Open Brand: 250,000 Amazon Brand Annotations Across 8 Categories

Open Brand contains over 250,000 product brand-value annotations across eight Amazon product profile categories. Released by kassemsabeh in early 2024, the dataset focuses on attribute-value extraction and named entity recognition within the e-commerce domain.

Use Cases

• Training Named Entity Recognition (NER) models to identify brand names in e-commerce text
• Developing attribute-value extraction systems using the 50,000+ unique brand values
• Benchmarking brand normalization algorithms across eight different product categories

Strengths

• 250,000+ brand-value annotations
• 50,000+ unique brand values providing high label diversity
• Covers 8 distinct Amazon product categories
• MIT licensed for permissive reuse

Limitations

• Lack of documented column names or sample data in the repository
• Potential platform bias as data is sourced exclusively from Amazon product profiles
• Unknown file format and size

Provenance

Source

Amazon product profiles

Collection Method

Annotated

Freshness

Last updated in January 2024.