How Groupon Automatically Categorizes Merchant Services Using Text Analysis
A system that automatically scans merchant websites and uses high-precision search queries to label their services, helping platforms like Groupon organize thousands of business listings.
Patent Number
US 9330167
Status
Active
Filing Date
May 13, 2013
Grant Date
May 3, 2016
Expiration
~May 2033 (estimated)
Claims
18
Assignee
Groupon Inc
Inventors
Nick Pendar
Citations
25 forward · 10 backward
What it covers
This patent describes a way to teach a computer how to read a merchant's website and figure out what service they offer, such as 'spa' or 'auto repair.' It works by first looking at a large collection of known documents to build search queries that are highly accurate, measured by a precision score. The system calculates this score by dividing the number of 'true positive' results by the total number of results found. Once these high-quality queries are refined, the system applies them to new, unlabeled text from a merchant's site to automatically assign the correct category label.
What it doesn't cover
- —Does not cover general-purpose search engines that do not use a specific precision-threshold-based query selection process.
- —Does not cover manual tagging or human-in-the-loop classification systems.
- —Does not cover machine learning models that classify text without using the specific feature-index-pair and distance-measure query generation method described.
- —Does not cover image-based merchant classification.
The clever bit
The system generates queries based on 'feature index pairs'—tracking not just the words, but their specific positions in a sentence—and then filters those queries by a precision threshold to ensure only the most reliable indicators are used for classification.
Why it matters
For companies like Groupon that aggregate thousands of local deals, manually categorizing every merchant is impossible. This technology allowed them to scale their marketplace by automating the classification of unstructured text found on merchant websites, ensuring that a user searching for 'massage' actually finds relevant spa deals.
Real-world examples
- 1.Automated categorization of local business listings on Groupon
- 2.Internal merchant classification tools for daily deal platforms
- 3.Automated tagging of unstructured web content for recommendation engines
Generated by PatentBrief · Not legal advice · patentbrief.org
US 9330167 · 2026