Search Filter Scoring Using Popularity Tables
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for filtering and presenting search results on network-based marketplaces face challenges in accurately interpreting keyword queries and efficiently generating relevant listings, as they struggle to distinguish between competing interpretations of attribute-value pairs and prioritize results based on popularity.
Innovation Solution
A system that receives a query, identifies multiple filter sets, scores them based on probabilities of attribute-value pairs in completed listings, and uses the highest scoring filter set to generate search results, while also presenting filters and their values in an ordered manner based on popularity, utilizing a popularity table to determine the relevance and co-occurrence of attributes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple filter sets are identified and scored based on probabilities, then search result accuracy is improved, but system complexity increases
Solution Approach 1:
The system segments the search filtering process into multiple independent filter sets, each representing a different interpretation of the query. Each filter set is scored separately based on probability calculations from historical data, allowing the system to evaluate and compare multiple filtering approaches independently before selecting the best one.
Solution Approach 2:
The system performs preliminary actions by pre-calculating probabilities of attribute-value pairs from completed listings and storing them in a popularity table. This pre-computed data is then used to quickly score different filter sets during search operations, improving accuracy without adding significant complexity to real-time processing.
2Measurement precision
If filter sets are scored using probability calculations from historical data, then relevance of search results is improved, but processing time increases
Solution Approach 1:
The system performs probability calculations in advance by analyzing completed listings and storing the results in a popularity table. This pre-computation allows the system to quickly retrieve and use probability values during search operations without performing complex calculations in real-time, thus maintaining high relevance accuracy while reducing processing time.
Solution Approach 2:
The system creates a simplified copy of the complex probability relationships by storing pre-calculated probability values in a popularity table. This copy allows fast access and comparison of filter set scores without requiring the system to re-calculate probabilities during each search operation.
3Ease of operation
If the system presents filters and values in ordered manner based on popularity, then user experience is improved, but computational overhead increases
Solution Approach 1:
The system pre-calculates popularity metrics for filters and values based on historical transaction data and stores them in the popularity table. During search operations, the system simply retrieves and applies these pre-computed rankings to order filters and values, improving user experience without requiring significant computational resources during active use.
Data Source
AI summary
Systems and methods to seed a search are described. The system identifies a seed listing included in listings that describe items being offered for sale on a network-based marketplace. The system identifies a seed filter context based on the seed listing and probabilities. The probabilities describe occurrences of attribute-value pairs in a plurality of listings that respectively describe items that were previously transacted on the network-based marketplace. The system extracts values from the seed listing based on the seed filter context. The system initializes the seed filter context based on the values and generates search results based on the seed filter context. The search results include a second plurality of listings that are identified from the first plurality of listings. Finally, the system communicates interface information to a client machine including the seed filter context and at least one listing from the second plurality of listings.