ecommerce_ops · ecommerce · workflow

Amazon builds COSMO commonsense knowledge graph to improve product recommendation F1 by 60%

Amazon's product recommendation engine lacked commonsense reasoning to infer contextual product needs, such as that pregnant women searching for shoes might want slip-resistant footwear.

How it works
Common implementation structure
How this type of workflow is generally built, generalized across documented cases — not tied to any one vendor's stack. Click any stage to read what happens there. Specific products that implement these stages appear in “Tools commonly seen” below.
Stage 1 · Customer query submitted
A customer submits a query to the Amazon Store, initiating the product recommendation process.
Tools used
COSMOLLMs
Outcome

The COSMO-enhanced recommendation model achieved a 60% increase in macro F1 with frozen encoders, and maintained a 28% macro F1 edge and 22% micro F1 edge over the best baseline after fine-tuning.

What failed first

LLMs used to generate commonsense relationship hypotheses tend to produce empty or trivial rationales, requiring heuristic filtering and additional validation steps to remove low-quality outputs.

Results
Volume60%
Source

https://www.amazon.science/blog/building-commonsense-knowledge-graphs-to-aid-product-recommendation?tag=soumet-20

How we source this →

Grounding & classification
Source type: technical build writeup
20 fields verified against source quotes.
data extractionrecommendation systemknowledge baseproduct catalogfailure mode describedhuman review describedmetric backednamed customersource backedtools describedworkflow describedecommerceaccuracy improvementtechnical build writeupecommerce opsextract classify route