Amazon Reviews 2023

McAuley-Lab released this collection of Amazon product reviews and metadata in late 2023, covering user interactions up to September 2023. The dataset contains between 10 billion and 100 billion records, including text reviews, numerical ratings, and product images. It serves as a significant update to the previous 2018 version with improved metadata cleanliness and temporal precision.

Use Cases

• Sentiment analysis using the review text and numerical rating columns
• Recommendation system training using user-item interaction history and category information
• Price sensitivity modeling using the price and brand metadata fields
• Multi-modal product classification combining images with text descriptions

Strengths

• Millisecond-level timestamps for high-resolution temporal modeling
• Includes product images for multi-modal analysis
• Estimated scale of 10 billion to 100 billion records
• Cleaner metadata compared to the 2018 version

Limitations

• Geographic bias toward the US region as indicated by dataset tags
• Potential for label noise inherent in crowdsourced user ratings
• Significant storage and compute requirements due to the 10B+ record scale

Provenance

Source

McAuley-Lab

Collection Method

Scraped from Amazon web platforms

Time Range

Ending September 2023

Freshness

Contains reviews updated through September 2023.

Geography

United States