Semantic GUI Modification for More Relevant Product Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Marketplaces face challenges in helping customers find relevant products amidst a vast array of offerings, as free-form text searches often yield irrelevant results due to the sheer volume of products available.
Innovation Solution
A system that utilizes semantic analysis to generate query embeddings, determine entity embeddings with similarity scores, filter based on criteria, and modify the graphical user interface to prioritize relevant products by associating keywords with entities, thereby enhancing search relevance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If free-form text search is used to search through millions of products, then the system can handle large product volumes, but the search accuracy deteriorates and irrelevant products are returned
Solution Approach 1:
The patent introduces semantic embeddings as an intermediary layer between the search query and product database. Instead of direct text matching, the system converts both queries and product titles into semantic vector representations, allowing for more accurate similarity-based matching even across large product volumes
Solution Approach 2:
The system transforms the search problem from traditional text parameter matching to semantic space parameter matching. By changing the search parameters from exact text matches to cosine similarity measurements in embedding space, the system maintains accuracy while scaling to millions of products
2Measurement precision
If semantic analysis is implemented to improve search accuracy, then relevant products are prioritized, but the system complexity increases
Solution Approach 1:
The system performs preliminary semantic analysis by pre-computing embeddings for product titles and storing them in the database. This advance preparation allows the search system to quickly compare query embeddings against pre-processed product embeddings without performing complex analysis during the actual search operation
Solution Approach 2:
The patent replaces traditional mechanical text-matching mechanisms with semantic embedding-based similarity computation. Instead of using keyword matching algorithms, the system substitutes them with vector space model operations, which are computationally more efficient for capturing semantic meaning
3Measurement precision
If entity embeddings and similarity scoring are used to filter results, then search relevance improves, but the processing time increases
Solution Approach 1:
The system applies partial filtering by first computing similarity scores for all products, then applying threshold-based filtering to retain only the most relevant results. This selective approach processes the full dataset initially but only performs detailed filtering on a reduced subset, balancing comprehensiveness with efficiency
Data Source
AI summary
Systems and methods including one or more processors and one or more non-transitory storage devices storing computing instructions configured to run on the one or more processors and perform: receiving a search query from a user during a current user session; generating a query embedding for the search query; determining one or more entity embeddings with similarity scores that satisfy a threshold, the one or more entity embeddings including keywords associated with entities; filtering the one or more entity embeddings based on filtering criteria; generating a mapping for the search query based on the one or more entity embeddings, the mapping including an association between keywords in the search query and an entity; and modifying a graphical user interface to display items associated with the entity in an upper portion of the graphical user interface during the current user session. Other embodiments are disclosed herein.