Creative Content Ranking Using Pairwise Resonance Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current techniques for image selection and placement in digital advertisements are inadequate due to the large number of potential images available, with only a few enhancing audience resonance, and there is a lack of effective methods to predict which images will engage specific audience segments.
Innovation Solution
A system using artificial intelligence and machine learning algorithms to classify images based on human subjective preferences, assigning resonance values and confidence scores through pairwise comparisons, and generating models to predict image performance in different contexts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a large number of images are provided for advertisement placement, then the variety and potential resonance with different audience segments increases, but the complexity of image selection and the time required to identify effective images increases significantly
Solution Approach 1:
The patent replaces manual image selection processes with an automated machine learning system that uses trained models to predict image performance. The system automatically analyzes images, extracts features, and ranks them based on predicted resonance with target audiences, eliminating the need for manual evaluation of large image collections.
Solution Approach 2:
The patent introduces an intermediary machine learning model that acts as a bridge between the large pool of available images and the advertisement placement decision. This model pre-evaluates and scores images based on their potential performance, serving as an intermediate filtering layer that simplifies the final selection process.
2Ease of operation
If manual image selection processes are used, then flexibility in evaluating subjective image appeal exists, but the time and resources required to evaluate large numbers of images increase significantly
Solution Approach 1:
The patent performs preliminary evaluation of images by training machine learning models on historical performance data before actual advertisement placement decisions are made. The models are pre-trained to recognize patterns and features that correlate with successful image performance, enabling rapid evaluation when real selection is needed.
Solution Approach 2:
The patent substitutes manual subjective evaluation with an automated machine learning system that can process and evaluate images at scale. The system uses trained models to predict image performance metrics, providing both speed and consistency in the evaluation process while maintaining the ability to capture subjective appeal through pattern recognition.
3Device complexity
If traditional image selection methods are used without predictive analytics, then simplicity of the process is maintained, but the ability to accurately predict which images will resonate with specific audience segments is limited
Solution Approach 1:
The patent segments the audience into distinct groups based on demographics, interests, and behaviors, and trains separate machine learning models for each segment. This allows the system to predict image performance specifically for each audience segment rather than using a generic evaluation approach, improving prediction accuracy for targeted advertising.
Solution Approach 2:
The patent changes the parameters used for image evaluation by incorporating multiple features beyond basic visual characteristics, including contextual information, audience segment characteristics, and historical performance data. The machine learning models analyze these varied parameters to generate more accurate predictions of image resonance with specific audiences.
Data Source
AI summary
Methods and systems for predicting performance of creative content are disclosed. Exemplary implementations may: receive a collection of images; provide a context to a user; serially cause display of pairs of images on a computer interface; receive user responses indicating which image of each pair is preferred given the context; determine a resonance value for each image based on a number of times the user responses indicate each image is preferred when displayed in a pair of images; determine a confidence score for each image; generate one or more models for predicting image performance based on one or more of the resonance value and the confidence score for each image; receive a plurality of candidate images; determine, using at least one model, a first metric set for each candidate; and cause display of a listing of the candidate images, the listing including the first metric set for each candidate image.