Search Ranking via Corpus Statistics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current search engines face challenges in accurately ranking search results based on user needs, as they rely on traditional methods that do not effectively utilize corpus search statistics to determine relevance, leading to suboptimal results.
Innovation Solution
A system and method that determine the relevance of search results by using corpus search statistics, including click measures and search fractions, to provide a measure of relative relevance for ranking search results, which can modify the baseline ranking function and improve search result accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional ranking methods are used, then device complexity is reduced, but search result relevance deteriorates
Solution Approach 1:
The system pre-computes corpus search statistics (search fractions, click measures) and stores them in a database before they are needed for ranking. This preliminary action allows the ranking engine to quickly retrieve and use pre-analyzed statistics without performing complex real-time computations, thus improving search result relevance while keeping the ranking process relatively simple
Solution Approach 2:
The patent introduces corpus search statistics as an intermediary layer between the search query and the final ranking. These statistics (search fractions and click measures) act as mediators that translate raw search data into meaningful relevance signals, allowing the ranking engine to improve accuracy without directly handling complex raw data analysis
2Reliability
If corpus search statistics are utilized, then search result relevance is improved, but computational overhead increases
Solution Approach 1:
The system performs computationally intensive statistical analysis (calculating search fractions and click measures) in advance and stores the results. When a search query arrives, the ranking engine only needs to retrieve pre-computed statistics and apply simple weighting, dramatically reducing real-time computational energy consumption while maintaining improved search relevance
Solution Approach 2:
The patent computes statistics at a corpus level rather than at the individual document level. This partial action approach aggregates data to a higher level of abstraction, reducing the total computational work needed while still providing sufficient signal for improving search result relevance through the relative relevance measure
3Measurement precision
If multiple corpus statistics are computed, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent divides the relevance measurement task into separate statistical components: search fractions (measuring query distribution across corpora) and click measures (measuring user preference). By segmenting the analysis into these distinct metrics, the system achieves precise multi-dimensional measurement while keeping each individual computation relatively simple and manageable
Solution Approach 2:
The system transforms raw search data into standardized statistical parameters (search fractions and click measures) with defined mathematical formulations. This parameter transformation allows precise relevance measurement through controlled mathematical operations while maintaining computational simplicity through consistent formulas that can be efficiently implemented and scaled
Data Source
AI summary
Methods, systems, and apparatus, including computer program products, for ranking search results of a search query using corpus search statistics. In one aspect, a method includes determining a first relevance of a first corpus to a search query, determining a second relevance of a second corpus to the search query, determining a measure of relative relevance of the first corpus and the second corpus to the search query, and providing the measure of relative relevance to a ranking engine for ranking of search results for a new search corresponding to the search query.