Machine-Learning Search Refinement Using User Feedback Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional search techniques often fail to provide desired results due to improper use of search criteria and correlation of terms, leading to the provision of irrelevant information, wasting resources and user time.
Innovation Solution
A refined search method using machine learning to select and alter search criteria based on user interactions and preferences, allowing for improved search results through user feedback and machine learning adjustments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional search algorithms use complex criteria to tailor search results, then search result relevance is improved, but search accuracy deteriorates due to improper use and correlation of terms
Solution Approach 1:
The patent implements feedback mechanisms where user interactions with search results (clicks, selections, refinements) are captured and used to adjust search criteria weighting. The system learns from user behavior patterns to improve term correlation and criteria application, resolving the contradiction between using complex criteria for relevance while maintaining accuracy through continuous refinement based on actual user needs
Solution Approach 2:
The search system dynamically adjusts search criteria weights and correlations based on real-time user interactions and historical data. Rather than using static complex algorithms, the system adapts term importance and criteria application dynamically, allowing it to maintain both relevance through complexity and accuracy through adaptability to specific user contexts
2Loss of information
If search engines provide detailed tailored results based on multiple criteria, then information completeness is improved, but user time consumption deteriorates due to inability to quickly identify desired results
Solution Approach 1:
The patent segments search results into hierarchical levels and categories based on user selections and interaction patterns. By organizing comprehensive information into structured segments, the system maintains information completeness while enabling users to quickly navigate and identify desired results without being overwhelmed by undifferentiated detailed information
Solution Approach 2:
The system performs preliminary actions by pre-processing and organizing search criteria and results based on historical user behavior and selection patterns. Search results are pre-segmented and weighted according to likely user preferences, so when users perform searches, they receive already-organized comprehensive information that reduces their time to find desired results
3Adaptability or versatility
If search algorithms use prior user actions as search criteria, then search personalization is improved, but search reliability deteriorates when prior actions do not reflect current user needs
Solution Approach 1:
The patent implements dynamic adjustment of personalization based on user feedback and current search context. The system continuously evaluates whether prior user actions remain relevant to current search needs and adjusts the weight of historical criteria accordingly, maintaining personalization adaptability while ensuring reliability by down-weighting or discarding outdated contextual information
Solution Approach 2:
User feedback on search results (selections, refinements, corrections) is captured and used to adjust the influence of prior user actions on current search criteria. The system learns when historical behavior patterns remain valid and when they should be overridden, resolving the contradiction between personalization through prior actions and reliability when those actions no longer reflect current needs
Data Source
AI summary
This document describes techniques and devices for a refined search with machine learning. These techniques improve computer-aided searches through enabling selection of search criteria used in a prior search and providing a refined search result based on that selection. Furthermore, a machine-learning component of a search engine can be altered to improve future search results based on the selection and an indication of the desirability of the refined search result.