Search Query Generation Using Similarity Profiles

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing search systems face challenges in generating accurate search queries for images based on text inputs, particularly when the text does not provide sufficient terms, leading to inefficient image search results.

Innovation Solution

The system generates a search query by determining a set of terms from both text-based and image-based content items, using similarity profiles and relevance scores to select optimal search terms for an image search, thereby enhancing the accuracy and completeness of the search query.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional text-based search queries are used for image search, then the search process is simple, but the search accuracy deteriorates when text inputs are limited or insufficient

Engineering Contradiction:
Improvesearch accuracyVSAvoidsearch system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent combines text-based search terms with image-based visual features to create a hybrid search query system. The system extracts terms from both the input text and associated content items, then uses similarity profiles to identify visually similar images that complement the text-based search results, thereby improving overall search accuracy without requiring a complete system overhaul

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent introduces a similarity profile as an intermediary mechanism that bridges text-based queries and image-based retrieval. The similarity profile analyzes relationships between content items and identifies visually similar images that may not be directly matched by text keywords alone, acting as a mediator to enhance search precision when text inputs are limited

Inventive Principle:
Principle #24Intermediary (Mediator)

2Loss of information

If multiple content items are analyzed to generate search terms, then the completeness of search queries improves, but the processing time increases

Engineering Contradiction:
Improvesearch query completenessVSAvoidquery processing time
Core Design Contradiction:
Loss of informationVSLoss of time

Solution Approach 1:

The patent applies partial action by selectively analyzing only the most relevant content items for term extraction rather than processing all available content. The system uses similarity scores to identify and focus on content items that are most likely to contribute valuable search terms, thereby reducing processing time while maintaining query completeness

Inventive Principle:
Principle #16Partial or excessive action

Solution Approach 2:

The patent performs preliminary analysis by pre-computing similarity profiles and term relationships between content items before actual search queries are executed. This preliminary processing allows the system to quickly retrieve and combine relevant terms during search operations without performing extensive analysis in real-time, thus reducing query processing time

Inventive Principle:
Principle #10Preliminary action

3Reliability

If similarity profiles are generated from multiple content items, then the relevance of search terms improves, but the computational complexity increases

Engineering Contradiction:
Improveterm relevanceVSAvoidcomputational complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent segments the computational task of generating similarity profiles by processing content items in groups or batches rather than simultaneously analyzing all content. The system divides the large-scale similarity computation into smaller, manageable segments that can be processed independently and then combined, reducing peak computational complexity while maintaining term relevance

Inventive Principle:
Principle #1Segmentation

Data Source

PatentUS20240386067A1Search Query Generation Based Upon Received Text
Publication Date: 2024.11.21 YAHOO ASSETS LLC
  • US20240386067A1 patent drawing
  • US20240386067A1 patent drawing
  • US20240386067A1 patent drawing

AI summary

In an example, a first set of text may be received from a client device. A set of content items may be selected from among content items based upon the first set of text and a plurality of sets of content item text associated with the content items. A set of terms may be determined based upon the first set of text and the set of content items. A similarity profile associated with the set of terms may be generated. The similarity profile is indicative of similarity scores associated with similarities between terms of the set of terms. Relevance scores associated with the set of terms may be determined based upon the similarity profile. One or more search terms may be selected from among the set of terms based upon the relevance scores. A search may be performed based upon the one or more search terms.