains the user's emotional data from the emotion engine, which includes the user's emotional state, such as positive, negative, or neutral.

[0242] Step 8:

[0243] The server adjusts the banner's design elements based on the acquired emotional data, for example, using bright colors when the user is expressing positive emotions and muted colors when expressing negative emotions. Furthermore, the tagline dynamically changes depending on the emotion.

[0244] Step 9:

[0245] The device receives the generated banner and automatically expands it to multiple sizes, for example, resizing it to different sizes such as 300x250, 728x90, and 160x600, and then saving it. This allows a banner that can be used in multiple ad spaces to be prepared at once.

[0246] Step 10:

[0247] The generated banners in multiple sizes are provided to the user, who can then use them in actual advertising operations to check the effectiveness of CTR.

[0248] As a concrete example, suppose a user sets up a landing page with promotional information for "New Year Sale!" and requests the generation of a banner ad. Let's also assume that the user's sentiment is positive. The server uses a learning model and sentiment engine to generate an optimal banner that combines the "New Year Sale!" tagline with bright colors. The device then expands this banner into multiple sizes and provides it to the user. This series of processes allows users to efficiently use banner ads that aim for a high CTR.

[0249] Example 2

[0250] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0251] Conventional audiovisual advertisement generation systems do not take into account the user's emotional response and therefore have difficulty generating optimal advertisements. Furthermore, the process of automatically expanding the generated advertisements into multiple sizes is also time-consuming. Therefore, a new system for creating more effective audiovisual advertisements and automatically expanding them is needed.

[0252] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0253] In this invention, the server includes means for collecting daily selection rate data, means for analyzing characteristics of past audiovisual advertisements, means for extracting characteristics of audiovisual advertisements with high selection rates, means for generating audiovisual advertisements using a learning algorithm, means for optimizing the generated audiovisual advertisements based on guide page information, means for adjusting design elements based on user emotions using an emotion processing engine, and means for automatically expanding the generated audiovisual advertisements into multiple dimensions, thereby enabling efficient generation of optimal audiovisual advertisements according to user emotions and automatic expansion thereof.

[0254] "Selection rate data" refers to data that indicates the rate at which users click or select an audiovisual advertisement displayed.

[0255] "Audiovisual advertising" refers to advertising that uses both visual and auditory sensations to provide information to users.

[0256] A "learning algorithm" is a machine learning method for predicting optimal design elements for audiovisual advertisements based on collected data.

[0257] "Guide page information" refers to the content of a web page that contains promotional information for a particular product or service.

[0258] An "emotion processing engine" is a system that analyzes a user's emotional state and adjusts design elements based on the results.

[0259] "Design elements" are the components of an audiovisual advertisement, and specifically include layout, color, wording, image materials, theme, line of sight, and font.

[0260] "Dimensions" refers to the physical size and proportions of an audiovisual advertisement.

[0261] "User" means an individual or organization that utilizes the system to generate and display audiovisual advertisements.

[0262] This invention relates to a system that collects and analyzes audiovisual advertisement selection rate data, combines a learning algorithm and an emotion processing engine to generate optimal audiovisual advertisements, and optimizes and automatically deploys the advertisements based on guide page information.

[0263] The hardware required to implement this system includes a database server, a web server, an emotion processing engine server, and a terminal. The software used includes a database management system (e.g., MySQL), a data processing library (e.g., Pandas, NumPy), a machine learning library (e.g., Scikit-learn, TensorFlow), and an image processing library (e.g., PIL).

[0264] The specific program processing is as follows.

[0265] The server collects audiovisual ad selection rate data daily from a database and log files, including design elements such as layout, color, wording, image materials, theme, eye direction, and typeface. The server uses SQL queries to extract relevant data from the database and Python scripts to retrieve additional information from the log files.

[0266] The server then analyzes the collected data using statistical methods. It extracts design elements of audiovisual advertisements with high selection rates, filters advertisements that exceed a certain threshold, and calculates average features. Specifically, the server creates a data frame using the Pandas library and calculates statistical information such as the mean and standard deviation using the NumPy library.

[0267] The server builds a training dataset based on features extracted from high-selection audiovisual ads. This dataset includes detailed information about each design element. The server saves the extracted features in a CSV file, which is then fed into machine learning algorithms. The training algorithms include random forests and neural networks, which are trained using Scikit-learn and TensorFlow.

[0268] The user inputs information about the page through the system interface and requests the generation of an audiovisual advertisement. The input information includes the promotional content and key messages. The user submits this information using a web form.

[0269] The server combines the information on the user's guide page with the characteristics of high-selection audiovisual ads estimated from the learning model to generate optimal audiovisual ads. Specifically, the server uses PIL (Python Imaging Library) to combine elements such as layout and catchy copy to generate ad images. A generative AI model is also used in this process.

[0270] The server obtains user emotional data in real time from an emotion processing engine. This data includes the user's emotional state, such as positive, negative, or neutral. The server adjusts design elements based on this emotional data. For example, bright colors are used to indicate positive emotions, and muted colors to indicate negative emotions. The tagline also changes dynamically depending on the emotion.

[0271] The generated audiovisual advertisement is automatically deployed to multiple dimensions by the device, which uses an image processing library to resize and reposition the advertisement image to the appropriate ratio.

[0272] As a concrete example, consider the case where a user sets up an information page with promotional information for "New Year Sale!" and requests the generation of a banner ad. Assume the user's sentiment is positive. In this case, the server collects and analyzes past data to extract features of audiovisual ads with high selection rates. Based on the extracted features, the server trains a model to generate an optimal audiovisual ad containing the slogan "New Year Sale!". Furthermore, the server obtains positive sentiment data from the sentiment engine and adjusts the design to a brighter color tone. The generated ad is automatically displayed in multiple sizes by the device.

[0273] Example prompt sentence:

[0274] "Generate optimal audiovisual ads taking into account sentiment data. The promotion is 'Summer Sale' and users are showing positive sentiment."

[0275] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0276] Step 1:

[0277] Collecting CTR data

[0278] The server collects audiovisual ad selection rate data from databases and log files daily. The databases and log files are used as input, and the extracted data includes design elements such as layout, color, wording, image materials, theme, eye line, and typeface. This data is collected using SQL queries and Python scripts and saved as a data frame.

[0279] Specifically, the server performs the following steps:

[0280] Use a SQL query to extract daily selectivity data from the database.

[0281] Run a Python script and retrieve additional information from the log file.

[0282] The collected data is converted into a data frame using the Pandas library and saved.

[0283] Step 2:

[0284] Analyzing CTR data

[0285] The server analyzes the collected data using statistical methods. The input is the data frame collected in step 1, and the output is to extract the characteristics of audiovisual advertisements with high selection rates. Specifically, it calculates the average and standard deviation, and filters out advertisements that exceed a certain threshold.

[0286] Specifically, the server performs the following steps:

[0287] Use the Pandas library to filter high-selectivity design elements from a data frame.

[0288] Using the NumPy library, calculate the mean and standard deviation of each design element.

[0289] The characteristics of high selection rate audiovisual advertisements are extracted and saved in a list.

[0290] Step 3:

[0291] Building a feature dataset

[0292] The server constructs a training dataset based on the extracted features of high-selectivity audiovisual advertisements. The input is the feature list extracted in step 2, and the output is a training dataset (CSV file).

[0293] Specifically, the server performs the following steps:

[0294] The extracted features are organized to build a dataset containing detailed information about each design element.

[0295] Convert the dataset to a CSV file and save it.

[0296] Step 4:

[0297] Training the learning model

[0298] The server trains a learning algorithm using the constructed training dataset. The input is the CSV file constructed in step 3, and the output is the trained model. The learning algorithms used include Scikit-learn's random forest and TensorFlow's neural network.

[0299] Specifically, the server performs the following steps:

[0300] Read the CSV file and generate a training dataset.

[0301] Train random forest and neural network models using Scikit-learn and TensorFlow libraries.

[0302] Save the trained model.

[0303] Step 5:

[0304] Receiving landing page information

[0305] The user inputs the information about the user's guide page through the system interface and requests the generation of an audiovisual advertisement. The input is the user's guide page information, and the output is that this information is sent to the server.

[0306] Specifically, the user performs the following steps:

[0307] Use the web form to enter your information.

[0308] The information is sent to the server.

[0309] Step 6:

[0310] Banner generation

[0311] The server generates an optimal audiovisual advertisement by integrating the information on the guide page received from the user and the features of the high-selection rate audiovisual advertisement estimated from the learning model. The input is the landing page information and the trained model, and the output is the generated audiovisual advertisement (image file).

[0312] Specifically, the server performs the following steps:

[0313] Analyze landing page information and extract key messages and promotional content.

[0314] Use a trained model to select the best design elements.

[0315] Audiovisual advertising images are generated using PIL (Python Imaging Library).

[0316] Step 7:

[0317] Acquiring emotion data

[0318] The server receives user emotion data from the emotion processing engine. The input is the user emotion data (via API), and the output is data used to adjust design elements according to the emotion.

[0319] Specifically, the server performs the following steps:

[0320] A request is sent to the emotion processing engine via the API to obtain emotion data.

[0321] Analyze the acquired emotion data.

[0322] Step 8:

[0323] Adjusting design elements

[0324] The server adjusts the design elements of the banner based on the acquired emotional data. The input is the emotional data and the generated audiovisual advertisement, and the output is the adjusted final audiovisual advertisement.

[0325] Specifically, the server performs the following steps:

[0326] Colors and catchy slogans are dynamically changed based on emotional data.

[0327] The audiovisual advertising images are re-edited using PIL.

[0328] Step 9:

[0329] Banner size options

[0330] The terminal automatically unwraps the generated audiovisual advertisement into multiple dimensions: the input is the adjusted final audiovisual advertisement, and the output is the audiovisual advertisement (multiple image files) in different sizes.

[0331] Specifically, the device performs the following steps:

[0332] Load the generated audiovisual advertisement and resize it to the different sizes you need.

[0333] Generate audiovisual ads optimized for each size, taking into account existing placement and proportions.

[0334] (Application example 2)

[0335] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0336] Conventional banner ad generation systems focus on optimizing click-through rates (CTR), but do not take into account the user's emotional state when generating banners. Therefore, further optimization is required to maximize CTR. Furthermore, it is necessary to support dynamic design adjustments based on user emotional data when displaying ads on different devices, such as smartphones and head-mounted displays.

[0337] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0338] In this invention, the server includes a means for collecting daily click-through rate (CTR) data, a means for analyzing the characteristics of past banner advertisements, a means for extracting the characteristics of high-CTR banners, a means for generating banners using a learning model, a means for optimizing the generated banners based on landing page information, a means for automatically expanding the generated banners to multiple sizes, a means for collecting user emotion data in real time, a means for adjusting the banner design elements based on the collected emotion data, and a means for displaying banners optimized for smartphones and head-mounted displays. This allows for the generation of optimal banners that take the user's emotional state into consideration, further maximizing CTR. Furthermore, dynamic design adjustments based on user emotion data can be realized even when displaying advertisements on different devices.

[0339] "Click-through rate (CTR)" is an indicator that indicates the ratio of the number of times an ad is clicked by a user to the number of times it is displayed.

[0340] A "banner ad" is an image or graphic advertisement that appears on a web page or within an application.

[0341] A "learning model" is a model that uses algorithms to learn patterns and characteristics based on past data and makes predictions and classifications for new data.

[0342] "Landing page information" refers to the content and information of the web page that a user who clicks on an advertisement first arrives at.

[0343] "Emotion data" is data that indicates the user's emotional state, such as positive, negative, or neutral, obtained from the user's facial expression and behavior.

[0344] "Design elements" refers to the components of a banner ad, such as layout, color, catchy copy, stock photos, motifs, line of sight, and font.

[0345] A "smartphone" is a mobile information terminal that, in addition to the functions of a mobile phone, can also browse the Internet and use applications.

[0346] A "head-mounted display" is a display device worn on the head, allowing users to experience virtual reality (VR) and augmented reality (AR).

[0347] "Real-time" means that data collection and processing occur simultaneously with actual time.

[0348] A "high CTR banner" is a banner ad that has a higher-than-average click-through rate (CTR).

[0349] An "emotion engine" is software or a system that analyzes a user's emotional state and provides that data.

[0350] The system of the present invention aims to dynamically generate advertising banners based on user emotional data and maximize click-through rates (CTR). The system uses advanced machine learning models to analyze past data and apply real-time emotional information to generate optimal banners that will attract users' attention.

[0351] 1. Generate a program

[0352] The system consists of the following main components:

[0353] 1. Data Collection Module

[0354] 2. Data Analysis Module

[0355] 3. Learning Model Training Module

[0356] 4. Banner generation module

[0357] 5. Emotion Data Acquisition Module

[0358] 6. Banner Adjustment Module

[0359] 7. Display module

[0360] 2. Explain the system's processing in natural language

[0361] Data Collection Module

[0362] The server collects daily click-through rate (CTR) data, which includes the number of clicks and impressions for each banner ad.

[0363] Data Analysis Module

[0364] The server analyzes the characteristics of past banner ads to determine which features have a high CTR. This involves analyzing design elements such as banner layout, color, tagline, stock photos, motifs, line of sight, and font.

[0365] Learning Model Training Module

[0366] The server trains a learning model based on the analysis results, using TensorFlow to build a neural network to generate the optimal banner.

[0367] Banner Generation Module

[0368] Using the learning model, the server generates the optimal banner based on the landing page information, automatically selecting taglines and images that correspond to the promotional information provided by the user.

[0369] Emotion data acquisition module

[0370] The user's device (smartphone or head-mounted display) uses a camera to collect the user's emotional data in real time. OpenCV and FaceAPI are used to analyze the user's emotional state (positive, negative, neutral, etc.) from their facial expressions.

[0371] Banner Adjustment Module

[0372] The server dynamically adjusts the banner's design elements based on the acquired emotional data: for example, if the user is expressing positive emotions, the banner's colors will be brighter and the copy's tone will be more optimistic.

[0373] Display Module

[0374] The generated banners are displayed on the user's device in the optimal size and format, which includes a process to automatically generate banner images of different sizes.

[0375] 3. Specific examples and prompt sentences

[0376] Specific examples

[0377] When a landing page with promotion information "New Year Sale!" is input, the server generates an optimal banner with the tagline "50% off on all items." If the user's sentiment data is analyzed as positive, the banner design will be composed of bright colors. The displayed banner will also be optimized for the screen size of smartphones and head-mounted displays.

[0378] Prompt Sentence Examples

[0379] Input data:

[0380] Landing page information: {'message': 'New Year Sale!', 'promotion_details': '50% off on all items'}

[0381] Emotion data: {'emotion': 'positive'}

[0382] Requirements:

[0383] Always include the tagline "New Year Sale!"

[0384] Use bright colors.

[0385] Use high CTR elements when generating images.

[0386] Generation goal:

[0387] To generate an optimal advertisement banner based on a user's positive emotional state.

[0388] This system can generate advertising banners that reflect the user's emotional state, maximizing click-through rates (CTR). It can also dynamically adjust the design of ads displayed on different devices based on the user's emotional data.

[0389] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0390] Step 1:

[0391] The server collects daily click-through rate (CTR) data, which includes the number of clicks and impressions for each banner ad. The data is retrieved from databases and log files and stored in a database.

[0392] Input: Number of clicks and impressions of past banner ads

[0393] Data processing: Data collection and organization

[0394] Output: Organized CTR data

[0395] Step 2:

[0396] The server analyzes the characteristics of past banner ads to identify banner design elements (layout, color, tagline, stock photo, motif, gaze, font) that cause CTRs above a certain threshold.

[0397] Input: CTR data and banner ad characteristics information

[0398] Data calculations: statistical analysis

[0399] Output: High CTR banner feature data

[0400] Step 3:

[0401] The server extracts features of high-CTR banners and builds a learning dataset. TensorFlow is used to train the learning model. A neural network is used to build a model to predict the optimal banner design.

[0402] Input: High CTR banner feature data

[0403] Data Computing: Training Machine Learning Models

[0404] Output: A trained learning model

[0405] Step 4:

[0406] The user's device (smartphone or head-mounted display) uses a camera to collect the user's emotional data in real time. OpenCV and FaceAPI are used to analyze the user's emotional state (positive, negative, neutral, etc.) from their facial expressions.

[0407] Input: Camera image

[0408] Data Computing: Sentiment Analysis

[0409] Output: Real-time emotion data

[0410] Step 5:

[0411] Once the user enters the landing page information, the server uses this information to generate the optimal banner. The generative AI model inputs prompt text and creates a banner with the desired tagline and image.

[0412] Input: Landing page information

[0413] Data calculation: Banner generation

[0414] Output: The generated banner

[0415] Prompt Sentence Examples

[0416] Input data:

[0417] Landing page information: {'message': 'New Year Sale!', 'promotion_details': '50% off on all items'}

[0418] Emotion data: {'emotion': 'positive'}

[0419] Requirements:

[0420] Always include the tagline "New Year Sale!"

[0421] Use bright colors.

[0422] Use high CTR elements when generating images.

[0423] Generation goal:

[0424] To generate an optimal advertisement banner based on a user's positive emotional state.

[0425] Step 6:

[0426] The server dynamically adjusts the banner's design elements based on the acquired emotional data: for example, if the user is expressing positive emotions, the banner's color may be brightened and the tagline may be changed to a more positive tone.

[0427] Input: Generated banner and sentiment data

[0428] Data calculation: Dynamic banner adjustment

[0429] Output: Adjusted banner

[0430] Step 7:

[0431] The generated banner is displayed on the user's device in the optimal size and format. The server automatically generates banner images of different sizes and displays them in a format suitable for smartphones and head-mounted displays.

[0432] Input: Adjusted banner

[0433] Data processing: Size expansion

[0434] Output: Banner optimized for display

[0435] Through these steps, the system can generate optimal banners that take into account the user's emotional state and maximize click-through rates (CTR).

[0436] The specific processing unit 290 transmits the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0437] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0438] In the above embodiment, an example in which the specific process is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific process may be performed by the smart device 14.

[0439] [Second embodiment]

[0440] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0441] 3, the data processing system 210 includes the data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0442] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0443] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, and the camera 42 are also connected to the bus 52.

[0444] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0445] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0446] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0447] Fig. 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Fig. 4, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0448] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0449] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0450] In the smart glasses 214, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0451] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal."

[0452] The present invention relates to a system that collects daily click-through rate (CTR) data, analyzes the characteristics of past banner advertisements, extracts the characteristics of high-CTR banners, generates banners using a learning model, optimizes the generated banners based on landing page information, and automatically expands the generated banners into multiple sizes.

[0453] As an embodiment of the present invention, a specific program and its processing content will be described below.

[0454] Data collection and analysis phase

[0455] collection

[0456] The server collects CTR data for banner ads from a database every day. This data includes many characteristic information about past banner ads (layout, color, catchy copy, photo material, motif, gaze, font, etc.).

[0457] analysis

[0458] The server analyzes the collected data and extracts the characteristics of banners with high CTRs, using statistical methods to analyze the data and identify design elements and patterns that indicate high CTRs.

[0459] Feature extraction and training phase

[0460] Feature Extraction

[0461] The server builds a training dataset based on features extracted from high-CTR banners, including detailed information about each element (such as which layout is most effective or which color is most eye-catching).

[0462] study

[0463] This data is then used to train a learning model, using machine learning algorithms such as random forests and neural networks to build a model that predicts optimal design elements.

[0464] Banner generation phase

[0465] Receiving information

[0466] The user provides landing page information, including the landing page's key message and promotion.

[0467] Banner Generation

[0468] The server uses the learned model to generate an optimal banner based on the provided landing page information, including layout, color, tagline, stock photos, motif, line of sight, font, etc.

[0469] Size development phase

[0470] Size range

[0471] The generated banner will be automatically expanded to multiple sizes. The device will handle this process and create banner images of different sizes, for example 300x250, 728x90, 160x600, etc.

[0472] Specific examples

[0473] User operations

[0474] A user sets up a landing page with promotional information for "New Year Sale!" and requests to generate a banner ad.

[0475] Server Processing

[0476] 1. The server collects and analyzes past data and extracts the characteristics of high CTR banners.

[0477] 2. The learning model is trained based on the extracted features so that it can generate optimal banners.

[0478] 3. Based on the landing page information provided by the user, the server generates a suitable banner containing the tagline "New Year Sale!"

[0479] 4. The device will automatically expand the generated banner to various sizes such as 300x250, 728x90, and 160x600.

[0480] The above is a detailed description of the embodiment of the present invention. This system reduces the workload of designers and automatically generates optimal banners that maximize CTR.

[0481] The processing flow will be explained below.

[0482] Step 1:

[0483] The server collects daily banner ad click-through rate (CTR) data. It obtains past banner ad features such as layout, color, catchy copy, stock photos, motifs, gaze, and font, as well as their CTR, from databases and log files.

[0484] Step 2:

[0485] The server analyzes the collected data and extracts the characteristics of banners with high CTR. Specifically, it filters data for banner ads with CTRs above a certain threshold and calculates the average characteristics. During this process, it uses statistical methods to identify which design elements contribute to a high CTR.

[0486] Step 3:

[0487] The server builds a training dataset based on the extracted features. It organizes the design elements of the filtered high-CTR banner ads into a feature vector and formats it to be fed into the training model.

[0488] Step 4:

[0489] The server uses the training dataset to train a model. For example, it uses machine learning algorithms such as random forests and neural networks to build a model for generating optimal banners. This model is responsible for predicting high CTR.

[0490] Step 5:

[0491] The user inputs landing page information, including key messaging and promotional content, and requests the creation of a banner ad.

[0492] Step 6:

[0493] The server receives the provided landing page information and generates the optimal banner based on it. The landing page information and the extracted characteristics of the high CTR banner are input into the learning model to determine the optimal design elements. The generated banner includes layout, color, catchy copy, stock photos, motif, line of sight, font, etc.

[0494] Step 7:

[0495] The device receives the generated banner and automatically expands it to multiple sizes, for example, resizing it to different sizes such as 300x250, 728x90, and 160x600, and then saving it. This allows you to prepare a banner that can be used in multiple ad spaces at once.

[0496] Step 8:

[0497] The generated banners in multiple sizes are provided to the user, who can then use them in actual advertising operations to check the effectiveness of CTR.

[0498] The above is the specific processing flow of the system that automatically generates banner ads to maximize CTR and deploys them in various sizes. This series of processes makes it possible to efficiently generate optimal banners that achieve a high CTR while reducing the workload on designers.

[0499] Example 1

[0500] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0501] Conventional banner ad generation systems lack the functionality to automatically analyze and apply design elements that are expected to improve click-through rates (CTR), and require manual design and adjustment. This places a heavy workload on designers, making it difficult to create effective banner ads. The objective of this invention is to provide a system that optimizes design elements and automatically expands to multiple sizes, efficiently generating effective banner ads.

[0502] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0503] In this invention, the server includes means for collecting daily click-through rate (CTR) data, means for analyzing characteristics of past banner ads, means for extracting characteristics of high CTR banners, means for generating banners using a learning model, means for optimizing the generated banners based on landing page information, means for automatically expanding the generated banners into multiple sizes, and means for saving the optimized banners in cloud storage. This makes it possible to automatically and efficiently generate and save optimal banner ads that are expected to have a high CTR.

[0504] "Click-through rate (CTR)" indicates the percentage of clicks on a displayed ad, and is an indicator used to measure the effectiveness of an ad.

[0505] A "banner ad" is advertising content in the form of an image or video displayed on a web page, and is created with the purpose of encouraging users to click on it.

[0506] "Features" refers to identifiable characteristics of a banner ad, such as specific design elements, layout, color, catchy copy, photographic material, motif, line of sight, font, etc.

[0507] A "high CTR banner" refers to a banner ad that has a relatively high click-through rate and contains an effective design and message.

[0508] A "learning model" is a computational model that uses machine learning algorithms to learn patterns and features from data and make predictions and classifications based on future data.

[0509] A "landing page" is a specific web page accessed from a web advertisement or search engine result, and is the first page a user visits.

[0510] "Optimization" refers to the process of adjusting conditions or factors to most effectively achieve a particular objective.

[0511] "Cloud storage" refers to services that allow data to be stored remotely and made accessible over the internet.

[0512] "Sizing" refers to the process of automatically adapting a particular design to different dimensions to accommodate multiple ad slots.

[0513] This invention provides a system for optimizing the click-through rate (CTR) of banner ads. This system collects CTR data, analyzes and extracts the characteristics of banner ads that show high CTR, generates optimal banners using a learning model, optimizes the generated banners based on landing page information, and deploys them in multiple sizes.

[0514] Data collection and analysis

[0515] The server first connects to a database to collect daily banner ad CTR data. This database stores information about past banner ad CTRs, layouts, colors, taglines, stock photos, motifs, gaze patterns, fonts, etc. The server extracts this information using SQL queries and performs data cleaning, such as removing missing and outlier values ​​and normalizing the data.

[0516] The server then analyzes the collected data using statistical methods, such as calculating average CTR, variance, and correlation coefficients, to identify design elements and patterns that indicate high CTR.

[0517] Feature extraction and model training

[0518] The server extracts the features of high-CTR banners and builds a training dataset based on a multidimensional feature vector. This dataset contains the features of each banner and the corresponding CTR.

[0519] Next, the server selects a machine learning algorithm (e.g., random forest, neural network, etc.) and trains the model using the training dataset. This training model is then used to generate banners with optimal design elements to predict high CTR. Specifically, the random forest model can be trained using Scikit-learn, an educational Python library.

[0520] Banner generation and size development

[0521] The user enters the landing page information, which includes the main message and promotional content. This information is sent to the server via the terminal. For example, the promotional information entered may be "New Year Sale!"

[0522] The server takes this input information and uses a trained model to generate an optimal banner, including layout, color, tagline, stock image, motif, line of sight, font, etc. The actual banner image is generated using the Pillow library and other image processing tools.

[0523] The device then converts the generated banner to different sizes, using OpenCV or other image conversion libraries to automatically convert it to different ad sizes such as 300x250, 728x90, and 160x600.

[0524] Final generation and saving

[0525] The user can check the generated banner and request corrections if necessary. The server then saves the optimized banner to cloud storage, often using Amazon S3 for example. The saved banner is then provided to the user as a URL.

[0526] For example, the prompt is:

[0527] Generate the optimal banner based on the promotion information "New Year Sale!"

[0528] Using this system will significantly reduce the workload of designers and enable the automatic generation, deployment, and storage of banner ads that are expected to have a high CTR.

[0529] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0530] Step 1:

[0531] Data collection

[0532] The server collects daily click-through rate (CTR) data from a database. Specifically, it connects to the database and executes SQL queries to extract information related to past banner ads, such as CTR, layout, color, tagline, stock image, motif, gaze, and font. It specifies a date range and ad ID as input, and obtains a dataset containing CTR data as output. For example, it executes the query "SELECT FROM Banners WHERE date >= CURDATE() - INTERVAL 30 DAY;".

[0533] Step 2:

[0534] Data analysis

[0535] The server analyzes the collected data. It uses the CTR data collected in step 1 as input, performs data cleaning, and removes missing and outlier values. It then analyzes the data using statistical methods to identify design elements and patterns that indicate high CTR. Specifically, it calculates the average CTR, variance, correlation coefficient, etc., and obtains a list of elements that indicate high CTR as output.

[0536] Step 3:

[0537] Feature Extraction

[0538] The server extracts features from the high CTR banners identified in step 2. It uses the analysis result dataset as input and builds a feature vector for each banner. For example, it creates a matrix-format dataset containing detailed information such as which layout is most effective or which color is most eye-catching. As output, it obtains a training dataset.

[0539] Step 4:

[0540] Learning Phase

[0541] The server trains the machine learning model. It uses the training dataset constructed in step 3 as input and selects a machine learning algorithm such as a random forest or neural network. Specifically, it uses the Scikit-learn library to run the model's fit method and train the optimal predictive model. The output is the trained learning model.

[0542] Step 5:

[0543] Receiving information

[0544] The user inputs landing page information, specifically key messages and promotional content, through an input form, and the device sends this information to the server. The input is the landing page information, and the output is the data sent to the server.

[0545] Step 6:

[0546] Banner Generation

[0547] The server uses the trained model to generate the optimal banner based on the provided landing page information. Using the landing page information and the trained model as input, it calculates the banner design parameters that predict the highest CTR. Specifically, it determines the tagline, layout, color, stock photos, etc., and generates the actual banner image using the Pillow library. The generated banner image is obtained as output.

[0548] Step 7:

[0549] Size range

[0550] The device will unpack the generated banner into multiple sizes. As input, it uses the banner image generated in step 6 and automatically converts it to different sizes (e.g., 300x250, 728x90, 160x600) using an image conversion library such as OpenCV. As output, it gets the banner image for each size.

[0551] Step 8:

[0552] Final generation and saving

[0553] The server saves the generated banner in cloud storage. As input, it uses banner images of multiple sizes and uploads them to a cloud storage service. Specifically, it saves the banner images using a storage service such as Amazon S3, generates a URL for the saved banner, and provides it to the user. As output, it obtains the URL for the saved banner.

[0554] This allows the system to automatically generate optimal banner ads aiming for high CTR, and efficiently expand and save them in multiple sizes.

[0555] (Application example 1)

[0556] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0557] Conventional banner ad generation systems simply generate and display static banners based on past data. This makes it difficult to provide effective ads that maximize CTR, as it is not possible to optimize in real time or respond to dynamically changing market and user preferences. Furthermore, manual banner size deployment and optimization requires a great deal of effort, hindering efficient advertising operations.

[0558] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0559] In this invention, the server includes means for collecting daily click-through rate (CTR) data, means for analyzing the characteristics of past banner advertisements, means for extracting the characteristics of high-CTR banners, means for generating banners using a learning model, means for optimizing the generated banners based on landing page information, means for automatically expanding the generated banners to multiple sizes, and means for generating and displaying dynamically optimized banner advertisements in real time. This makes it possible to respond to changes in the market and users in real time and generate dynamically optimized banner advertisements. Furthermore, automatic expansion of banner sizes enables efficient advertising management.

[0560] "Click-through rate (CTR)" is an indicator that shows the ratio of the number of clicks to the number of times an ad is displayed.

[0561] A "banner ad" is a graphic advertisement that appears on a web page or within an application.

[0562] "Landing page information" refers to information about the content and structure of the web page to which a user is directed when they click on an advertisement.

[0563] A "learning model" is a machine learning algorithm that learns patterns and rules based on data and makes predictions and classifications.

[0564] "Feature extraction" is the process of extracting useful characteristics or patterns from data.

[0565] "Real-time" refers to the time characteristics in which data is processed and results are provided immediately.

[0566] "Optimization" is the process of adjusting parameters and conditions to achieve the best results in order to achieve a specific objective.

[0567] "Sizing" is the process of adjusting the original design to different sizes to generate multiple sizes of banners.

[0568] "Dynamic optimization" is the process of instantly adjusting the optimum state in response to changing conditions in real time.

[0569] A "prompt" refers to a sentence of instructions or questions that is input into a generative AI model.

[0570] MODE FOR CARRYING OUT THE INVENTION

[0571] To implement the present invention, a system in which a server, a terminal, and a user work in cooperation with each other is required. A specific embodiment of this system will be described below.

[0572] System Overview

[0573] The system collects daily click-through rate (CTR) data, analyzes the characteristics of past banner ads, extracts the characteristics of high-CTR banners, and generates banners using a generative AI model.The generated banners are optimized based on landing page information, automatically expanded into multiple sizes, and dynamically generated and displayed in real time.

[0574] Hardware and software used

[0575] Server: Performs primary data processing and training of learning models. Software used is Python, TensorFlow (Keras), and PIL (Pillow).

[0576] Terminal: Operated by the user, it displays the generated banner and inputs the prompt text. A smartphone or tablet is used as the interface.

[0577] Database: Uses data storage to accumulate CTR data and banner ad feature data.

[0578] Data collection and analysis phase

[0579] The server collects historical and daily banner ad click-through rate (CTR) data from a database. This data includes many characteristic information related to layout, color, tagline, stock photos, motifs, gaze, fonts, etc. The server analyzes this data using statistical methods to identify design elements and patterns that indicate high CTR.

[0580] Feature extraction and training phase

[0581] The server builds a training dataset based on features extracted from high-CTR banners, creating a dataset with detailed information about each element. The training model is then trained using this data to build a model that predicts optimal design elements using machine learning algorithms such as random forests and neural networks.

[0582] Banner generation phase

[0583] The user provides landing page information, including key messaging and promotional content. The server uses the trained model to generate an optimal banner based on the provided landing page information. The generated banner includes layout, color, tagline, stock photos, motif, line of sight, font, etc.

[0584] Size development phase

[0585] The generated banner will be automatically expanded to multiple sizes. The device will process this and create banner images of different sizes. For example, various sizes such as 300x250, 728x90, 160x600 etc. will be generated.

[0586] Dynamic optimization and real-time generation

[0587] The system dynamically optimizes to respond to changes in the market and users in real time: the server instantly adjusts banner ads based on current click-through rate data and trends, ensuring they are displayed optimally to users.

[0588] Specific examples

[0589] A user sets up a landing page with promotional information such as "Summer Sale Begins!" and requests the system to generate a banner ad. The server collects and analyzes past data to extract the characteristics of high-CTR banners. After the model is trained, the server generates an optimal banner containing the tagline "Summer Sale Begins!" based on the landing page information provided by the user. The generated banner is automatically expanded to various sizes, such as 300x250, 728x90, and 160x600, and is dynamically optimized for display depending on the device.

[0590] Prompt Sentence Examples

[0591] "Generate a 'Summer Sale!' banner that has a high click-through rate based on historical data. Resize the generated banner to 300x250, 728x90, and 160x600."

[0592] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0593] Step 1: Data collection

[0594] The server collects daily click-through rate (CTR) data from a database, which includes characteristic information about past banner ads (layout, color, catchy copy, stock photos, motifs, gaze, font, etc.). Input data is obtained from the database, and CTR data is generated as output.

[0595] Step 2: Data analysis

[0596] The server analyzes the collected CTR data using statistical methods to identify banner design elements and patterns that result in high click-through rates. The input data is CTR data, and the output is characteristic data of high CTR banners. This identified data indicates the design elements of high CTR banners.

[0597] Step 3: Feature extraction

[0598] The server builds a training dataset based on the features of high-CTR banners, including detailed information on each element (layout, color, catchy copy, stock photos, motif, gaze, font, etc.). The input data is a dataset containing the features, and the output is a training dataset.

[0599] Step 4: Model training

[0600] The server trains a learning model using a training dataset. Specifically, it builds the model using machine learning algorithms such as random forests and neural networks. The input data is the training dataset, and the output is the trained learning model.

[0601] Step 5: Receive landing page information

[0602] The user provides landing page information that will be the basis for the banner ad they want to generate. The input data is the landing page information entered by the user, and the output is the appropriately formatted landing page information.

[0603] Step 6: Generate a banner

[0604] The server uses the learning model to generate an optimal banner based on the provided landing page information. The generated banner includes layout, color, catchy copy, stock photos, motif, gaze, font, etc. The input data is the landing page information and the learning model, and the output is the generated banner.

[0605] Step 7: Size development

[0606] The server automatically expands the generated banner to multiple sizes, for example, banner images of different sizes such as 300x250, 728x90, 160x600, etc. The input data is the generated banner, and the output is the banner images resized to different sizes.

[0607] Step 8: Dynamic optimization and real-time generation

[0608] The server adjusts banner ads in real time based on current click-through rate data and trends, and displays them optimally to users. The input data is the latest click-through rate data and the generated banner, and the output is a dynamically optimized banner.

[0609] Step 9: Displaying the banner

[0610] Users can view the optimized banner ads on their smartphones or other devices and request modifications or regeneration as necessary. The input data is the dynamically optimized banner, and the output is the displayed banner ad.

[0611] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[0612] This invention relates to a system that combines click-through rate (CTR) data collection and analysis, banner generation using a learning model, optimization based on landing page information, and an emotion engine. The system aims to further maximize CTR by adjusting banner design elements based on user emotion data.

[0613] Data collection and analysis phase

[0614] collection

[0615] The server collects banner ad CTR data daily, retrieving information about past banner ad features such as layout, color, catchy copy, stock photos, motifs, gaze, and font, as well as their CTR, from databases and log files.

[0616] analysis

[0617] The server analyzes the collected data and extracts the characteristics of banners with high CTR. Specifically, it filters data for banner ads with CTRs above a certain threshold and calculates the average characteristics. In this process, it uses statistical methods to identify which design elements contribute to a high CTR.

[0618] Feature extraction and training phase

[0619] Feature Extraction

[0620] The server builds a training dataset based on the features extracted from high-CTR banners, including detailed information about each element (layout, color, catchy copy, stock photos, motif, gaze, and font).

[0621] study

[0622] This data is then used to train a learning model, using machine learning algorithms such as random forests and neural networks to build a model that predicts optimal design elements.

[0623] Banner generation phase

[0624] Receiving information

[0625] The user inputs landing page information, including key messaging and promotional content, and requests the creation of a banner ad.

[0626] Banner Generation

[0627] The server generates an optimal banner based on the landing page information and the characteristics of high-CTR banners obtained from the learning model, including layout, color, catchy copy, stock photos, motif, line of sight, font, etc.

[0628] Collaboration with emotion engine

[0629] Acquiring emotion data

[0630] The server obtains the user's emotional data from the emotion engine, including whether the user is in a positive, negative, or neutral emotional state.

[0631] Adjusting design elements

[0632] The server adjusts the banner's design elements based on the acquired emotional data, for example, using bright colors when the user is expressing positive emotions and muted colors when expressing negative emotions, and dynamically changes the tagline accordingly.

[0633] Size development phase

[0634] Size range

[0635] The generated banner will be automatically expanded to multiple sizes. The device will handle this process and create banner images of different sizes, for example 300x250, 728x90, 160x600, etc.

[0636] Specific examples

[0637] User operations

[0638] Assume that a user sets up a landing page with promotion information for "New Year Sale!" and requests the generation of a banner ad. Also, assume that the user's sentiment is positive.

[0639] Server Processing

[0640] 1. The server collects and analyzes past data and extracts the characteristics of high CTR banners.

[0641] 2. The learning model is trained based on the extracted features so that it can generate optimal banners.

[0642] 3. Based on the landing page information provided by the user, the server generates a suitable banner containing the tagline "New Year Sale!"

[0643] 4. The server retrieves the user's emotional data from the emotion engine and adjusts the design by using brighter colors if the emotion indicates a positive one.

[0644] 5. The device will automatically expand the generated banner to various sizes such as 300x250, 728x90, and 160x600.

[0645] The above is a detailed description of the embodiment of the present invention. This system reduces the workload of designers, efficiently generates optimal banners that maximize CTR, and also enables design adjustments based on user emotions.

[0646] The processing flow will be explained below.

[0647] Step 1:

[0648] The server collects daily banner ad click-through rate (CTR) data. It obtains past banner ad features such as layout, color, catchy copy, stock photos, motifs, gaze, and font, as well as their CTR, from databases and log files.

[0649] Step 2:

[0650] The server analyzes the collected data and extracts the characteristics of banners with high CTR. Specifically, it filters data for banner ads with CTRs above a certain threshold and calculates the average characteristics. In this process, it uses statistical methods to identify which design elements contribute to a high CTR.

[0651] Step 3:

[0652] The server builds a training dataset based on the features extracted from high-CTR banners, including detailed information about each element (layout, color, catchy copy, stock photos, motif, gaze, font, etc.).

[0653] Step 4:

[0654] The server uses the constructed training dataset to train a model. For example, it uses machine learning algorithms such as random forests and neural networks to build a model for generating optimal banners. This model plays a role in predicting high CTR.

[0655] Step 5:

[0656] The user inputs landing page information, including key messaging and promotional content, and requests the creation of a banner ad.

[0657] Step 6:

[0658] The server receives the provided landing page information and generates the optimal banner based on it. The landing page information and the extracted characteristics of the high CTR banner are input into the learning model to determine the optimal design elements. The generated banner includes layout, color, catchy copy, stock photos, motif, line of sight, font, etc.

[0659] Step 7:

[0660] The server obtains the user's emotional data from the emotion engine, which includes the user's emotional state, such as positive, negative, or neutral.

[0661] Step 8:

[0662] The server adjusts the banner's design elements based on the acquired emotional data, for example, using bright colors when the user is expressing positive emotions and muted colors when expressing negative emotions. Furthermore, the tagline dynamically changes depending on the emotion.

[0663] Step 9:

[0664] The device receives the generated banner and automatically expands it to multiple sizes, for example, resizing it to different sizes such as 300x250, 728x90, and 160x600, and then saving it. This allows a banner that can be used in multiple ad spaces to be prepared at once.

[0665] Step 10:

[0666] The generated banners in multiple sizes are provided to the user, who can then use them in actual advertising operations to check the effectiveness of CTR.

[0667] As a concrete example, suppose a user sets up a landing page with promotional information for "New Year Sale!" and requests the generation of a banner ad. Let's also assume that the user's sentiment is positive. The server uses a learning model and sentiment engine to generate an optimal banner that combines the "New Year Sale!" tagline with bright colors. The device then expands this banner into multiple sizes and provides it to the user. This series of processes allows users to efficiently use banner ads that aim for a high CTR.

[0668] Example 2

[0669] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0670] Conventional audiovisual advertisement generation systems do not take into account the user's emotional response and therefore have difficulty generating optimal advertisements. Furthermore, the process of automatically expanding the generated advertisements into multiple sizes is also time-consuming. Therefore, a new system for creating more effective audiovisual advertisements and automatically expanding them is needed.

[0671] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0672] In this invention, the server includes means for collecting daily selection rate data, means for analyzing characteristics of past audiovisual advertisements, means for extracting characteristics of audiovisual advertisements with high selection rates, means for generating audiovisual advertisements using a learning algorithm, means for optimizing the generated audiovisual advertisements based on guide page information, means for adjusting design elements based on user emotions using an emotion processing engine, and means for automatically expanding the generated audiovisual advertisements into multiple dimensions, thereby enabling efficient generation of optimal audiovisual advertisements according to user emotions and automatic expansion thereof.

[0673] "Selection rate data" refers to data that indicates the rate at which users click or select an audiovisual advertisement displayed.

[0674] "Audiovisual advertising" refers to advertising that uses both visual and auditory sensations to provide information to users.

[0675] A "learning algorithm" is a machine learning method for predicting optimal design elements for audiovisual advertisements based on collected data.

[0676] "Guide page information" refers to the content of a web page that contains promotional information for a particular product or service.

[0677] An "emotion processing engine" is a system that analyzes a user's emotional state and adjusts design elements based on the results.

[0678] "Design elements" are the components of an audiovisual advertisement, and specifically include layout, color, wording, image materials, theme, line of sight, and font.

[0679] "Dimensions" refers to the physical size and proportions of an audiovisual advertisement.

[0680] "User" means an individual or organization that utilizes the system to generate and display audiovisual advertisements.

[0681] This invention relates to a system that collects and analyzes audiovisual advertisement selection rate data, combines a learning algorithm and an emotion processing engine to generate optimal audiovisual advertisements, and optimizes and automatically deploys the advertisements based on guide page information.

[0682] The hardware required to implement this system includes a database server, a web server, an emotion processing engine server, and a terminal. The software used includes a database management system (e.g., MySQL), a data processing library (e.g., Pandas, NumPy), a machine learning library (e.g., Scikit-learn, TensorFlow), and an image processing library (e.g., PIL).

[0683] The specific program processing is as follows.

[0684] The server collects audiovisual ad selection rate data daily from a database and log files, including design elements such as layout, color, wording, image materials, theme, eye direction, and typeface. The server uses SQL queries to extract relevant data from the database and Python scripts to retrieve additional information from the log files.

[0685] The server then analyzes the collected data using statistical methods. It extracts design elements of audiovisual advertisements with high selection rates, filters advertisements that exceed a certain threshold, and calculates average features. Specifically, the server creates a data frame using the Pandas library and calculates statistical information such as the mean and standard deviation using the NumPy library.

[0686] The server builds a training dataset based on features extracted from high-selection audiovisual ads. This dataset includes detailed information about each design element. The server saves the extracted features in a CSV file, which is then fed into machine learning algorithms. The training algorithms include random forests and neural networks, which are trained using Scikit-learn and TensorFlow.

[0687] The user inputs information about the page through the system interface and requests the generation of an audiovisual advertisement. The input information includes the promotional content and key messages. The user submits this information using a web form.

[0688] The server combines the information on the user's guide page with the characteristics of high-selection audiovisual ads estimated from the learning model to generate optimal audiovisual ads. Specifically, the server uses PIL (Python Imaging Library) to combine elements such as layout and catchy copy to generate ad images. A generative AI model is also used in this process.

[0689] The server obtains user emotional data in real time from an emotion processing engine. This data includes the user's emotional state, such as positive, negative, or neutral. The server adjusts design elements based on this emotional data. For example, bright colors are used to indicate positive emotions, and muted colors to indicate negative emotions. The tagline also changes dynamically depending on the emotion.

[0690] The generated audiovisual advertisement is automatically deployed to multiple dimensions by the device, which uses an image processing library to resize and reposition the advertisement image to the appropriate ratio.

[0691] As a concrete example, consider the case where a user sets up an information page with promotional information for "New Year Sale!" and requests the generation of a banner ad. Assume the user's sentiment is positive. In this case, the server collects and analyzes past data to extract features of audiovisual ads with high selection rates. Based on the extracted features, the server trains a model to generate an optimal audiovisual ad containing the slogan "New Year Sale!". Furthermore, the server obtains positive sentiment data from the sentiment engine and adjusts the design to a brighter color tone. The generated ad is automatically displayed in multiple sizes by the device.

[0692] Example prompt sentence:

[0693] "Generate optimal audiovisual ads taking into account sentiment data. The promotion is 'Summer Sale' and users are showing positive sentiment."

[0694] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0695] Step 1:

[0696] Collecting CTR data

[0697] The server collects audiovisual ad selection rate data from databases and log files daily. The databases and log files are used as input, and the extracted data includes design elements such as layout, color, wording, image materials, theme, eye line, and typeface. This data is collected using SQL queries and Python scripts and saved as a data frame.

[0698] Specifically, the server performs the following steps:

[0699] Use a SQL query to extract daily selectivity data from the database.

[0700] Run a Python script and retrieve additional information from the log file.

[0701] The collected data is converted into a data frame using the Pandas library and saved.

[0702] Step 2:

[0703] Analyzing CTR data

[0704] The server analyzes the collected data using statistical methods. The input is the data frame collected in step 1, and the output is to extract the characteristics of audiovisual advertisements with high selection rates. Specifically, it calculates the average and standard deviation, and filters out advertisements that exceed a certain threshold.

[0705] Specifically, the server performs the following steps:

[0706] Use the Pandas library to filter high-selectivity design elements from a data frame.

[0707] Using the NumPy library, calculate the mean and standard deviation of each design element.

[0708] The characteristics of high selection rate audiovisual advertisements are extracted and saved in a list.

[0709] Step 3:

[0710] Building a feature dataset

[0711] The server constructs a training dataset based on the extracted features of high-selectivity audiovisual advertisements. The input is the feature list extracted in step 2, and the output is a training dataset (CSV file).

[0712] Specifically, the server performs the following steps:

[0713] The extracted features are organized to build a dataset containing detailed information about each design element.

[0714] Convert the dataset to a CSV file and save it.

[0715] Step 4:

[0716] Training the learning model

[0717] The server trains a learning algorithm using the constructed training dataset. The input is the CSV file constructed in step 3, and the output is the trained model. The learning algorithms used include Scikit-learn's random forest and TensorFlow's neural network.

[0718] Specifically, the server performs the following steps:

[0719] Read the CSV file and generate a training dataset.

[0720] Train random forest and neural network models using Scikit-learn and TensorFlow libraries.

[0721] Save the trained model.

[0722] Step 5:

[0723] Receiving landing page information

[0724] The user inputs the information about the user's guide page through the system interface and requests the generation of an audiovisual advertisement. The input is the user's guide page information, and the output is that this information is sent to the server.

[0725] Specifically, the user performs the following steps:

[0726] Use the web form to enter your information.

[0727] The information is sent to the server.

[0728] Step 6:

[0729] Banner generation

[0730] The server generates an optimal audiovisual advertisement by integrating the information on the guide page received from the user and the features of the high-selection rate audiovisual advertisement estimated from the learning model. The input is the landing page information and the trained model, and the output is the generated audiovisual advertisement (image file).

[0731] Specifically, the server performs the following steps:

[0732] Analyze landing page information and extract key messages and promotional content.

[0733] Use a trained model to select the best design elements.

[0734] Audiovisual advertising images are generated using PIL (Python Imaging Library).

[0735] Step 7:

[0736] Acquiring emotion data

[0737] The server receives user emotion data from the emotion processing engine. The input is the user emotion data (via API), and the output is data used to adjust design elements according to the emotion.

[0738] Specifically, the server performs the following steps:

[0739] A request is sent to the emotion processing engine via the API to obtain emotion data.

[0740] Analyze the acquired emotion data.

[0741] Step 8:

[0742] Adjusting design elements

[0743] The server adjusts the design elements of the banner based on the acquired emotional data. The input is the emotional data and the generated audiovisual advertisement, and the output is the adjusted final audiovisual advertisement.

[0744] Specifically, the server performs the following steps:

[0745] Colors and catchy slogans are dynamically changed based on emotional data.

[0746] The audiovisual advertising images are re-edited using PIL.

[0747] Step 9:

[0748] Banner size options

[0749] The terminal automatically unwraps the generated audiovisual advertisement into multiple dimensions: the input is the adjusted final audiovisual advertisement, and the output is the audiovisual advertisement (multiple image files) in different sizes.

[0750] Specifically, the device performs the following steps:

[0751] Load the generated audiovisual advertisement and resize it to the different sizes you need.

[0752] Generate audiovisual ads optimized for each size, taking into account existing placement and proportions.

[0753] (Application example 2)

[0754] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0755] Conventional banner ad generation systems focus on optimizing click-through rates (CTR), but do not take into account the user's emotional state when generating banners. Therefore, further optimization is required to maximize CTR. Furthermore, it is necessary to support dynamic design adjustments based on user emotional data when displaying ads on different devices, such as smartphones and head-mounted displays.

[0756] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0757] In this invention, the server includes a means for collecting daily click-through rate (CTR) data, a means for analyzing the characteristics of past banner advertisements, a means for extracting the characteristics of high-CTR banners, a means for generating banners using a learning model, a means for optimizing the generated banners based on landing page information, a means for automatically expanding the generated banners to multiple sizes, a means for collecting user emotion data in real time, a means for adjusting the banner design elements based on the collected emotion data, and a means for displaying banners optimized for smartphones and head-mounted displays. This allows for the generation of optimal banners that take the user's emotional state into consideration, further maximizing CTR. Furthermore, dynamic design adjustments based on user emotion data can be realized even when displaying advertisements on different devices.

[0758] "Click-through rate (CTR)" is an indicator that indicates the ratio of the number of times an ad is clicked by a user to the number of times it is displayed.

[0759] A "banner ad" is an image or graphic advertisement that appears on a web page or within an application.

[0760] A "learning model" is a model that uses algorithms to learn patterns and characteristics based on past data and makes predictions and classifications for new data.

[0761] "Landing page information" refers to the content and information of the web page that a user who clicks on an advertisement first arrives at.

[0762] "Emotion data" is data that indicates the user's emotional state, such as positive, negative, or neutral, obtained from the user's facial expression and behavior.

[0763] "Design elements" refers to the components of a banner ad, such as layout, color, catchy copy, stock photos, motifs, line of sight, and font.

[0764] A "smartphone" is a mobile information terminal that, in addition to the functions of a mobile phone, can also browse the Internet and use applications.

[0765] A "head-mounted display" is a display device worn on the head, allowing users to experience virtual reality (VR) and augmented reality (AR).

[0766] "Real-time" means that data collection and processing occur simultaneously with actual time.

[0767] A "high CTR banner" is a banner ad that has a higher-than-average click-through rate (CTR).

[0768] An "emotion engine" is software or a system that analyzes a user's emotional state and provides that data.

[0769] The system of the present invention aims to dynamically generate advertising banners based on user emotional data and maximize click-through rates (CTR). The system uses advanced machine learning models to analyze past data and apply real-time emotional information to generate optimal banners that will attract users' attention.

[0770] 1. Generate a program

[0771] The system consists of the following main components:

[0772] 1. Data Collection Module

[0773] 2. Data Analysis Module

[0774] 3. Learning Model Training Module

[0775] 4. Banner generation module

[0776] 5. Emotion Data Acquisition Module

[0777] 6. Banner Adjustment Module

[0778] 7. Display module

[0779] 2. Explain the system's processing in natural language

[0780] Data Collection Module

[0781] The server collects daily click-through rate (CTR) data, which includes the number of clicks and impressions for each banner ad.

[0782] Data Analysis Module

[0783] The server analyzes the characteristics of past banner ads to determine which features have a high CTR. This involves analyzing design elements such as banner layout, color, tagline, stock photos, motifs, line of sight, and font.

[0784] Learning Model Training Module

[0785] The server trains a learning model based on the analysis results, using TensorFlow to build a neural network to generate the optimal banner.

[0786] Banner Generation Module

[0787] Using the learning model, the server generates the optimal banner based on the landing page information, automatically selecting taglines and images that correspond to the promotional information provided by the user.

[0788] Emotion data acquisition module

[0789] The user's device (smartphone or head-mounted display) uses a camera to collect the user's emotional data in real time. OpenCV and FaceAPI are used to analyze the user's emotional state (positive, negative, neutral, etc.) from their facial expressions.

[0790] Banner Adjustment Module

[0791] The server dynamically adjusts the banner's design elements based on the acquired emotional data: for example, if the user is expressing positive emotions, the banner's colors will be brighter and the copy's tone will be more optimistic.

[0792] Display Module

[0793] The generated banners are displayed on the user's device in the optimal size and format, which includes a process to automatically generate banner images of different sizes.

[0794] 3. Specific examples and prompt sentences

[0795] Specific examples

[0796] When a landing page with promotion information "New Year Sale!" is input, the server generates an optimal banner with the tagline "50% off on all items." If the user's sentiment data is analyzed as positive, the banner design will be composed of bright colors. The displayed banner will also be optimized for the screen size of smartphones and head-mounted displays.

[0797] Prompt Sentence Examples

[0798] Input data:

[0799] Landing page information: {'message': 'New Year Sale!', 'promotion_details': '50% off on all items'}

[0800] Emotion data: {'emotion': 'positive'}

[0801] Requirements:

[0802] Always include the tagline "New Year Sale!"

[0803] Use bright colors.

[0804] Use high CTR elements when generating images.

[0805] Generation goal:

[0806] To generate an optimal advertisement banner based on a user's positive emotional state.

[0807] This system can generate advertising banners that reflect the user's emotional state, maximizing click-through rates (CTR). It can also dynamically adjust the design of ads displayed on different devices based on the user's emotional data.

[0808] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0809] Step 1:

[0810] The server collects daily click-through rate (CTR) data, which includes the number of clicks and impressions for each banner ad. The data is retrieved from databases and log files and stored in a database.

[0811] Input: Number of clicks and impressions of past banner ads

[0812] Data processing: Data collection and organization

[0813] Output: Organized CTR data

[0814] Step 2:

[0815] The server analyzes the characteristics of past banner ads to identify banner design elements (layout, color, tagline, stock photo, motif, gaze, font) that cause CTRs above a certain threshold.

[0816] Input: CTR data and banner ad characteristics information

[0817] Data calculations: statistical analysis

[0818] Output: High CTR banner feature data

[0819] Step 3:

[0820] The server extracts features of high-CTR banners and builds a learning dataset. TensorFlow is used to train the learning model. A neural network is used to build a model to predict the optimal banner design.

[0821] Input: High CTR banner feature data

[0822] Data Computing: Training Machine Learning Models

[0823] Output: A trained learning model

[0824] Step 4:

[0825] The user's device (smartphone or head-mounted display) uses a camera to collect the user's emotional data in real time. OpenCV and FaceAPI are used to analyze the user's emotional state (positive, negative, neutral, etc.) from their facial expressions.

[0826] Input: Camera image

[0827] Data Computing: Sentiment Analysis

[0828] Output: Real-time emotion data

[0829] Step 5:

[0830] Once the user enters the landing page information, the server uses this information to generate the optimal banner. The generative AI model inputs prompt text and creates a banner with the desired tagline and image.

[0831] Input: Landing page information

[0832] Data calculation: Banner generation

[0833] Output: The generated banner

[0834] Prompt Sentence Examples

[0835] Input data:

[0836] Landing page information: {'message': 'New Year Sale!', 'promotion_details': '50% off on all items'}

[0837] Emotion data: {'emotion': 'positive'}

[0838] Requirements:

[0839] Always include the tagline "New Year Sale!"

[0840] Use bright colors.

[0841] Use high CTR elements when generating images.

[0842] Generation goal:

[0843] To generate an optimal advertisement banner based on a user's positive emotional state.

[0844] Step 6:

[0845] The server dynamically adjusts the banner's design elements based on the acquired emotional data: for example, if the user is expressing positive emotions, the banner's color may be brightened and the tagline may be changed to a more positive tone.

[0846] Input: Generated banner and sentiment data

[0847] Data calculation: Dynamic banner adjustment

[0848] Output: Adjusted banner

[0849] Step 7:

[0850] The generated banner is displayed on the user's device in the optimal size and format. The server automatically generates banner images of different sizes and displays them in a format suitable for smartphones and head-mounted displays.

[0851] Input: Adjusted banner

[0852] Data processing: Size expansion

[0853] Output: Banner optimized for display

[0854] Through these steps, the system can generate optimal banners that take into account the user's emotional state and maximize click-through rates (CTR).

[0855] The specific processing unit 290 transmits the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0856] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0857] In the above embodiment, an example in which the specific processing is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the smart glasses 214.

[0858] [Third embodiment]

[0859] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0860] 5, the data processing system 310 includes the data processing device 12 and a headset type terminal 314. An example of the data processing device 12 is a server.

[0861] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0862] The headset type terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a display 343. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the display 343 are also connected to the bus 52.

[0863] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0864] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0865] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0866] Fig. 6 shows an example of the main functions of the data processing device 12 and the headset type terminal 314. As shown in Fig. 6, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0867] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0868] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0869] In the headset type terminal 314, a reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0870] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the headset type terminal 314 will be referred to as the "terminal."

[0871] The present invention relates to a system that collects daily click-through rate (CTR) data, analyzes the characteristics of past banner advertisements, extracts the characteristics of high-CTR banners, generates banners using a learning model, optimizes the generated banners based on landing page information, and automatically expands the generated banners into multiple sizes.

[0872] As an embodiment of the present invention, a specific program and its processing content will be described below.

[0873] Data collection and analysis phase

[0874] collection

[0875] The server collects CTR data for banner ads from a database every day. This data includes many characteristic information about past banner ads (layout, color, catchy copy, photo material, motif, gaze, font, etc.).

[0876] analysis

[0877] The server analyzes the collected data and extracts the characteristics of banners with high CTRs, using statistical methods to analyze the data and identify design elements and patterns that indicate high CTRs.

[0878] Feature extraction and training phase

[0879] Feature Extraction

[0880] The server builds a training dataset based on features extracted from high-CTR banners, including detailed information about each element (such as which layout is most effective or which color is most eye-catching).

[0881] study

[0882] This data is then used to train a learning model, using machine learning algorithms such as random forests and neural networks to build a model that predicts optimal design elements.

[0883] Banner generation phase

[0884] Receiving information

[0885] The user provides landing page information, including the landing page's key message and promotion.

[0886] Banner Generation

[0887] The server uses the learned model to generate an optimal banner based on the provided landing page information, including layout, color, tagline, stock photos, motif, line of sight, font, etc.

[0888] Size development phase

[0889] Size range

[0890] The generated banner will be automatically expanded to multiple sizes. The device will handle this process and create banner images of different sizes, for example 300x250, 728x90, 160x600, etc.

[0891] Specific examples

[0892] User operations

[0893] A user sets up a landing page with promotional information for "New Year Sale!" and requests to generate a banner ad.

[0894] Server Processing

[0895] 1. The server collects and analyzes past data and extracts the characteristics of high CTR banners.

[0896] 2. The learning model is trained based on the extracted features so that it can generate optimal banners.

[0897] 3. Based on the landing page information provided by the user, the server generates a suitable banner containing the tagline "New Year Sale!"

[0898] 4. The device will automatically expand the generated banner to various sizes such as 300x250, 728x90, and 160x600.

[0899] The above is a detailed description of the embodiment of the present invention. This system reduces the workload of designers and automatically generates optimal banners that maximize CTR.

[0900] The processing flow will be explained below.

[0901] Step 1:

[0902] The server collects daily banner ad click-through rate (CTR) data. It obtains past banner ad features such as layout, color, catchy copy, stock photos, motifs, gaze, and font, as well as their CTR, from databases and log files.

[0903] Step 2:

[0904] The server analyzes the collected data and extracts the characteristics of banners with high CTR. Specifically, it filters data for banner ads with CTRs above a certain threshold and calculates the average characteristics. During this process, it uses statistical methods to identify which design elements contribute to a high CTR.

[0905] Step 3:

[0906] The server builds a training dataset based on the extracted features. It organizes the design elements of the filtered high-CTR banner ads into a feature vector and formats it to be fed into the training model.

[0907] Step 4:

[0908] The server uses the training dataset to train a model. For example, it uses machine learning algorithms such as random forests and neural networks to build a model for generating optimal banners. This model is responsible for predicting high CTR.

[0909] Step 5:

[0910] The user inputs landing page information, including key messaging and promotional content, and requests the creation of a banner ad.

[0911] Step 6:

[0912] The server receives the provided landing page information and generates the optimal banner based on it. The landing page information and the extracted characteristics of the high CTR banner are input into the learning model to determine the optimal design elements. The generated banner includes layout, color, catchy copy, stock photos, motif, line of sight, font, etc.

[0913] Step 7:

[0914] The device receives the generated banner and automatically expands it to multiple sizes, for example, resizing it to different sizes such as 300x250, 728x90, and 160x600, and then saving it. This allows you to prepare a banner that can be used in multiple ad spaces at once.

[0915] Step 8:

[0916] The generated banners in multiple sizes are provided to the user, who can then use them in actual advertising operations to check the effectiveness of CTR.

[0917] The above is the specific processing flow of the system that automatically generates banner ads to maximize CTR and deploys them in various sizes. This series of processes makes it possible to efficiently generate optimal banners that achieve a high CTR while reducing the workload on designers.

[0918] Example 1

[0919] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0920] Conventional banner ad generation systems lack the functionality to automatically analyze and apply design elements that are expected to improve click-through rates (CTR), and require manual design and adjustment. This places a heavy workload on designers, making it difficult to create effective banner ads. The objective of this invention is to provide a system that optimizes design elements and automatically expands to multiple sizes, efficiently generating effective banner ads.

[0921] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0922] In this invention, the server includes means for collecting daily click-through rate (CTR) data, means for analyzing characteristics of past banner ads, means for extracting characteristics of high CTR banners, means for generating banners using a learning model, means for optimizing the generated banners based on landing page information, means for automatically expanding the generated banners into multiple sizes, and means for saving the optimized banners in cloud storage. This makes it possible to automatically and efficiently generate and save optimal banner ads that are expected to have a high CTR.

[0923] "Click-through rate (CTR)" indicates the percentage of clicks on a displayed ad, and is an indicator used to measure the effectiveness of an ad.

[0924] A "banner ad" is advertising content in the form of an image or video displayed on a web page, and is created with the purpose of encouraging users to click on it.

[0925] "Features" refers to identifiable characteristics of a banner ad, such as specific design elements, layout, color, catchy copy, photographic material, motif, line of sight, font, etc.

[0926] A "high CTR banner" refers to a banner ad that has a relatively high click-through rate and contains an effective design and message.

[0927] A "learning model" is a computational model that uses machine learning algorithms to learn patterns and features from data and make predictions and classifications based on future data.

[0928] A "landing page" is a specific web page accessed from a web advertisement or search engine result, and is the first page a user visits.

[0929] "Optimization" refers to the process of adjusting conditions or factors to most effectively achieve a particular objective.

[0930] "Cloud storage" refers to services that allow data to be stored remotely and made accessible over the internet.

[0931] "Sizing" refers to the process of automatically adapting a particular design to different dimensions to accommodate multiple ad slots.

[0932] This invention provides a system for optimizing the click-through rate (CTR) of banner ads. This system collects CTR data, analyzes and extracts the characteristics of banner ads that show high CTR, generates optimal banners using a learning model, optimizes the generated banners based on landing page information, and deploys them in multiple sizes.

[0933] Data collection and analysis

[0934] The server first connects to a database to collect daily banner ad CTR data. This database stores information about past banner ad CTRs, layouts, colors, taglines, stock photos, motifs, gaze patterns, fonts, etc. The server extracts this information using SQL queries and performs data cleaning, such as removing missing and outlier values ​​and normalizing the data.

[0935] The server then analyzes the collected data using statistical methods, such as calculating average CTR, variance, and correlation coefficients, to identify design elements and patterns that indicate high CTR.

[0936] Feature extraction and model training

[0937] The server extracts the features of high-CTR banners and builds a training dataset based on a multidimensional feature vector. This dataset contains the features of each banner and the corresponding CTR.

[0938] Next, the server selects a machine learning algorithm (e.g., random forest, neural network, etc.) and trains the model using the training dataset. This training model is then used to generate banners with optimal design elements to predict high CTR. Specifically, the random forest model can be trained using Scikit-learn, an educational Python library.

[0939] Banner generation and size development

[0940] The user enters the landing page information, which includes the main message and promotional content. This information is sent to the server via the terminal. For example, the promotional information entered may be "New Year Sale!"

[0941] The server takes this input information and uses a trained model to generate an optimal banner, including layout, color, tagline, stock image, motif, line of sight, font, etc. The actual banner image is generated using the Pillow library and other image processing tools.

[0942] The device then converts the generated banner to different sizes, using OpenCV or other image conversion libraries to automatically convert it to different ad sizes such as 300x250, 728x90, and 160x600.

[0943] Final generation and saving

[0944] The user can check the generated banner and request corrections if necessary. The server then saves the optimized banner to cloud storage, often using Amazon S3 for example. The saved banner is then provided to the user as a URL.

[0945] For example, the prompt is:

[0946] Generate the optimal banner based on the promotion information "New Year Sale!"

[0947] Using this system will significantly reduce the workload of designers and enable the automatic generation, deployment, and storage of banner ads that are expected to have a high CTR.

[0948] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0949] Step 1:

[0950] Data collection

[0951] The server collects daily click-through rate (CTR) data from a database. Specifically, it connects to the database and executes SQL queries to extract information related to past banner ads, such as CTR, layout, color, tagline, stock image, motif, gaze, and font. It specifies a date range and ad ID as input, and obtains a dataset containing CTR data as output. For example, it executes the query "SELECT FROM Banners WHERE date >= CURDATE() - INTERVAL 30 DAY;".

[0952] Step 2:

[0953] Data analysis

[0954] The server analyzes the collected data. It uses the CTR data collected in step 1 as input, performs data cleaning, and removes missing and outlier values. It then analyzes the data using statistical methods to identify design elements and patterns that indicate high CTR. Specifically, it calculates the average CTR, variance, correlation coefficient, etc., and obtains a list of elements that indicate high CTR as output.

[0955] Step 3:

[0956] Feature Extraction

[0957] The server extracts features from the high CTR banners identified in step 2. It uses the analysis result dataset as input and builds a feature vector for each banner. For example, it creates a matrix-format dataset containing detailed information such as which layout is most effective or which color is most eye-catching. As output, it obtains a training dataset.

[0958] Step 4:

[0959] Learning Phase

[0960] The server trains the machine learning model. It uses the training dataset constructed in step 3 as input and selects a machine learning algorithm such as a random forest or neural network. Specifically, it uses the Scikit-learn library to run the model's fit method and train the optimal predictive model. The output is the trained learning model.

[0961] Step 5:

[0962] Receiving information

[0963] The user inputs landing page information, specifically key messages and promotional content, through an input form, and the device sends this information to the server. The input is the landing page information, and the output is the data sent to the server.

[0964] Step 6:

[0965] Banner Generation

[0966] The server uses the trained model to generate the optimal banner based on the provided landing page information. Using the landing page information and the trained model as input, it calculates the banner design parameters that predict the highest CTR. Specifically, it determines the tagline, layout, color, stock photos, etc., and generates the actual banner image using the Pillow library. The generated banner image is obtained as output.

[0967] Step 7:

[0968] Size range

[0969] The device will unpack the generated banner into multiple sizes. As input, it uses the banner image generated in step 6 and automatically converts it to different sizes (e.g., 300x250, 728x90, 160x600) using an image conversion library such as OpenCV. As output, it gets the banner image for each size.

[0970] Step 8:

[0971] Final generation and saving

[0972] The server saves the generated banner in cloud storage. As input, it uses banner images of multiple sizes and uploads them to a cloud storage service. Specifically, it saves the banner images using a storage service such as Amazon S3, generates a URL for the saved banner, and provides it to the user. As output, it obtains the URL for the saved banner.

[0973] This allows the system to automatically generate optimal banner ads aiming for high CTR, and efficiently expand and save them in multiple sizes.

[0974] (Application example 1)

[0975] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0976] Conventional banner ad generation systems simply generate and display static banners based on past data. This makes it difficult to provide effective ads that maximize CTR, as it is not possible to optimize in real time or respond to dynamically changing market and user preferences. Furthermore, manual banner size deployment and optimization requires a great deal of effort, hindering efficient advertising operations.

[0977] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0978] In this invention, the server includes means for collecting daily click-through rate (CTR) data, means for analyzing the characteristics of past banner advertisements, means for extracting the characteristics of high-CTR banners, means for generating banners using a learning model, means for optimizing the generated banners based on landing page information, means for automatically expanding the generated banners to multiple sizes, and means for generating and displaying dynamically optimized banner advertisements in real time. This makes it possible to respond to changes in the market and users in real time and generate dynamically optimized banner advertisements. Furthermore, automatic expansion of banner sizes enables efficient advertising management.

[0979] "Click-through rate (CTR)" is an indicator that shows the ratio of the number of clicks to the number of times an ad is displayed.

[0980] A "banner ad" is a graphic advertisement that appears on a web page or within an application.

[0981] "Landing page information" refers to information about the content and structure of the web page to which a user is directed when they click on an advertisement.

[0982] A "learning model" is a machine learning algorithm that learns patterns and rules based on data and makes predictions and classifications.

[0983] "Feature extraction" is the process of extracting useful characteristics or patterns from data.

[0984] "Real-time" refers to the time characteristics in which data is processed and results are provided immediately.

[0985] "Optimization" is the process of adjusting parameters and conditions to achieve the best results in order to achieve a specific objective.

[0986] "Sizing" is the process of adjusting the original design to different sizes to generate multiple sizes of banners.

[0987] "Dynamic optimization" is the process of instantly adjusting the optimum state in response to changing conditions in real time.

[0988] A "prompt" refers to a sentence of instructions or questions that is input into a generative AI model.

[0989] MODE FOR CARRYING OUT THE INVENTION

[0990] To implement the present invention, a system in which a server, a terminal, and a user work in cooperation with each other is required. A specific embodiment of this system will be described below.

[0991] System Overview

[0992] The system collects daily click-through rate (CTR) data, analyzes the characteristics of past banner ads, extracts the characteristics of high-CTR banners, and generates banners using a generative AI model.The generated banners are optimized based on landing page information, automatically expanded into multiple sizes, and dynamically generated and displayed in real time.

[0993] Hardware and software used

[0994] Server: Performs primary data processing and training of learning models. Software used is Python, TensorFlow (Keras), and PIL (Pillow).

[0995] Terminal: Operated by the user, it displays the generated banner and inputs the prompt text. A smartphone or tablet is used as the interface.

[0996] Database: Uses data storage to accumulate CTR data and banner ad feature data.

[0997] Data collection and analysis phase

[0998] The server collects historical and daily banner ad click-through rate (CTR) data from a database. This data includes many characteristic information related to layout, color, tagline, stock photos, motifs, gaze, fonts, etc. The server analyzes this data using statistical methods to identify design elements and patterns that indicate high CTR.

[0999] Feature extraction and training phase

[1000] The server builds a training dataset based on features extracted from high-CTR banners, creating a dataset with detailed information about each element. The training model is then trained using this data to build a model that predicts optimal design elements using machine learning algorithms such as random forests and neural networks.

[1001] Banner generation phase

[1002] The user provides landing page information, including key messaging and promotional content. The server uses the trained model to generate an optimal banner based on the provided landing page information. The generated banner includes layout, color, tagline, stock photos, motif, line of sight, font, etc.

[1003] Size development phase

[1004] The generated banner will be automatically expanded to multiple sizes. The device will process this and create banner images of different sizes. For example, various sizes such as 300x250, 728x90, 160x600 etc. will be generated.

[1005] Dynamic optimization and real-time generation

[1006] The system dynamically optimizes to respond to changes in the market and users in real time: the server instantly adjusts banner ads based on current click-through rate data and trends, ensuring they are displayed optimally to users.

[1007] Specific examples

[1008] A user sets up a landing page with promotional information such as "Summer Sale Begins!" and requests the system to generate a banner ad. The server collects and analyzes past data to extract the characteristics of high-CTR banners. After the model is trained, the server generates an optimal banner containing the tagline "Summer Sale Begins!" based on the landing page information provided by the user. The generated banner is automatically expanded to various sizes, such as 300x250, 728x90, and 160x600, and is dynamically optimized for display depending on the device.

[1009] Prompt Sentence Examples

[1010] "Generate a 'Summer Sale!' banner that has a high click-through rate based on historical data. Resize the generated banner to 300x250, 728x90, and 160x600."

[1011] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1012] Step 1: Data collection

[1013] The server collects daily click-through rate (CTR) data from a database, which includes characteristic information about past banner ads (layout, color, catchy copy, stock photos, motifs, gaze, font, etc.). Input data is obtained from the database, and CTR data is generated as output.

[1014] Step 2: Data analysis

[1015] The server analyzes the collected CTR data using statistical methods to identify banner design elements and patterns that result in high click-through rates. The input data is CTR data, and the output is characteristic data of high CTR banners. This identified data indicates the design elements of high CTR banners.

[1016] Step 3: Feature extraction

[1017] The server builds a training dataset based on the features of high-CTR banners, including detailed information on each element (layout, color, catchy copy, stock photos, motif, gaze, font, etc.). The input data is a dataset containing the features, and the output is a training dataset.

[1018] Step 4: Model training

[1019] The server trains a learning model using a training dataset. Specifically, it builds the model using machine learning algorithms such as random forests and neural networks. The input data is the training dataset, and the output is the trained learning model.

[1020] Step 5: Receive landing page information

[1021] The user provides landing page information that will be the basis for the banner ad they want to generate. The input data is the landing page information entered by the user, and the output is the appropriately formatted landing page information.

[1022] Step 6: Generate a banner

[1023] The server uses the learning model to generate an optimal banner based on the provided landing page information. The generated banner includes layout, color, catchy copy, stock photos, motif, gaze, font, etc. The input data is the landing page information and the learning model, and the output is the generated banner.

[1024] Step 7: Size development

[1025] The server automatically expands the generated banner to multiple sizes, for example, banner images of different sizes such as 300x250, 728x90, 160x600, etc. The input data is the generated banner, and the output is the banner images resized to different sizes.

[1026] Step 8: Dynamic optimization and real-time generation

[1027] The server adjusts banner ads in real time based on current click-through rate data and trends, and displays them optimally to users. The input data is the latest click-through rate data and the generated banner, and the output is a dynamically optimized banner.

[1028] Step 9: Displaying the banner

[1029] Users can view the optimized banner ads on their smartphones or other devices and request modifications or regeneration as necessary. The input data is the dynamically optimized banner, and the output is the displayed banner ad.

[1030] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1031] This invention relates to a system that combines click-through rate (CTR) data collection and analysis, banner generation using a learning model, optimization based on landing page information, and an emotion engine. The system aims to further maximize CTR by adjusting banner design elements based on user emotion data.

[1032] Data collection and analysis phase

[1033] collection

[1034] The server collects banner ad CTR data daily, retrieving information about past banner ad features such as layout, color, catchy copy, stock photos, motifs, gaze, and font, as well as their CTR, from databases and log files.

[1035] analysis

[1036] The server analyzes the collected data and extracts the characteristics of banners with high CTR. Specifically, it filters data for banner ads with CTRs above a certain threshold and calculates the average characteristics. In this process, it uses statistical methods to identify which design elements contribute to a high CTR.

[1037] Feature extraction and training phase

[1038] Feature Extraction

[1039] The server builds a training dataset based on the features extracted from high-CTR banners, including detailed information about each element (layout, color, catchy copy, stock photos, motif, gaze, and font).

[1040] study

[1041] This data is then used to train a learning model, using machine learning algorithms such as random forests and neural networks to build a model that predicts optimal design elements.

[1042] Banner generation phase

[1043] Receiving information

[1044] The user inputs landing page information, including key messaging and promotional content, and requests the creation of a banner ad.

[1045] Banner Generation

[1046] The server generates an optimal banner based on the landing page information and the characteristics of high-CTR banners obtained from the learning model, including layout, color, catchy copy, stock photos, motif, line of sight, font, etc.

[1047] Collaboration with emotion engine

[1048] Acquiring emotion data

[1049] The server obtains the user's emotional data from the emotion engine, including whether the user is in a positive, negative, or neutral emotional state.

[1050] Adjusting design elements

[1051] The server adjusts the banner's design elements based on the acquired emotional data, for example, using bright colors when the user is expressing positive emotions and muted colors when expressing negative emotions, and dynamically changes the tagline accordingly.

[1052] Size development phase

[1053] Size range

[1054] The generated banner will be automatically expanded to multiple sizes. The device will handle this process and create banner images of different sizes, for example 300x250, 728x90, 160x600, etc.

[1055] Specific examples

[1056] User operations

[1057] Assume that a user sets up a landing page with promotion information for "New Year Sale!" and requests the generation of a banner ad. Also, assume that the user's sentiment is positive.

[1058] Server Processing

[1059] 1. The server collects and analyzes past data and extracts the characteristics of high CTR banners.

[1060] 2. The learning model is trained based on the extracted features so that it can generate optimal banners.

[1061] 3. Based on the landing page information provided by the user, the server generates a suitable banner containing the tagline "New Year Sale!"

[1062] 4. The server retrieves the user's emotional data from the emotion engine and adjusts the design by using brighter colors if the emotion indicates a positive one.

[1063] 5. The device will automatically expand the generated banner to various sizes such as 300x250, 728x90, and 160x600.

[1064] The above is a detailed description of the embodiment of the present invention. This system reduces the workload of designers, efficiently generates optimal banners that maximize CTR, and also enables design adjustments based on user emotions.

[1065] The processing flow will be explained below.

[1066] Step 1:

[1067] The server collects daily banner ad click-through rate (CTR) data. It obtains past banner ad features such as layout, color, catchy copy, stock photos, motifs, gaze, and font, as well as their CTR, from databases and log files.

[1068] Step 2:

[1069] The server analyzes the collected data and extracts the characteristics of banners with high CTR. Specifically, it filters data for banner ads with CTRs above a certain threshold and calculates the average characteristics. In this process, it uses statistical methods to identify which design elements contribute to a high CTR.

[1070] Step 3:

[1071] The server builds a training dataset based on the features extracted from high-CTR banners, including detailed information about each element (layout, color, catchy copy, stock photos, motif, gaze, font, etc.).

[1072] Step 4:

[1073] The server uses the constructed training dataset to train a model. For example, it uses machine learning algorithms such as random forests and neural networks to build a model for generating optimal banners. This model plays a role in predicting high CTR.

[1074] Step 5:

[1075] The user inputs landing page information, including key messaging and promotional content, and requests the creation of a banner ad.

[1076] Step 6:

[1077] The server receives the provided landing page information and generates the optimal banner based on it. The landing page information and the extracted characteristics of the high CTR banner are input into the learning model to determine the optimal design elements. The generated banner includes layout, color, catchy copy, stock photos, motif, line of sight, font, etc.

[1078] Step 7:

[1079] The server obtains the user's emotional data from the emotion engine, which includes the user's emotional state, such as positive, negative, or neutral.

[1080] Step 8:

[1081] The server adjusts the banner's design elements based on the acquired emotional data, for example, using bright colors when the user is expressing positive emotions and muted colors when expressing negative emotions. Furthermore, the tagline dynamically changes depending on the emotion.

[1082] Step 9:

[1083] The device receives the generated banner and automatically expands it to multiple sizes, for example, resizing it to different sizes such as 300x250, 728x90, and 160x600, and then saving it. This allows a banner that can be used in multiple ad spaces to be prepared at once.

[1084] Step 10:

[1085] The generated banners in multiple sizes are provided to the user, who can then use them in actual advertising operations to check the effectiveness of CTR.

[1086] As a concrete example, suppose a user sets up a landing page with promotional information for "New Year Sale!" and requests the generation of a banner ad. Let's also assume that the user's sentiment is positive. The server uses a learning model and sentiment engine to generate an optimal banner that combines the "New Year Sale!" tagline with bright colors. The device then expands this banner into multiple sizes and provides it to the user. This series of processes allows users to efficiently use banner ads that aim for a high CTR.

[1087] Example 2

[1088] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1089] Conventional audiovisual advertisement generation systems do not take into account the user's emotional response and therefore have difficulty generating optimal advertisements. Furthermore, the process of automatically expanding the generated advertisements into multiple sizes is also time-consuming. Therefore, a new system for creating more effective audiovisual advertisements and automatically expanding them is needed.

[1090] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1091] In this invention, the server includes means for collecting daily selection rate data, means for analyzing characteristics of past audiovisual advertisements, means for extracting characteristics of audiovisual advertisements with high selection rates, means for generating audiovisual advertisements using a learning algorithm, means for optimizing the generated audiovisual advertisements based on guide page information, means for adjusting design elements based on user emotions using an emotion processing engine, and means for automatically expanding the generated audiovisual advertisements into multiple dimensions, thereby enabling efficient generation of optimal audiovisual advertisements according to user emotions and automatic expansion thereof.

[1092] "Selection rate data" refers to data that indicates the rate at which users click or select an audiovisual advertisement displayed.

[1093] "Audiovisual advertising" refers to advertising that uses both visual and auditory sensations to provide information to users.

[1094] A "learning algorithm" is a machine learning method for predicting optimal design elements for audiovisual advertisements based on collected data.

[1095] "Guide page information" refers to the content of a web page that contains promotional information for a particular product or service.

[1096] An "emotion processing engine" is a system that analyzes a user's emotional state and adjusts design elements based on the results.

[1097] "Design elements" are the components of an audiovisual advertisement, and specifically include layout, color, wording, image materials, theme, line of sight, and font.

[1098] "Dimensions" refers to the physical size and proportions of an audiovisual advertisement.

[1099] "User" means an individual or organization that utilizes the system to generate and display audiovisual advertisements.

[1100] This invention relates to a system that collects and analyzes audiovisual advertisement selection rate data, combines a learning algorithm and an emotion processing engine to generate optimal audiovisual advertisements, and optimizes and automatically deploys the advertisements based on guide page information.

[1101] The hardware required to implement this system includes a database server, a web server, an emotion processing engine server, and a terminal. The software used includes a database management system (e.g., MySQL), a data processing library (e.g., Pandas, NumPy), a machine learning library (e.g., Scikit-learn, TensorFlow), and an image processing library (e.g., PIL).

[1102] The specific program processing is as follows.

[1103] The server collects audiovisual ad selection rate data daily from a database and log files, including design elements such as layout, color, wording, image materials, theme, eye direction, and typeface. The server uses SQL queries to extract relevant data from the database and Python scripts to retrieve additional information from the log files.

[1104] The server then analyzes the collected data using statistical methods. It extracts design elements of audiovisual advertisements with high selection rates, filters advertisements that exceed a certain threshold, and calculates average features. Specifically, the server creates a data frame using the Pandas library and calculates statistical information such as the mean and standard deviation using the NumPy library.

[1105] The server builds a training dataset based on features extracted from high-selection audiovisual ads. This dataset includes detailed information about each design element. The server saves the extracted features in a CSV file, which is then fed into machine learning algorithms. The training algorithms include random forests and neural networks, which are trained using Scikit-learn and TensorFlow.

[1106] The user inputs information about the page through the system interface and requests the generation of an audiovisual advertisement. The input information includes the promotional content and key messages. The user submits this information using a web form.

[1107] The server combines the information on the user's guide page with the characteristics of high-selection audiovisual ads estimated from the learning model to generate optimal audiovisual ads. Specifically, the server uses PIL (Python Imaging Library) to combine elements such as layout and catchy copy to generate ad images. A generative AI model is also used in this process.

[1108] The server obtains user emotional data in real time from an emotion processing engine. This data includes the user's emotional state, such as positive, negative, or neutral. The server adjusts design elements based on this emotional data. For example, bright colors are used to indicate positive emotions, and muted colors to indicate negative emotions. The tagline also changes dynamically depending on the emotion.

[1109] The generated audiovisual advertisement is automatically deployed to multiple dimensions by the device, which uses an image processing library to resize and reposition the advertisement image to the appropriate ratio.

[1110] As a concrete example, consider the case where a user sets up an information page with promotional information for "New Year Sale!" and requests the generation of a banner ad. Assume the user's sentiment is positive. In this case, the server collects and analyzes past data to extract features of audiovisual ads with high selection rates. Based on the extracted features, the server trains a model to generate an optimal audiovisual ad containing the slogan "New Year Sale!". Furthermore, the server obtains positive sentiment data from the sentiment engine and adjusts the design to a brighter color tone. The generated ad is automatically displayed in multiple sizes by the device.

[1111] Example prompt sentence:

[1112] "Generate optimal audiovisual ads taking into account sentiment data. The promotion is 'Summer Sale' and users are showing positive sentiment."

[1113] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1114] Step 1:

[1115] Collecting CTR data

[1116] The server collects audiovisual ad selection rate data from databases and log files daily. The databases and log files are used as input, and the extracted data includes design elements such as layout, color, wording, image materials, theme, eye line, and typeface. This data is collected using SQL queries and Python scripts and saved as a data frame.

[1117] Specifically, the server performs the following steps:

[1118] Use a SQL query to extract daily selectivity data from the database.

[1119] Run a Python script and retrieve additional information from the log file.

[1120] The collected data is converted into a data frame using the Pandas library and saved.

[1121] Step 2:

[1122] Analyzing CTR data

[1123] The server analyzes the collected data using statistical methods. The input is the data frame collected in step 1, and the output is to extract the characteristics of audiovisual advertisements with high selection rates. Specifically, it calculates the average and standard deviation, and filters out advertisements that exceed a certain threshold.

[1124] Specifically, the server performs the following steps:

[1125] Use the Pandas library to filter high-selectivity design elements from a data frame.

[1126] Using the NumPy library, calculate the mean and standard deviation of each design element.

[1127] The characteristics of high selection rate audiovisual advertisements are extracted and saved in a list.

[1128] Step 3:

[1129] Building a feature dataset

[1130] The server constructs a training dataset based on the extracted features of high-selectivity audiovisual advertisements. The input is the feature list extracted in step 2, and the output is a training dataset (CSV file).

[1131] Specifically, the server performs the following steps:

[1132] The extracted features are organized to build a dataset containing detailed information about each design element.

[1133] Convert the dataset to a CSV file and save it.

[1134] Step 4:

[1135] Training the learning model

[1136] The server trains a learning algorithm using the constructed training dataset. The input is the CSV file constructed in step 3, and the output is the trained model. The learning algorithms used include Scikit-learn's random forest and TensorFlow's neural network.

[1137] Specifically, the server performs the following steps:

[1138] Read the CSV file and generate a training dataset.

[1139] Train random forest and neural network models using Scikit-learn and TensorFlow libraries.

[1140] Save the trained model.

[1141] Step 5:

[1142] Receiving landing page information

[1143] The user inputs the information about the user's guide page through the system interface and requests the generation of an audiovisual advertisement. The input is the user's guide page information, and the output is that this information is sent to the server.

[1144] Specifically, the user performs the following steps:

[1145] Use the web form to enter your information.

[1146] The information is sent to the server.

[1147] Step 6:

[1148] Banner generation

[1149] The server generates an optimal audiovisual advertisement by integrating the information on the guide page received from the user and the features of the high-selection rate audiovisual advertisement estimated from the learning model. The input is the landing page information and the trained model, and the output is the generated audiovisual advertisement (image file).

[1150] Specifically, the server performs the following steps:

[1151] Analyze landing page information and extract key messages and promotional content.

[1152] Use a trained model to select the best design elements.

[1153] Audiovisual advertising images are generated using PIL (Python Imaging Library).

[1154] Step 7:

[1155] Acquiring emotion data

[1156] The server receives user emotion data from the emotion processing engine. The input is the user emotion data (via API), and the output is data used to adjust design elements according to the emotion.

[1157] Specifically, the server performs the following steps:

[1158] A request is sent to the emotion processing engine via the API to obtain emotion data.

[1159] Analyze the acquired emotion data.

[1160] Step 8:

[1161] Adjusting design elements

[1162] The server adjusts the design elements of the banner based on the acquired emotional data. The input is the emotional data and the generated audiovisual advertisement, and the output is the adjusted final audiovisual advertisement.

[1163] Specifically, the server performs the following steps:

[1164] Colors and catchy slogans are dynamically changed based on emotional data.

[1165] The audiovisual advertising images are re-edited using PIL.

[1166] Step 9:

[1167] Banner size options

[1168] The terminal automatically unwraps the generated audiovisual advertisement into multiple dimensions: the input is the adjusted final audiovisual advertisement, and the output is the audiovisual advertisement (multiple image files) in different sizes.

[1169] Specifically, the device performs the following steps:

[1170] Load the generated audiovisual advertisement and resize it to the different sizes you need.

[1171] Generate audiovisual ads optimized for each size, taking into account existing placement and proportions.

[1172] (Application example 2)

[1173] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1174] Conventional banner ad generation systems focus on optimizing click-through rates (CTR), but do not take into account the user's emotional state when generating banners. Therefore, further optimization is required to maximize CTR. Furthermore, it is necessary to support dynamic design adjustments based on user emotional data when displaying ads on different devices, such as smartphones and head-mounted displays.

[1175] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1176] In this invention, the server includes a means for collecting daily click-through rate (CTR) data, a means for analyzing the characteristics of past banner advertisements, a means for extracting the characteristics of high-CTR banners, a means for generating banners using a learning model, a means for optimizing the generated banners based on landing page information, a means for automatically expanding the generated banners to multiple sizes, a means for collecting user emotion data in real time, a means for adjusting the banner design elements based on the collected emotion data, and a means for displaying banners optimized for smartphones and head-mounted displays. This allows for the generation of optimal banners that take the user's emotional state into consideration, further maximizing CTR. Furthermore, dynamic design adjustments based on user emotion data can be realized even when displaying advertisements on different devices.

[1177] "Click-through rate (CTR)" is an indicator that indicates the ratio of the number of times an ad is clicked by a user to the number of times it is displayed.

[1178] A "banner ad" is an image or graphic advertisement that appears on a web page or within an application.

[1179] A "learning model" is a model that uses algorithms to learn patterns and characteristics based on past data and makes predictions and classifications for new data.

[1180] "Landing page information" refers to the content and information of the web page that a user who clicks on an advertisement first arrives at.

[1181] "Emotion data" is data that indicates the user's emotional state, such as positive, negative, or neutral, obtained from the user's facial expression and behavior.

[1182] "Design elements" refers to the components of a banner ad, such as layout, color, catchy copy, stock photos, motifs, line of sight, and font.

[1183] A "smartphone" is a mobile information terminal that, in addition to the functions of a mobile phone, can also browse the Internet and use applications.

[1184] A "head-mounted display" is a display device worn on the head, allowing users to experience virtual reality (VR) and augmented reality (AR).

[1185] "Real-time" means that data collection and processing occur simultaneously with actual time.

[1186] A "high CTR banner" is a banner ad that has a higher-than-average click-through rate (CTR).

[1187] An "emotion engine" is software or a system that analyzes a user's emotional state and provides that data.

[1188] The system of the present invention aims to dynamically generate advertising banners based on user emotional data and maximize click-through rates (CTR). The system uses advanced machine learning models to analyze past data and apply real-time emotional information to generate optimal banners that will attract users' attention.

[1189] 1. Generate a program

[1190] The system consists of the following main components:

[1191] 1. Data Collection Module

[1192] 2. Data Analysis Module

[1193] 3. Learning Model Training Module

[1194] 4. Banner generation module

[1195] 5. Emotion Data Acquisition Module

[1196] 6. Banner Adjustment Module

[1197] 7. Display module

[1198] 2. Explain the system's processing in natural language

[1199] Data Collection Module

[1200] The server collects daily click-through rate (CTR) data, which includes the number of clicks and impressions for each banner ad.

[1201] Data Analysis Module

[1202] The server analyzes the characteristics of past banner ads to determine which features have a high CTR. This involves analyzing design elements such as banner layout, color, tagline, stock photos, motifs, line of sight, and font.

[1203] Learning Model Training Module

[1204] The server trains a learning model based on the analysis results, using TensorFlow to build a neural network to generate the optimal banner.

[1205] Banner Generation Module

[1206] Using the learning model, the server generates the optimal banner based on the landing page information, automatically selecting taglines and images that correspond to the promotional information provided by the user.

[1207] Emotion data acquisition module

[1208] The user's device (smartphone or head-mounted display) uses a camera to collect the user's emotional data in real time. OpenCV and FaceAPI are used to analyze the user's emotional state (positive, negative, neutral, etc.) from their facial expressions.

[1209] Banner Adjustment Module

[1210] The server dynamically adjusts the banner's design elements based on the acquired emotional data: for example, if the user is expressing positive emotions, the banner's colors will be brighter and the copy's tone will be more optimistic.

[1211] Display Module

[1212] The generated banners are displayed on the user's device in the optimal size and format, which includes a process to automatically generate banner images of different sizes.

[1213] 3. Specific examples and prompt sentences

[1214] Specific examples

[1215] When a landing page with promotion information "New Year Sale!" is input, the server generates an optimal banner with the tagline "50% off on all items." If the user's sentiment data is analyzed as positive, the banner design will be composed of bright colors. The displayed banner will also be optimized for the screen size of smartphones and head-mounted displays.

[1216] Prompt Sentence Examples

[1217] Input data:

[1218] Landing page information: {'message': 'New Year Sale!', 'promotion_details': '50% off on all items'}

[1219] Emotion data: {'emotion': 'positive'}

[1220] Requirements:

[1221] Always include the tagline "New Year Sale!"

[1222] Use bright colors.

[1223] Use high CTR elements when generating images.

[1224] Generation goal:

[1225] To generate an optimal advertisement banner based on a user's positive emotional state.

[1226] This system can generate advertising banners that reflect the user's emotional state, maximizing click-through rates (CTR). It can also dynamically adjust the design of ads displayed on different devices based on the user's emotional data.

[1227] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1228] Step 1:

[1229] The server collects daily click-through rate (CTR) data, which includes the number of clicks and impressions for each banner ad. The data is retrieved from databases and log files and stored in a database.

[1230] Input: Number of clicks and impressions of past banner ads

[1231] Data processing: Data collection and organization

[1232] Output: Organized CTR data

[1233] Step 2:

[1234] The server analyzes the characteristics of past banner ads to identify banner design elements (layout, color, tagline, stock photo, motif, gaze, font) that cause CTRs above a certain threshold.

[1235] Input: CTR data and banner ad characteristics information

[1236] Data calculations: statistical analysis

[1237] Output: High CTR banner feature data

[1238] Step 3:

[1239] The server extracts features of high-CTR banners and builds a learning dataset. TensorFlow is used to train the learning model. A neural network is used to build a model to predict the optimal banner design.

[1240] Input: High CTR banner feature data

[1241] Data Computing: Training Machine Learning Models

[1242] Output: A trained learning model

[1243] Step 4:

[1244] The user's device (smartphone or head-mounted display) uses a camera to collect the user's emotional data in real time. OpenCV and FaceAPI are used to analyze the user's emotional state (positive, negative, neutral, etc.) from their facial expressions.

[1245] Input: Camera image

[1246] Data Computing: Sentiment Analysis

[1247] Output: Real-time emotion data

[1248] Step 5:

[1249] Once the user enters the landing page information, the server uses this information to generate the optimal banner. The generative AI model inputs prompt text and creates a banner with the desired tagline and image.

[1250] Input: Landing page information

[1251] Data calculation: Banner generation

[1252] Output: The generated banner

[1253] Prompt Sentence Examples

[1254] Input data:

[1255] Landing page information: {'message': 'New Year Sale!', 'promotion_details': '50% off on all items'}

[1256] Emotion data: {'emotion': 'positive'}

[1257] Requirements:

[1258] Always include the tagline "New Year Sale!"

[1259] Use bright colors.

[1260] Use high CTR elements when generating images.

[1261] Generation goal:

[1262] To generate an optimal advertisement banner based on a user's positive emotional state.

[1263] Step 6:

[1264] The server dynamically adjusts the banner's design elements based on the acquired emotional data: for example, if the user is expressing positive emotions, the banner's color may be brightened and the tagline may be changed to a more positive tone.

[1265] Input: Generated banner and sentiment data

[1266] Data calculation: Dynamic banner adjustment

[1267] Output: Adjusted banner

[1268] Step 7:

[1269] The generated banner is displayed on the user's device in the optimal size and format. The server automatically generates banner images of different sizes and displays them in a format suitable for smartphones and head-mounted displays.

[1270] Input: Adjusted banner

[1271] Data processing: Size expansion

[1272] Output: Banner optimized for display

[1273] Through these steps, the system can generate optimal banners that take into account the user's emotional state and maximize click-through rates (CTR).

[1274] The specific processing unit 290 transmits the result of the specific processing to the headset type terminal 314. In the headset type terminal 314, the control unit 46A causes the speaker 240 and the display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[1275] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1276] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the headset type terminal 314.

[1277] [Fourth embodiment]

[1278] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1279] 7, a data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[1280] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1281] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a control target 443. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the control target 443 are also connected to the bus 52.

[1282] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[1283] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[1284] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1285] The control object 443 includes a display device, LEDs in the eyes, and motors for driving the arms, hands, and feet. The posture and gestures of the robot 414 are controlled by controlling the motors of the arms, hands, and feet. Some of the emotions of the robot 414 can be expressed by controlling these motors. In addition, the facial expressions of the robot 414 can also be expressed by controlling the light emission state of the LEDs in the eyes of the robot 414.

[1286] Fig. 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Fig. 8, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1287] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[1288] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1289] In the robot 414, the processor 46 performs the reception output process. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1290] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1291] The present invention relates to a system that collects daily click-through rate (CTR) data, analyzes the characteristics of past banner advertisements, extracts the characteristics of high-CTR banners, generates banners using a learning model, optimizes the generated banners based on landing page information, and automatically expands the generated banners into multiple sizes.

[1292] As an embodiment of the present invention, a specific program and its processing content will be described below.

[1293] Data collection and analysis phase

[1294] collection

[1295] The server collects CTR data for banner ads from a database every day. This data includes many characteristic information about past banner ads (layout, color, catchy copy, photo material, motif, gaze, font, etc.).

[1296] analysis

[1297] The server analyzes the collected data and extracts the characteristics of banners with high CTRs, using statistical methods to analyze the data and identify design elements and patterns that indicate high CTRs.

[1298] Feature extraction and training phase

[1299] Feature Extraction

[1300] The server builds a training dataset based on features extracted from high-CTR banners, including detailed information about each element (such as which layout is most effective or which color is most eye-catching).

[1301] study

[1302] This data is then used to train a learning model, using machine learning algorithms such as random forests and neural networks to build a model that predicts optimal design elements.

[1303] Banner generation phase

[1304] Receiving information

[1305] The user provides landing page information, including the landing page's key message and promotion.

[1306] Banner Generation

[1307] The server uses the learned model to generate an optimal banner based on the provided landing page information, including layout, color, tagline, stock photos, motif, line of sight, font, etc.

[1308] Size development phase

[1309] Size range

[1310] The generated banner will be automatically expanded to multiple sizes. The device will handle this process and create banner images of different sizes, for example 300x250, 728x90, 160x600, etc.

[1311] Specific examples

[1312] User operations

[1313] A user sets up a landing page with promotional information for "New Year Sale!" and requests to generate a banner ad.

[1314] Server Processing

[1315] 1. The server collects and analyzes past data and extracts the characteristics of high CTR banners.

[1316] 2. The learning model is trained based on the extracted features so that it can generate optimal banners.

[1317] 3. Based on the landing page information provided by the user, the server generates a suitable banner containing the tagline "New Year Sale!"

[1318] 4. The device will automatically expand the generated banner to various sizes such as 300x250, 728x90, and 160x600.

[1319] The above is a detailed description of the embodiment of the present invention. This system reduces the workload of designers and automatically generates optimal banners that maximize CTR.

[1320] The processing flow will be explained below.

[1321] Step 1:

[1322] The server collects daily banner ad click-through rate (CTR) data. It obtains past banner ad features such as layout, color, catchy copy, stock photos, motifs, gaze, and font, as well as their CTR, from databases and log files.

[1323] Step 2:

[1324] The server analyzes the collected data and extracts the characteristics of banners with high CTR. Specifically, it filters data for banner ads with CTRs above a certain threshold and calculates the average characteristics. During this process, it uses statistical methods to identify which design elements contribute to a high CTR.

[1325] Step 3:

[1326] The server builds a training dataset based on the extracted features. It organizes the design elements of the filtered high-CTR banner ads into a feature vector and formats it to be fed into the training model.

[1327] Step 4:

[1328] The server uses the training dataset to train a model. For example, it uses machine learning algorithms such as random forests and neural networks to build a model for generating optimal banners. This model is responsible for predicting high CTR.

[1329] Step 5:

[1330] The user inputs landing page information, including key messaging and promotional content, and requests the creation of a banner ad.

[1331] Step 6:

[1332] The server receives the provided landing page information and generates the optimal banner based on it. The landing page information and the extracted characteristics of the high CTR banner are input into the learning model to determine the optimal design elements. The generated banner includes layout, color, catchy copy, stock photos, motif, line of sight, font, etc.

[1333] Step 7:

[1334] The device receives the generated banner and automatically expands it to multiple sizes, for example, resizing it to different sizes such as 300x250, 728x90, and 160x600, and then saving it. This allows you to prepare a banner that can be used in multiple ad spaces at once.

[1335] Step 8:

[1336] The generated banners in multiple sizes are provided to the user, who can then use them in actual advertising operations to check the effectiveness of CTR.

[1337] The above is the specific processing flow of the system that automatically generates banner ads to maximize CTR and deploys them in various sizes. This series of processes makes it possible to efficiently generate optimal banners that achieve a high CTR while reducing the workload on designers.

[1338] Example 1

[1339] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1340] Conventional banner ad generation systems lack the functionality to automatically analyze and apply design elements that are expected to improve click-through rates (CTR), and require manual design and adjustment. This places a heavy workload on designers, making it difficult to create effective banner ads. The objective of this invention is to provide a system that optimizes design elements and automatically expands to multiple sizes, efficiently generating effective banner ads.

[1341] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1342] In this invention, the server includes means for collecting daily click-through rate (CTR) data, means for analyzing characteristics of past banner ads, means for extracting characteristics of high CTR banners, means for generating banners using a learning model, means for optimizing the generated banners based on landing page information, means for automatically expanding the generated banners into multiple sizes, and means for saving the optimized banners in cloud storage. This makes it possible to automatically and efficiently generate and save optimal banner ads that are expected to have a high CTR.

[1343] "Click-through rate (CTR)" indicates the percentage of clicks on a displayed ad, and is an indicator used to measure the effectiveness of an ad.

[1344] A "banner ad" is advertising content in the form of an image or video displayed on a web page, and is created with the purpose of encouraging users to click on it.

[1345] "Features" refers to identifiable characteristics of a banner ad, such as specific design elements, layout, color, catchy copy, photographic material, motif, line of sight, font, etc.

[1346] A "high CTR banner" refers to a banner ad that has a relatively high click-through rate and contains an effective design and message.

[1347] A "learning model" is a computational model that uses machine learning algorithms to learn patterns and features from data and make predictions and classifications based on future data.

[1348] A "landing page" is a specific web page accessed from a web advertisement or search engine result, and is the first page a user visits.

[1349] "Optimization" refers to the process of adjusting conditions or factors to most effectively achieve a particular objective.

[1350] "Cloud storage" refers to services that allow data to be stored remotely and made accessible over the internet.

[1351] "Sizing" refers to the process of automatically adapting a particular design to different dimensions to accommodate multiple ad slots.

[1352] This invention provides a system for optimizing the click-through rate (CTR) of banner ads. This system collects CTR data, analyzes and extracts the characteristics of banner ads that show high CTR, generates optimal banners using a learning model, optimizes the generated banners based on landing page information, and deploys them in multiple sizes.

[1353] Data collection and analysis

[1354] The server first connects to a database to collect daily banner ad CTR data. This database stores information about past banner ad CTRs, layouts, colors, taglines, stock photos, motifs, gaze patterns, fonts, etc. The server extracts this information using SQL queries and performs data cleaning, such as removing missing and outlier values ​​and normalizing the data.

[1355] The server then analyzes the collected data using statistical methods, such as calculating average CTR, variance, and correlation coefficients, to identify design elements and patterns that indicate high CTR.

[1356] Feature extraction and model training

[1357] The server extracts the features of high-CTR banners and builds a training dataset based on a multidimensional feature vector. This dataset contains the features of each banner and the corresponding CTR.

[1358] Next, the server selects a machine learning algorithm (e.g., random forest, neural network, etc.) and trains the model using the training dataset. This training model is then used to generate banners with optimal design elements to predict high CTR. Specifically, the random forest model can be trained using Scikit-learn, an educational Python library.

[1359] Banner generation and size development

[1360] The user enters the landing page information, which includes the main message and promotional content. This information is sent to the server via the terminal. For example, the promotional information entered may be "New Year Sale!"

[1361] The server takes this input information and uses a trained model to generate an optimal banner, including layout, color, tagline, stock image, motif, line of sight, font, etc. The actual banner image is generated using the Pillow library and other image processing tools.

[1362] The device then converts the generated banner to different sizes, using OpenCV or other image conversion libraries to automatically convert it to different ad sizes such as 300x250, 728x90, and 160x600.

[1363] Final generation and saving

[1364] The user can check the generated banner and request corrections if necessary. The server then saves the optimized banner to cloud storage, often using Amazon S3 for example. The saved banner is then provided to the user as a URL.

[1365] For example, the prompt is:

[1366] Generate the optimal banner based on the promotion information "New Year Sale!"

[1367] Using this system will significantly reduce the workload of designers and enable the automatic generation, deployment, and storage of banner ads that are expected to have a high CTR.

[1368] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1369] Step 1:

[1370] Data collection

[1371] The server collects daily click-through rate (CTR) data from a database. Specifically, it connects to the database and executes SQL queries to extract information related to past banner ads, such as CTR, layout, color, tagline, stock image, motif, gaze, and font. It specifies a date range and ad ID as input, and obtains a dataset containing CTR data as output. For example, it executes the query "SELECT FROM Banners WHERE date >= CURDATE() - INTERVAL 30 DAY;".

[1372] Step 2:

[1373] Data analysis

[1374] The server analyzes the collected data. It uses the CTR data collected in step 1 as input, performs data cleaning, and removes missing and outlier values. It then analyzes the data using statistical methods to identify design elements and patterns that indicate high CTR. Specifically, it calculates the average CTR, variance, correlation coefficient, etc., and obtains a list of elements that indicate high CTR as output.

[1375] Step 3:

[1376] Feature Extraction

[1377] The server extracts features from the high CTR banners identified in step 2. It uses the analysis result dataset as input and builds a feature vector for each banner. For example, it creates a matrix-format dataset containing detailed information such as which layout is most effective or which color is most eye-catching. As output, it obtains a training dataset.

[1378] Step 4:

[1379] Learning Phase

[1380] The server trains the machine learning model. It uses the training dataset constructed in step 3 as input and selects a machine learning algorithm such as a random forest or neural network. Specifically, it uses the Scikit-learn library to run the model's fit method and train the optimal predictive model. The output is the trained learning model.

[1381] Step 5:

[1382] Receiving information

[1383] The user inputs landing page information, specifically key messages and promotional content, through an input form, and the device sends this information to the server. The input is the landing page information, and the output is the data sent to the server.

[1384] Step 6:

[1385] Banner Generation

[1386] The server uses the trained model to generate the optimal banner based on the provided landing page information. Using the landing page information and the trained model as input, it calculates the banner design parameters that predict the highest CTR. Specifically, it determines the tagline, layout, color, stock photos, etc., and generates the actual banner image using the Pillow library. The generated banner image is obtained as output.

[1387] Step 7:

[1388] Size range

[1389] The device will unpack the generated banner into multiple sizes. As input, it uses the banner image generated in step 6 and automatically converts it to different sizes (e.g., 300x250, 728x90, 160x600) using an image conversion library such as OpenCV. As output, it gets the banner image for each size.

[1390] Step 8:

[1391] Final generation and saving

[1392] The server saves the generated banner in cloud storage. As input, it uses banner images of multiple sizes and uploads them to a cloud storage service. Specifically, it saves the banner images using a storage service such as Amazon S3, generates a URL for the saved banner, and provides it to the user. As output, it obtains the URL for the saved banner.

[1393] This allows the system to automatically generate optimal banner ads aiming for high CTR, and efficiently expand and save them in multiple sizes.

[1394] (Application example 1)

[1395] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1396] Conventional banner ad generation systems simply generate and display static banners based on past data. This makes it difficult to provide effective ads that maximize CTR, as it is not possible to optimize in real time or respond to dynamically changing market and user preferences. Furthermore, manual banner size deployment and optimization requires a great deal of effort, hindering efficient advertising operations.

[1397] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1398] In this invention, the server includes means for collecting daily click-through rate (CTR) data, means for analyzing the characteristics of past banner advertisements, means for extracting the characteristics of high-CTR banners, means for generating banners using a learning model, means for optimizing the generated banners based on landing page information, means for automatically expanding the generated banners to multiple sizes, and means for generating and displaying dynamically optimized banner advertisements in real time. This makes it possible to respond to changes in the market and users in real time and generate dynamically optimized banner advertisements. Furthermore, automatic expansion of banner sizes enables efficient advertising management.

[1399] "Click-through rate (CTR)" is an indicator that shows the ratio of the number of clicks to the number of times an ad is displayed.

[1400] A "banner ad" is a graphic advertisement that appears on a web page or within an application.

[1401] "Landing page information" refers to information about the content and structure of the web page to which a user is directed when they click on an advertisement.

[1402] A "learning model" is a machine learning algorithm that learns patterns and rules based on data and makes predictions and classifications.

[1403] "Feature extraction" is the process of extracting useful characteristics or patterns from data.

[1404] "Real-time" refers to the time characteristics in which data is processed and results are provided immediately.

[1405] "Optimization" is the process of adjusting parameters and conditions to achieve the best results in order to achieve a specific objective.

[1406] "Sizing" is the process of adjusting the original design to different sizes to generate multiple sizes of banners.

[1407] "Dynamic optimization" is the process of instantly adjusting the optimum state in response to changing conditions in real time.

[1408] A "prompt" refers to a sentence of instructions or questions that is input into a generative AI model.

[1409] MODE FOR CARRYING OUT THE INVENTION

[1410] To implement the present invention, a system in which a server, a terminal, and a user work in cooperation with each other is required. A specific embodiment of this system will be described below.

[1411] System Overview

[1412] The system collects daily click-through rate (CTR) data, analyzes the characteristics of past banner ads, extracts the characteristics of high-CTR banners, and generates banners using a generative AI model.The generated banners are optimized based on landing page information, automatically expanded into multiple sizes, and dynamically generated and displayed in real time.

[1413] Hardware and software used

[1414] Server: Performs primary data processing and training of learning models. Software used is Python, TensorFlow (Keras), and PIL (Pillow).

[1415] Terminal: Operated by the user, it displays the generated banner and inputs the prompt text. A smartphone or tablet is used as the interface.

[1416] Database: Uses data storage to accumulate CTR data and banner ad feature data.

[1417] Data collection and analysis phase

[1418] The server collects historical and daily banner ad click-through rate (CTR) data from a database. This data includes many characteristic information related to layout, color, tagline, stock photos, motifs, gaze, fonts, etc. The server analyzes this data using statistical methods to identify design elements and patterns that indicate high CTR.

[1419] Feature extraction and training phase

[1420] The server builds a training dataset based on features extracted from high-CTR banners, creating a dataset with detailed information about each element. The training model is then trained using this data to build a model that predicts optimal design elements using machine learning algorithms such as random forests and neural networks.

[1421] Banner generation phase

[1422] The user provides landing page information, including key messaging and promotional content. The server uses the trained model to generate an optimal banner based on the provided landing page information. The generated banner includes layout, color, tagline, stock photos, motif, line of sight, font, etc.

[1423] Size development phase

[1424] The generated banner will be automatically expanded to multiple sizes. The device will process this and create banner images of different sizes. For example, various sizes such as 300x250, 728x90, 160x600 etc. will be generated.

[1425] Dynamic optimization and real-time generation

[1426] The system dynamically optimizes to respond to changes in the market and users in real time: the server instantly adjusts banner ads based on current click-through rate data and trends, ensuring they are displayed optimally to users.

[1427] Specific examples

[1428] A user sets up a landing page with promotional information such as "Summer Sale Begins!" and requests the system to generate a banner ad. The server collects and analyzes past data to extract the characteristics of high-CTR banners. After the model is trained, the server generates an optimal banner containing the tagline "Summer Sale Begins!" based on the landing page information provided by the user. The generated banner is automatically expanded to various sizes, such as 300x250, 728x90, and 160x600, and is dynamically optimized for display depending on the device.

[1429] Prompt Sentence Examples

[1430] "Generate a 'Summer Sale!' banner that has a high click-through rate based on historical data. Resize the generated banner to 300x250, 728x90, and 160x600."

[1431] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1432] Step 1: Data collection

[1433] The server collects daily click-through rate (CTR) data from a database, which includes characteristic information about past banner ads (layout, color, catchy copy, stock photos, motifs, gaze, font, etc.). Input data is obtained from the database, and CTR data is generated as output.

[1434] Step 2: Data analysis

[1435] The server analyzes the collected CTR data using statistical methods to identify banner design elements and patterns that result in high click-through rates. The input data is CTR data, and the output is characteristic data of high CTR banners. This identified data indicates the design elements of high CTR banners.

[1436] Step 3: Feature extraction

[1437] The server builds a training dataset based on the features of high-CTR banners, including detailed information on each element (layout, color, catchy copy, stock photos, motif, gaze, font, etc.). The input data is a dataset containing the features, and the output is a training dataset.

[1438] Step 4: Model training

[1439] The server trains a learning model using a training dataset. Specifically, it builds the model using machine learning algorithms such as random forests and neural networks. The input data is the training dataset, and the output is the trained learning model.

[1440] Step 5: Receive landing page information

[1441] The user provides landing page information that will be the basis for the banner ad they want to generate. The input data is the landing page information entered by the user, and the output is the appropriately formatted landing page information.

[1442] Step 6: Generate a banner

[1443] The server uses the learning model to generate an optimal banner based on the provided landing page information. The generated banner includes layout, color, catchy copy, stock photos, motif, gaze, font, etc. The input data is the landing page information and the learning model, and the output is the generated banner.

[1444] Step 7: Size development

[1445] The server automatically expands the generated banner to multiple sizes, for example, banner images of different sizes such as 300x250, 728x90, 160x600, etc. The input data is the generated banner, and the output is the banner images resized to different sizes.

[1446] Step 8: Dynamic optimization and real-time generation

[1447] The server adjusts banner ads in real time based on current click-through rate data and trends, and displays them optimally to users. The input data is the latest click-through rate data and the generated banner, and the output is a dynamically optimized banner.

[1448] Step 9: Displaying the banner

[1449] Users can view the optimized banner ads on their smartphones or other devices and request modifications or regeneration as necessary. The input data is the dynamically optimized banner, and the output is the displayed banner ad.

[1450] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1451] This invention relates to a system that combines click-through rate (CTR) data collection and analysis, banner generation using a learning model, optimization based on landing page information, and an emotion engine. The system aims to further maximize CTR by adjusting banner design elements based on user emotion data.

[1452] Data collection and analysis phase

[1453] collection

[1454] The server collects banner ad CTR data daily, retrieving information about past banner ad features such as layout, color, catchy copy, stock photos, motifs, gaze, and font, as well as their CTR, from databases and log files.

[1455] analysis

[1456] The server analyzes the collected data and extracts the characteristics of banners with high CTR. Specifically, it filters data for banner ads with CTRs above a certain threshold and calculates the average characteristics. In this process, it uses statistical methods to identify which design elements contribute to a high CTR.

[1457] Feature extraction and training phase

[1458] Feature Extraction

[1459] The server builds a training dataset based on the features extracted from high-CTR banners, including detailed information about each element (layout, color, catchy copy, stock photos, motif, gaze, and font).

[1460] study

[1461] This data is then used to train a learning model, using machine learning algorithms such as random forests and neural networks to build a model that predicts optimal design elements.

[1462] Banner generation phase

[1463] Receiving information

[1464] The user inputs landing page information, including key messaging and promotional content, and requests the creation of a banner ad.

[1465] Banner Generation

[1466] The server generates an optimal banner based on the landing page information and the characteristics of high-CTR banners obtained from the learning model, including layout, color, catchy copy, stock photos, motif, line of sight, font, etc.

[1467] Collaboration with emotion engine

[1468] Acquiring emotion data

[1469] The server obtains the user's emotional data from the emotion engine, including whether the user is in a positive, negative, or neutral emotional state.

[1470] Adjusting design elements

[1471] The server adjusts the banner's design elements based on the acquired emotional data, for example, using bright colors when the user is expressing positive emotions and muted colors when expressing negative emotions, and dynamically changes the tagline accordingly.

[1472] Size development phase

[1473] Size range

[1474] The generated banner will be automatically expanded to multiple sizes. The device will handle this process and create banner images of different sizes, for example 300x250, 728x90, 160x600, etc.

[1475] Specific examples

[1476] User operations

[1477] Assume that a user sets up a landing page with promotion information for "New Year Sale!" and requests the generation of a banner ad. Also, assume that the user's sentiment is positive.

[1478] Server Processing

[1479] 1. The server collects and analyzes past data and extracts the characteristics of high CTR banners.

[1480] 2. The learning model is trained based on the extracted features so that it can generate optimal banners.

[1481] 3. Based on the landing page information provided by the user, the server generates a suitable banner containing the tagline "New Year Sale!"

[1482] 4. The server retrieves the user's emotional data from the emotion engine and adjusts the design by using brighter colors if the emotion indicates a positive one.

[1483] 5. The device will automatically expand the generated banner to various sizes such as 300x250, 728x90, and 160x600.

[1484] The above is a detailed description of the embodiment of the present invention. This system reduces the workload of designers, efficiently generates optimal banners that maximize CTR, and also enables design adjustments based on user emotions.

[1485] The processing flow will be explained below.

[1486] Step 1:

[1487] The server collects daily banner ad click-through rate (CTR) data. It obtains past banner ad features such as layout, color, catchy copy, stock photos, motifs, gaze, and font, as well as their CTR, from databases and log files.

[1488] Step 2:

[1489] The server analyzes the collected data and extracts the characteristics of banners with high CTR. Specifically, it filters data for banner ads with CTRs above a certain threshold and calculates the average characteristics. In this process, it uses statistical methods to identify which design elements contribute to a high CTR.

[1490] Step 3:

[1491] The server builds a training dataset based on the features extracted from high-CTR banners, including detailed information about each element (layout, color, catchy copy, stock photos, motif, gaze, font, etc.).

[1492] Step 4:

[1493] The server uses the constructed training dataset to train a model. For example, it uses machine learning algorithms such as random forests and neural networks to build a model for generating optimal banners. This model plays a role in predicting high CTR.

[1494] Step 5:

[1495] The user inputs landing page information, including key messaging and promotional content, and requests the creation of a banner ad.

[1496] Step 6:

[1497] The server receives the provided landing page information and generates the optimal banner based on it. The landing page information and the extracted characteristics of the high CTR banner are input into the learning model to determine the optimal design elements. The generated banner includes layout, color, catchy copy, stock photos, motif, line of sight, font, etc.

[1498] Step 7:

[1499] The server obtains the user's emotional data from the emotion engine, which includes the user's emotional state, such as positive, negative, or neutral.

[1500] Step 8:

[1501] The server adjusts the banner's design elements based on the acquired emotional data, for example, using bright colors when the user is expressing positive emotions and muted colors when expressing negative emotions. Furthermore, the tagline dynamically changes depending on the emotion.

[1502] Step 9:

[1503] The device receives the generated banner and automatically expands it to multiple sizes, for example, resizing it to different sizes such as 300x250, 728x90, and 160x600, and then saving it. This allows a banner that can be used in multiple ad spaces to be prepared at once.

[1504] Step 10:

[1505] The generated banners in multiple sizes are provided to the user, who can then use them in actual advertising operations to check the effectiveness of CTR.

[1506] As a concrete example, suppose a user sets up a landing page with promotional information for "New Year Sale!" and requests the generation of a banner ad. Let's also assume that the user's sentiment is positive. The server uses a learning model and sentiment engine to generate an optimal banner that combines the "New Year Sale!" tagline with bright colors. The device then expands this banner into multiple sizes and provides it to the user. This series of processes allows users to efficiently use banner ads that aim for a high CTR.

[1507] Example 2

[1508] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1509] Conventional audiovisual advertisement generation systems do not take into account the user's emotional response and therefore have difficulty generating optimal advertisements. Furthermore, the process of automatically expanding the generated advertisements into multiple sizes is also time-consuming. Therefore, a new system for creating more effective audiovisual advertisements and automatically expanding them is needed.

[1510] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1511] In this invention, the server includes means for collecting daily selection rate data, means for analyzing characteristics of past audiovisual advertisements, means for extracting characteristics of audiovisual advertisements with high selection rates, means for generating audiovisual advertisements using a learning algorithm, means for optimizing the generated audiovisual advertisements based on guide page information, means for adjusting design elements based on user emotions using an emotion processing engine, and means for automatically expanding the generated audiovisual advertisements into multiple dimensions, thereby enabling efficient generation of optimal audiovisual advertisements according to user emotions and automatic expansion thereof.

[1512] "Selection rate data" refers to data that indicates the rate at which users click or select an audiovisual advertisement displayed.

[1513] "Audiovisual advertising" refers to advertising that uses both visual and auditory sensations to provide information to users.

[1514] A "learning algorithm" is a machine learning method for predicting optimal design elements for audiovisual advertisements based on collected data.

[1515] "Guide page information" refers to the content of a web page that contains promotional information for a particular product or service.

[1516] An "emotion processing engine" is a system that analyzes a user's emotional state and adjusts design elements based on the results.

[1517] "Design elements" are the components of an audiovisual advertisement, and specifically include layout, color, wording, image materials, theme, line of sight, and font.

[1518] "Dimensions" refers to the physical size and proportions of an audiovisual advertisement.

[1519] "User" means an individual or organization that utilizes the system to generate and display audiovisual advertisements.

[1520] This invention relates to a system that collects and analyzes audiovisual advertisement selection rate data, combines a learning algorithm and an emotion processing engine to generate optimal audiovisual advertisements, and optimizes and automatically deploys the advertisements based on guide page information.

[1521] The hardware required to implement this system includes a database server, a web server, an emotion processing engine server, and a terminal. The software used includes a database management system (e.g., MySQL), a data processing library (e.g., Pandas, NumPy), a machine learning library (e.g., Scikit-learn, TensorFlow), and an image processing library (e.g., PIL).

[1522] The specific program processing is as follows.

[1523] The server collects audiovisual ad selection rate data daily from a database and log files, including design elements such as layout, color, wording, image materials, theme, eye direction, and typeface. The server uses SQL queries to extract relevant data from the database and Python scripts to retrieve additional information from the log files.

[1524] The server then analyzes the collected data using statistical methods. It extracts design elements of audiovisual advertisements with high selection rates, filters advertisements that exceed a certain threshold, and calculates average features. Specifically, the server creates a data frame using the Pandas library and calculates statistical information such as the mean and standard deviation using the NumPy library.

[1525] The server builds a training dataset based on features extracted from high-selection audiovisual ads. This dataset includes detailed information about each design element. The server saves the extracted features in a CSV file, which is then fed into machine learning algorithms. The training algorithms include random forests and neural networks, which are trained using Scikit-learn and TensorFlow.

[1526] The user inputs information about the page through the system interface and requests the generation of an audiovisual advertisement. The input information includes the promotional content and key messages. The user submits this information using a web form.

[1527] The server combines the information on the user's guide page with the characteristics of high-selection audiovisual ads estimated from the learning model to generate optimal audiovisual ads. Specifically, the server uses PIL (Python Imaging Library) to combine elements such as layout and catchy copy to generate ad images. A generative AI model is also used in this process.

[1528] The server obtains user emotional data in real time from an emotion processing engine. This data includes the user's emotional state, such as positive, negative, or neutral. The server adjusts design elements based on this emotional data. For example, bright colors are used to indicate positive emotions, and muted colors to indicate negative emotions. The tagline also changes dynamically depending on the emotion.

[1529] The generated audiovisual advertisement is automatically deployed to multiple dimensions by the device, which uses an image processing library to resize and reposition the advertisement image to the appropriate ratio.

[1530] As a concrete example, consider the case where a user sets up an information page with promotional information for "New Year Sale!" and requests the generation of a banner ad. Assume the user's sentiment is positive. In this case, the server collects and analyzes past data to extract features of audiovisual ads with high selection rates. Based on the extracted features, the server trains a model to generate an optimal audiovisual ad containing the slogan "New Year Sale!". Furthermore, the server obtains positive sentiment data from the sentiment engine and adjusts the design to a brighter color tone. The generated ad is automatically displayed in multiple sizes by the device.

[1531] Example prompt sentence:

[1532] "Generate optimal audiovisual ads taking into account sentiment data. The promotion is 'Summer Sale' and users are showing positive sentiment."

[1533] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1534] Step 1:

[1535] Collecting CTR data

[1536] The server collects audiovisual ad selection rate data from databases and log files daily. The databases and log files are used as input, and the extracted data includes design elements such as layout, color, wording, image materials, theme, eye line, and typeface. This data is collected using SQL queries and Python scripts and saved as a data frame.

[1537] Specifically, the server performs the following steps:

[1538] Use a SQL query to extract daily selectivity data from the database.

[1539] Run a Python script and retrieve additional information from the log file.

[1540] The collected data is converted into a data frame using the Pandas library and saved.

[1541] Step 2:

[1542] Analyzing CTR data

[1543] The server analyzes the collected data using statistical methods. The input is the data frame collected in step 1, and the output is to extract the characteristics of audiovisual advertisements with high selection rates. Specifically, it calculates the average and standard deviation, and filters out advertisements that exceed a certain threshold.

[1544] Specifically, the server performs the following steps:

[1545] Use the Pandas library to filter high-selectivity design elements from a data frame.

[1546] Using the NumPy library, calculate the mean and standard deviation of each design element.

[1547] The characteristics of high selection rate audiovisual advertisements are extracted and saved in a list.

[1548] Step 3:

[1549] Building a feature dataset

[1550] The server constructs a training dataset based on the extracted features of high-selectivity audiovisual advertisements. The input is the feature list extracted in step 2, and the output is a training dataset (CSV file).

[1551] Specifically, the server performs the following steps:

[1552] The extracted features are organized to build a dataset containing detailed information about each design element.

[1553] Convert the dataset to a CSV file and save it.

[1554] Step 4:

[1555] Training the learning model

[1556] The server trains a learning algorithm using the constructed training dataset. The input is the CSV file constructed in step 3, and the output is the trained model. The learning algorithms used include Scikit-learn's random forest and TensorFlow's neural network.

[1557] Specifically, the server performs the following steps:

[1558] Read the CSV file and generate a training dataset.

[1559] Train random forest and neural network models using Scikit-learn and TensorFlow libraries.

[1560] Save the trained model.

[1561] Step 5:

[1562] Receiving landing page information

[1563] The user inputs the information about the user's guide page through the system interface and requests the generation of an audiovisual advertisement. The input is the user's guide page information, and the output is that this information is sent to the server.

[1564] Specifically, the user performs the following steps:

[1565] Use the web form to enter your information.

[1566] The information is sent to the server.

[1567] Step 6:

[1568] Banner generation

[1569] The server generates an optimal audiovisual advertisement by integrating the information on the guide page received from the user and the features of the high-selection rate audiovisual advertisement estimated from the learning model. The input is the landing page information and the trained model, and the output is the generated audiovisual advertisement (image file).

[1570] Specifically, the server performs the following steps:

[1571] Analyze landing page information and extract key messages and promotional content.

[1572] Use a trained model to select the best design elements.

[1573] Audiovisual advertising images are generated using PIL (Python Imaging Library).

[1574] Step 7:

[1575] Acquiring emotion data

[1576] The server receives user emotion data from the emotion processing engine. The input is the user emotion data (via API), and the output is data used to adjust design elements according to the emotion.

[1577] Specifically, the server performs the following steps:

[1578] A request is sent to the emotion processing engine via the API to obtain emotion data.

[1579] Analyze the acquired emotion data.

[1580] Step 8:

[1581] Adjusting design elements

[1582] The server adjusts the design elements of the banner based on the acquired emotional data. The input is the emotional data and the generated audiovisual advertisement, and the output is the adjusted final audiovisual advertisement.

[1583] Specifically, the server performs the following steps:

[1584] Colors and catchy slogans are dynamically changed based on emotional data.

[1585] The audiovisual advertising images are re-edited using PIL.

[1586] Step 9:

[1587] Banner size options

[1588] The terminal automatically unwraps the generated audiovisual advertisement into multiple dimensions: the input is the adjusted final audiovisual advertisement, and the output is the audiovisual advertisement (multiple image files) in different sizes.

[1589] Specifically, the device performs the following steps:

[1590] Load the generated audiovisual advertisement and resize it to the different sizes you need.

[1591] Generate audiovisual ads optimized for each size, taking into account existing placement and proportions.

[1592] (Application example 2)

[1593] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1594] Conventional banner ad generation systems focus on optimizing click-through rates (CTR), but do not take into account the user's emotional state when generating banners. Therefore, further optimization is required to maximize CTR. Furthermore, it is necessary to support dynamic design adjustments based on user emotional data when displaying ads on different devices, such as smartphones and head-mounted displays.

[1595] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1596] In this invention, the server includes a means for collecting daily click-through rate (CTR) data, a means for analyzing the characteristics of past banner advertisements, a means for extracting the characteristics of high-CTR banners, a means for generating banners using a learning model, a means for optimizing the generated banners based on landing page information, a means for automatically expanding the generated banners to multiple sizes, a means for collecting user emotion data in real time, a means for adjusting the banner design elements based on the collected emotion data, and a means for displaying banners optimized for smartphones and head-mounted displays. This allows for the generation of optimal banners that take the user's emotional state into consideration, further maximizing CTR. Furthermore, dynamic design adjustments based on user emotion data can be realized even when displaying advertisements on different devices.

[1597] "Click-through rate (CTR)" is an indicator that indicates the ratio of the number of times an ad is clicked by a user to the number of times it is displayed.

[1598] A "banner ad" is an image or graphic advertisement that appears on a web page or within an application.

[1599] A "learning model" is a model that uses algorithms to learn patterns and characteristics based on past data and makes predictions and classifications for new data.

[1600] "Landing page information" refers to the content and information of the web page that a user who clicks on an advertisement first arrives at.

[1601] "Emotion data" is data that indicates the user's emotional state, such as positive, negative, or neutral, obtained from the user's facial expression and behavior.

[1602] "Design elements" refers to the components of a banner ad, such as layout, color, catchy copy, stock photos, motifs, line of sight, and font.

[1603] A "smartphone" is a mobile information terminal that, in addition to the functions of a mobile phone, can also browse the Internet and use applications.

[1604] A "head-mounted display" is a display device worn on the head, allowing users to experience virtual reality (VR) and augmented reality (AR).

[1605] "Real-time" means that data collection and processing occur simultaneously with actual time.

[1606] A "high CTR banner" is a banner ad that has a higher-than-average click-through rate (CTR).

[1607] An "emotion engine" is software or a system that analyzes a user's emotional state and provides that data.

[1608] The system of the present invention aims to dynamically generate advertising banners based on user emotional data and maximize click-through rates (CTR). The system uses advanced machine learning models to analyze past data and apply real-time emotional information to generate optimal banners that will attract users' attention.

[1609] 1. Generate a program

[1610] The system consists of the following main components:

[1611] 1. Data Collection Module

[1612] 2. Data Analysis Module

[1613] 3. Learning Model Training Module

[1614] 4. Banner generation module

[1615] 5. Emotion Data Acquisition Module

[1616] 6. Banner Adjustment Module

[1617] 7. Display module

[1618] 2. Explain the system's processing in natural language

[1619] Data Collection Module

[1620] The server collects daily click-through rate (CTR) data, which includes the number of clicks and impressions for each banner ad.

[1621] Data Analysis Module

[1622] The server analyzes the characteristics of past banner ads to determine which features have a high CTR. This involves analyzing design elements such as banner layout, color, tagline, stock photos, motifs, line of sight, and font.

[1623] Learning Model Training Module

[1624] The server trains a learning model based on the analysis results, using TensorFlow to build a neural network to generate the optimal banner.

[1625] Banner Generation Module

[1626] Using the learning model, the server generates the optimal banner based on the landing page information, automatically selecting taglines and images that correspond to the promotional information provided by the user.

[1627] Emotion data acquisition module

[1628] The user's device (smartphone or head-mounted display) uses a camera to collect the user's emotional data in real time. OpenCV and FaceAPI are used to analyze the user's emotional state (positive, negative, neutral, etc.) from their facial expressions.

[1629] Banner Adjustment Module

[1630] The server dynamically adjusts the banner's design elements based on the acquired emotional data: for example, if the user is expressing positive emotions, the banner's colors will be brighter and the copy's tone will be more optimistic.

[1631] Display Module

[1632] The generated banners are displayed on the user's device in the optimal size and format, which includes a process to automatically generate banner images of different sizes.

[1633] 3. Specific examples and prompt sentences

[1634] Specific examples

[1635] When a landing page with promotion information "New Year Sale!" is input, the server generates an optimal banner with the tagline "50% off on all items." If the user's sentiment data is analyzed as positive, the banner design will be composed of bright colors. The displayed banner will also be optimized for the screen size of smartphones and head-mounted displays.

[1636] Prompt Sentence Examples

[1637] Input data:

[1638] Landing page information: {'message': 'New Year Sale!', 'promotion_details': '50% off on all items'}

[1639] Emotion data: {'emotion': 'positive'}

[1640] Requirements:

[1641] Always include the tagline "New Year Sale!"

[1642] Use bright colors.

[1643] Use high CTR elements when generating images.

[1644] Generation goal:

[1645] To generate an optimal advertisement banner based on a user's positive emotional state.

[1646] This system can generate advertising banners that reflect the user's emotional state, maximizing click-through rates (CTR). It can also dynamically adjust the design of ads displayed on different devices based on the user's emotional data.

[1647] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1648] Step 1:

[1649] The server collects daily click-through rate (CTR) data, which includes the number of clicks and impressions for each banner ad. The data is retrieved from databases and log files and stored in a database.

[1650] Input: Number of clicks and impressions of past banner ads

[1651] Data processing: Data collection and organization

[1652] Output: Organized CTR data

[1653] Step 2:

[1654] The server analyzes the characteristics of past banner ads to identify banner design elements (layout, color, tagline, stock photo, motif, gaze, font) that cause CTRs above a certain threshold.

[1655] Input: CTR data and banner ad characteristics information

[1656] Data calculations: statistical analysis

[1657] Output: High CTR banner feature data

[1658] Step 3:

[1659] The server extracts features of high-CTR banners and builds a learning dataset. TensorFlow is used to train the learning model. A neural network is used to build a model to predict the optimal banner design.

[1660] Input: High CTR banner feature data

[1661] Data Computing: Training Machine Learning Models

[1662] Output: A trained learning model

[1663] Step 4:

[1664] The user's device (smartphone or head-mounted display) uses a camera to collect the user's emotional data in real time. OpenCV and FaceAPI are used to analyze the user's emotional state (positive, negative, neutral, etc.) from their facial expressions.

[1665] Input: Camera image

[1666] Data Computing: Sentiment Analysis

[1667] Output: Real-time emotion data

[1668] Step 5:

[1669] Once the user enters the landing page information, the server uses this information to generate the optimal banner. The generative AI model inputs prompt text and creates a banner with the desired tagline and image.

[1670] Input: Landing page information

[1671] Data calculation: Banner generation

[1672] Output: The generated banner

[1673] Prompt Sentence Examples

[1674] Input data:

[1675] Landing page information: {'message': 'New Year Sale!', 'promotion_details': '50% off on all items'}

[1676] Emotion data: {'emotion': 'positive'}

[1677] Requirements:

[1678] Always include the tagline "New Year Sale!"

[1679] Use bright colors.

[1680] Use high CTR elements when generating images.

[1681] Generation goal:

[1682] To generate an optimal advertisement banner based on a user's positive emotional state.

[1683] Step 6:

[1684] The server dynamically adjusts the banner's design elements based on the acquired emotional data: for example, if the user is expressing positive emotions, the banner's color may be brightened and the tagline may be changed to a more positive tone.

[1685] Input: Generated banner and sentiment data

[1686] Data calculation: Dynamic banner adjustment

[1687] Output: Adjusted banner

[1688] Step 7:

[1689] The generated banner is displayed on the user's device in the optimal size and format. The server automatically generates banner images of different sizes and displays them in a format suitable for smartphones and head-mounted displays.

[1690] Input: Adjusted banner

[1691] Data processing: Size expansion

[1692] Output: Banner optimized for display

[1693] Through these steps, the system can generate optimal banners that take into account the user's emotional state and maximize click-through rates (CTR).

[1694] The specific processing unit 290 transmits the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the control target 443 to output the result of the specific processing. The microphone 238 acquires voice indicating a user input regarding the result of the specific processing. The control unit 46A transmits voice data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[1695] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1696] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the robot 414.

[1697] The emotion identification model 59 as an emotion engine may determine the user's emotion according to a specific mapping. Specifically, the emotion identification model 59 may determine the user's emotion according to an emotion map (see FIG. 9), which is a specific mapping. Similarly, the emotion identification model 59 may determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[1698] FIG. 9 is a diagram illustrating an emotion map 400 on which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. Emotions closer to the center of the concentric circles are more primitive. Emotions representing states and actions arising from a state of mind are arranged on the outer edges of the concentric circles. The concept of emotion includes both affect and mental states. Emotions generally generated from reactions occurring in the brain are arranged on the left side of the concentric circles. Emotions generally induced by situational judgment are arranged on the right side of the concentric circles. Emotions generally generated from reactions occurring in the brain and induced by situational judgment are arranged on the upper and lower sides of the concentric circles. Furthermore, the emotion of "pleasure" is arranged on the upper side of the concentric circles, and the emotion of "discomfort" is arranged on the lower side. In this way, in the emotion map 400, multiple emotions are mapped based on the structure by which emotions are generated, and emotions that tend to occur simultaneously are mapped close to each other.

[1699] These emotions are distributed in the 3 o'clock direction on emotion map 400, and typically fluctuate between relief and anxiety. In the right half of emotion map 400, situational awareness dominates over internal sensations, resulting in a sense of calm.

[1700] The inside of emotion map 400 represents what is going on in the mind, and the outside of emotion map 400 represents behavior, so the further you go outside emotion map 400, the more visible the emotions become (the more they are expressed in behavior).

[1701] Human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. Emotions can also be created for robots, automobiles, and motorcycles, based on various balances, such as posture and remaining battery life. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. An emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on Voice Emotion Recognition and Emotional Brain Physiological Signal Analysis Systems, Tokushima University, Doctoral Dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map lists emotions belonging to the "reaction" domain, where sensation is dominant. The right half of the emotion map lists emotions belonging to the "situation" domain, where situational awareness is dominant.

[1702] The emotion map defines two emotions that promote learning. One is a negative emotion on the situation side, around the middle of "repentance" or "reflection." In other words, this occurs when the robot experiences negative emotions such as "I never want to feel this way again" or "I don't want to be scolded again." The other is a positive emotion on the response side, around "desire." In other words, this occurs when the robot experiences positive feelings such as "I want more" or "I want to know more."

[1703] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​indicating each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple pieces of training data that are combinations of user input and emotion values ​​indicating each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions that are located close to each other have similar values, as in the emotion map 900 shown in FIG. 10. FIG. 10 shows an example in which multiple emotions, "relieved," "calm," and "reassuring," have similar emotion values.

[1704] The system according to the present disclosure has been described above mainly with respect to the functions of the data processing device 12, but the system according to the present disclosure is not necessarily implemented on a server. The system according to the present disclosure may be implemented as a general information processing system. The present disclosure may be implemented, for example, as a software program running on a personal computer or an application running on a smartphone, etc. The method according to the present disclosure may be provided to users in the form of SaaS (Software as a Service).

[1705] In the above embodiment, an example was given in which the specific processing is performed by one computer 22, but the technology of the present disclosure is not limited to this, and the specific processing may be distributed and performed by a plurality of computers including the computer 22. For example, the data generation model 58 may be provided in an external device of the data processing device 12, and data may be generated in the external device in accordance with input data.

[1706] In the above embodiment, an example in which the specific processing program 56 is stored in the storage 32 has been described, but the technology of the present disclosure is not limited to this. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-transitory storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-transitory storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes the specific processing in accordance with the specific processing program 56.

[1707] Alternatively, the specific processing program 56 may be stored in a storage device such as a server connected to the data processing device 12 via the network 54, and the specific processing program 56 may be downloaded and installed on the computer 22 in response to a request from the data processing device 12.

[1708] It is not necessary to store all of the specific processing program 56 in a storage device such as a server connected to the data processing device 12 via the network 54, or to store all of the specific processing program 56 in the storage 32; only a portion of the specific processing program 56 may be stored.

[1709] The hardware resource for executing a specific process can be any of the following processors: An example of a processor is a CPU, which is a general-purpose processor that functions as a hardware resource for executing a specific process by executing software, i.e., a program. Another example of a processor is a dedicated electrical circuit, such as an FPGA (Field-Programmable Gate Array), a PLD (Programmable Logic Device), or an ASIC (Application Specific Integrated Circuit), which is a processor with a circuit configuration designed specifically for executing a specific process. Each processor has built-in or connected memory, and each processor uses the memory to execute the specific process.

[1710] The hardware resource that executes the specific processing may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Also, the hardware resource that executes the specific processing may be a single processor.

[1711] As an example of a system configured with a single processor, first, one processor is configured by combining one or more CPUs and software, and this processor functions as a hardware resource that executes a specific process. Second, there is a system that uses a processor that realizes the functions of an entire system including multiple hardware resources that execute a specific process on a single IC chip, as typified by SoC (System-on-a-chip). In this way, a specific process is realized using one or more of the above-mentioned various processors as hardware resources.

[1712] Furthermore, the hardware structure of these various processors can be, more specifically, an electric circuit that combines circuit elements such as semiconductor devices. The specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps may be deleted, new steps may be added, or the processing order may be rearranged, without departing from the spirit of the invention.

[1713] The above-described description and illustrations are a detailed explanation of the parts related to the technology of the present disclosure and are merely an example of the technology of the present disclosure. For example, the above description of the configuration, functions, actions, and effects is an explanation of an example of the configuration, functions, actions, and effects of the parts related to the technology of the present disclosure. Therefore, it goes without saying that unnecessary parts may be deleted, new elements may be added, or replacements may be made to the above-described description and illustrations within the scope of the gist of the technology of the present disclosure. Furthermore, to avoid confusion and facilitate understanding of the parts related to the technology of the present disclosure, the above-described description and illustrations omit explanations of common technical knowledge that do not require particular explanation to enable the implementation of the technology of the present disclosure.

[1714] All publications, patent applications, and technical standards mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent application, or technical standard was specifically and individually indicated to be incorporated by reference.

[1715] The following is further disclosed regarding the above embodiment.

[1716] (Claim 1)

[1717] A means of collecting daily click-through rate (CTR) data;

[1718] A means of analyzing the characteristics of past banner ads,

[1719] A method to extract the characteristics of high CTR banners,

[1720] A means for generating banners using a learning model;

[1721] A means for optimizing the generated banner based on landing page information;

[1722] A way to automatically expand the generated banner into multiple sizes,

[1723] A system including:

[1724] (Claim 2)

[1725] 10. A system comprising: means for receiving the landing page information of claim 1 and adding a catchy slogan to a banner generated based thereon.

[1726] (Claim 3)

[1727] A system in which the learning model of claim 1 includes means for optimizing each element of a banner's layout, color, catchy copy, photographic material, motif, line of sight, and font using the characteristics of high CTR banners.

[1728] "Example 1"

[1729] (Claim 1)

[1730] A means of collecting daily click-through rate (CTR) data;

[1731] A means of analyzing the characteristics of past banner ads,

[1732] A method to extract the characteristics of high CTR banners,

[1733] A means for generating banners using a learning model;

[1734] A means for optimizing the generated banner based on landing page information;

[1735] A way to automatically expand the generated banner into multiple sizes,

[1736] A means to save the optimized banners to cloud storage;

[1737] A system including:

[1738] (Claim 2)

[1739] 10. The system of claim 1, further comprising means for receiving landing page information and adding taglines and visual information to a banner generated based thereon.

[1740] (Claim 3)

[1741] The system of claim 1, wherein the learning model includes means for optimizing each element of the banner's layout, color, catchy copy, photographic material, motif, line of sight, and font using the characteristics of high CTR banners.

[1742] "Application Example 1"

[1743] (Claim 1)

[1744] A means of collecting daily click-through rate (CTR) data;

[1745] A means of analyzing the characteristics of past banner ads,

[1746] A method to extract the characteristics of high CTR banners,

[1747] A means for generating banners using a learning model;

[1748] A means for optimizing the generated banner based on landing page information;

[1749] A way to automatically expand the generated banner into multiple sizes,

[1750] means for generating and displaying dynamically optimized banner advertisements in real time;

[1751] A system including:

[1752] (Claim 2)

[1753] The system of claim 1, further comprising means for receiving the landing page information of claim 1 and adding a catchy slogan to a banner generated based thereon, and for generating dynamically optimized banners in real time based on click-through rate data.

[1754] (Claim 3)

[1755] The system of claim 1, wherein the learning model of claim 1 uses the characteristics of high CTR banners to optimize each element of the banner, including layout, color, catchy copy, photo material, motif, line of sight, and font, and performs processing to optimize the generated banner in real time based on landing page information.

[1756] "Example 2: Combining Emotion Engines"

[1757] (Claim 1)

[1758] a means of collecting daily selection rate data;

[1759] A means of analyzing the characteristics of past audiovisual advertising;

[1760] means for extracting features of high selection rate audiovisual advertisements;

[1761] means for generating audiovisual advertisements utilizing a learning algorithm;

[1762] means for optimizing the generated audiovisual advertisement based on the directory page information;

[1763] means for adjusting design elements based on a user's emotions using an emotion processing engine;

[1764] means for automatically expanding the generated audiovisual advertisement into multiple dimensions;

[1765] A system including:

[1766] (Claim 2)

[1767] 10. The system of claim 1, further comprising means for receiving the directory page information and adding text to the audiovisual advertisement generated based thereon.

[1768] (Claim 3)

[1769] 10. The system of claim 1, wherein the learning algorithm includes means for utilizing characteristics of high selection rate audiovisual advertisements to optimize the placement, color, wording, image material, theme, eye line, and typeface elements of the audiovisual advertisement.

[1770] "Application example 2 when combining emotion engines"

[1771] (Claim 1)

[1772] A means of collecting daily click-through rate (CTR) data;

[1773] A means of analyzing the characteristics of past banner ads,

[1774] A method to extract the characteristics of high CTR banners,

[1775] A means for generating banners using a learning model;

[1776] A means for optimizing the generated banner based on landing page information;

[1777] A way to automatically expand the generated banner into multiple sizes,

[1778] a means for collecting user emotion data in real time;

[1779] A means to adjust banner design elements based on collected emotional data;

[1780] A method to display banners optimized for smartphones and head-mounted displays,

[1781] A system including:

[1782] (Claim 2)

[1783] 2. The system of claim 1, further comprising: means for receiving landing page information and adding a catchy slogan to a banner generated based thereon; and means for dynamically adjusting the generated banner according to user emotion data.

[1784] (Claim 3)

[1785] The system of claim 1 includes a means for optimizing each element of the banner, including layout, color, catchy copy, photographic material, motif, gaze, and font, using the characteristics of high CTR banners, and a means for dynamically changing each element based on collected emotional data. [Explanation of symbols]

[1786] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot< / url:> < / url:> < / url:> < / url:>

Claims

1. A means of collecting daily click-through rate (CTR) data; A means of analyzing the characteristics of past banner ads, A method to extract the characteristics of high CTR banners, A means for generating banners using a learning model; A means for optimizing the generated banner based on landing page information; A way to automatically expand the generated banner into multiple sizes, A system including:

2. 10. A system comprising: means for receiving the landing page information according to claim 1 and adding a catchy slogan to a banner generated based thereon.

3. A system in which the learning model of claim 1 includes means for optimizing each element of the banner's layout, color, catchy copy, photographic material, motif, line of sight, and font by utilizing the characteristics of high CTR banners.

Images