Double Iterative Flavored Rank for Search Relevance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Search engines face challenges in providing contextually relevant results as query-independent scores fail to differentiate between results based on the user's intent, leading to irrelevant hits for queries like 'jaguar price' that could refer to a car, a cat, or an operating system.
Innovation Solution
A scoring engine assigns source and destination scores to web pages based on their relevance and reachability, using a random surfer model to recursively update scores, and employs parameters to control the diffusion of scores, ensuring that results are contextually relevant to the user's query.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If query-independent scores are used to rank search results, then the search engine can process queries efficiently, but the results fail to differentiate based on user intent and context
Solution Approach 1:
The patent changes the parameters used for scoring from static, query-independent metrics to dynamic, context-sensitive parameters. The scoring system incorporates query-specific terms, user intent signals, and contextual information to adjust relevance scores in real-time, thereby improving measurement precision without sacrificing processing efficiency through automated parameter adjustment
Solution Approach 2:
The patent introduces dynamic scoring mechanisms where relevance scores are not fixed but adapt based on query context, user behavior, and document characteristics. The system dynamically adjusts weighting factors and scoring criteria for each query, enabling differentiated results for ambiguous queries like 'jaguar price' based on detected user intent
2Quantity of substance
If the search engine returns all potential hits for a query, then comprehensive results are provided, but irrelevant results increase when query context is not considered
Solution Approach 1:
The patent applies local quality by treating different portions of the search result set differently based on their relevance to the query context. High-quality, contextually relevant results are prioritized and ranked higher, while less relevant results are either demoted or filtered out. The scoring system assigns different quality weights to different documents based on their alignment with query intent
Solution Approach 2:
The patent incorporates feedback mechanisms where user interactions with search results (clicks, dwell time, reQueries) are used to refine and adjust relevance scores. The system learns from user behavior patterns and adjusts the scoring algorithm to better differentiate relevant from irrelevant results in future queries, improving overall result reliability
Data Source
AI summary
Determining the relevance of a web node is disclosed. A seed score value of a first type is assigned to a seed set of nodes. A score value of a second type is derived for the web node based on a mapping of a reachability relationship between one or more seed nodes and the web node. A score value of the first type is derived for the web node based on a mapping of a reachability relationship between the web node and one or more evaluation nodes having derived weight values of the second type.