Search Engine Image Attractiveness Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Search engines face challenges in providing aesthetically pleasing search results that balance relevance, freshness, and attractiveness, as not all images relevant to a query are visually appealing, which can lead to decreased user engagement.
Innovation Solution
A system and method that selects and scores images based on an attractiveness value, in addition to relevance and freshness, to generate search results pages, using a machine learning model trained on curated image pairs to determine image attractiveness, and combines these scores to prioritize visually appealing content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If images are selected based solely on relevance to query intent, then search result relevance is improved, but aesthetic appeal deteriorates
Solution Approach 1:
The patent introduces a new parameter (attractiveness value) to the existing selection criteria. Instead of using only relevance, the system now evaluates images based on multiple parameters including relevance, freshness, and attractiveness. This parameter expansion allows the system to balance between search result relevance and aesthetic appeal, resolving the contradiction by finding optimal images that satisfy both criteria simultaneously.
Solution Approach 2:
The patent creates a composite scoring system that combines multiple evaluation dimensions (relevance score, freshness score, and attractiveness value) into an overall score. This composite approach allows the system to integrate different aspects of image quality and select images that perform well across multiple criteria, thereby resolving the trade-off between relevance and aesthetic appeal.
2Loss of information
If all relevant images are included in search results, then information completeness is improved, but user engagement deteriorates
Solution Approach 1:
The patent extracts and removes unattractive images from the search results based on their attractiveness values. By filtering out images with low attractiveness scores, the system maintains information completeness for relevant topics while improving user engagement by presenting only visually appealing images. This extraction process ensures that users encounter high-quality images that enhance their search experience.
Solution Approach 2:
The system changes the selection parameter from binary relevance inclusion to a scaled attractiveness-based filtering mechanism. By using attractiveness values as a continuous parameter for image selection, the system can maintain comprehensive information coverage while selectively presenting only the most engaging images to users, thereby improving user engagement without sacrificing information completeness.
3Measurement precision
If image attractiveness is evaluated using complex machine learning models, then aesthetic evaluation accuracy is improved, but system complexity increases
Solution Approach 1:
The patent introduces an intermediary component - the attractiveness value generation module - that bridges the gap between complex machine learning models and the search result generation system. This intermediary layer processes images through trained models to produce attractiveness values, which then feed into the overall scoring mechanism. This modular approach maintains high evaluation accuracy while managing system complexity through clear separation of concerns.
Solution Approach 2:
The system performs preliminary image evaluation and attractiveness scoring during the content retrieval phase, before the final search results are generated. By conducting these computations in advance, the system prepares pre-scored image data that can be quickly integrated into the final results without adding significant complexity to the main search processing pipeline. This preliminary action allows accurate aesthetic evaluation while minimizing impact on overall system complexity.
Data Source
AI summary
Systems and methods for identifying search results in response to a search query are presented. More particularly, images are selected as search results, at least in part, according to an attractiveness value associated with the images. Upon receiving a search query, a set of content is identified according to the query intent of the search query and includes at least one image. The identified set of content is ordered according an overall score determined according to relevance and, in the case of the at least one image, according to an attractiveness value. A search results generator selects items from the set of content according to their overall scores, including the at least one image, generates a search results page, and returns the search results page to the requesting party.