Query Sense Hyperplane Ranking for Image Search Relevance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Image search engines often return irrelevant results due to the reliance on text association, where images may not match the search query despite relevant text nearby, leading to low relevance and user inefficiency in finding desired images.
Innovation Solution
The method involves determining images associated with multiple senses of a query by mapping them into a space, identifying hyperplanes representing these senses, calculating scores based on distances to these hyperplanes, and ranking images accordingly to separate relevant images from irrelevant ones, using click-through information and labeling terms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If image search engines rely on text association to locate images, then the search process becomes simpler and faster, but the relevance of search results deteriorates because images may not match the search query
Solution Approach 1:
The patent segments the image search process into multiple stages: initial text-based filtering to obtain candidate images, followed by sense-based classification to categorize images according to multiple meanings of the query, and finally relevance ranking. This segmentation allows the system to maintain fast search operations while improving relevance by addressing the limitations of text-only association through multi-stage processing.
Solution Approach 2:
The patent introduces a new dimension to image search by classifying images according to multiple senses or meanings of the query terms. Instead of relying solely on text proximity (one dimension), the system adds semantic sense classification (additional dimensions) to create a multi-dimensional search space where images are ranked based on their alignment with different interpretations of the query, thereby improving relevance without sacrificing speed.
2Quantity of substance
If image search engines return hundreds of results based on text association, then the quantity of search results increases, but the quality of results deteriorates due to inclusion of irrelevant images
Solution Approach 1:
The patent applies local quality by differentiating and ranking images based on their specific alignment with different senses of the query. Rather than treating all results uniformly, the system identifies which images best match each particular sense interpretation and prioritizes them accordingly. This allows the system to maintain a comprehensive result set while improving local quality by ensuring the top results are highly relevant to the user's intent.
Solution Approach 2:
The patent incorporates feedback mechanisms where user interactions (such as clicks on images) are used to refine and re-rank search results. The system continuously learns from user behavior patterns to adjust the relevance scoring, thereby improving the quality of results over time while maintaining a sufficient quantity of diverse results. This feedback loop allows the system to balance result volume with result quality dynamically.
3Device complexity
If the image search engine uses only text near images for ranking, then the system complexity is reduced, but the ability to detect and measure image relevance becomes insufficient
Solution Approach 1:
The patent makes the image search system multi-functional by integrating multiple capabilities: text-based filtering, sense classification, image feature analysis, and relevance ranking. The system performs multiple functions (searching, classifying, measuring relevance) within a unified framework, which increases measurement precision for relevance detection while managing complexity through modular integration of these functions rather than requiring completely separate systems for each task.
Data Source
AI summary
A server device determines a plurality of images for a query. One or more images, of the plurality of images, are associated with one or more senses of the query. The server device maps the plurality of images into a space by representing the plurality of images with corresponding points in the space; determines one or more hyperplanes in the space based on the corresponding points in the space; calculates one or more scores for the plurality of images based on the corresponding points and the one or more hyperplanes; and ranks the one or more images based on the one or more scores.