Image Search Ranking Using Hover Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional search systems rely heavily on click-through rates to rank image search results, which may not accurately reflect user interest, especially for uncommon queries with limited selection data, and do not effectively utilize hover data to adjust rankings.
Innovation Solution
The method involves determining a quality measure for image search results based on a combination of click counts and hover counts, adjusting hover counts using a weighting factor that varies with query popularity, and calculating a click-to-hover ratio to refine the ranking process, ensuring that images receiving many hovers but few clicks are demoted and that hover data is weighted appropriately for unpopular queries.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional search systems rely heavily on click-through rates to rank image search results, then the ranking process is simple and based on available data, but the ranking accuracy does not reflect user interest accurately, especially for uncommon queries
Solution Approach 1:
The patent combines multiple user interaction signals (clicks, hovers, time spent) into a unified quality measure for ranking images. By merging these different types of user engagement data, the system achieves more accurate ranking especially for uncommon queries where click data is limited, as hover data provides additional information about user interest.
Solution Approach 2:
The patent introduces hover data as an intermediary signal between image display and click selection. Hover data serves as a preliminary indicator of user interest that occurs before the actual click, providing the system with earlier and more nuanced feedback about which images attract user attention, thereby improving ranking accuracy for queries with limited click data.
2Measurement precision
If the system incorporates hover data to adjust rankings, then user interest is reflected more accurately, but the system complexity increases due to additional data processing
Solution Approach 1:
The patent makes the hover data collection and processing system universal by implementing it across all image search results consistently. The same hover tracking and quality measure calculation mechanisms are applied to all images regardless of query type, allowing the system to handle both common and uncommon queries uniformly while maintaining accurate user interest measurement without requiring separate complex processing paths.
Solution Approach 2:
The patent changes the parameters used for ranking by introducing hover count and time spent as new measurable parameters alongside traditional click-through rate. By adding these parameters and integrating them into the quality measure calculation, the system achieves more accurate user interest measurement while managing complexity through systematic parameter integration rather than fundamentally redesigning the ranking architecture.
3Measurement precision
If hover counts are adjusted using weighting factors based on query popularity, then rankings are refined for uncommon queries, but the calculation complexity increases
Solution Approach 1:
The patent implements dynamic weighting factors that adjust hover count importance based on query popularity. For uncommon queries where click data is scarce, the system dynamically increases the weight of hover data in the quality measure calculation. For common queries with abundant click data, the hover weight is reduced. This dynamic adjustment refines rankings for uncommon queries while managing processing efficiency by adapting calculations to data availability rather than applying uniform complex processing to all queries.
4Measurement precision
If the system demotes images with high hover counts but low click counts, then click magnet images are identified and removed from top rankings, but additional processing steps are required
Solution Approach 1:
The patent uses feedback from the relationship between hover counts and click counts to identify and demote click magnet images. By continuously monitoring the ratio of clicks to hovers for each image, the system detects when an image attracts excessive hover attention but fails to convert to clicks, indicating it may be a click magnet. This feedback mechanism enables automatic detection and ranking adjustment without requiring separate complex identification processes.
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for ranking images using hover data. In one aspect, a method includes determining a click count and a hover count for an image and a search query pair. The click count specifies a number of times that an image search result that includes a representation of the image has been selected when provided in response to the search query. The hover count specifies a number of times that the representation of the image has been hovered over when the image search result has been provided in response to the search query. A quality measure for the image with respect to the search query is determined. The quality measure is based on the click count and the hover count. A ranking of the image is adjusted for the search query based on the quality measure for the image.