ProfanityBench: Multilingual Profanity and Abuse Annotations Across 715 Languages

ProfanityBench is a multilingual dataset of profanity and abusive language. It contains approximately 82,000 annotated entries across 715 languages and 459 countries or dialects. The dataset was created by BibbyResearch and was last updated on Hugging Face in May 2026.

Use Cases

• Train profanity detection models based on the annotated entries across many languages.
• Benchmark hate speech classifiers using the provided hate-speech flags.
• Study linguistic and cultural variation in offensive language based on the included etymology and cultural context.
• Analyze the evolution of slang and tone in abusive language across different regions.

Strengths

• Approximately 82,000 annotated entries provides a substantial corpus.
• Covers 715 languages and 459 countries or dialects, indicating broad linguistic coverage.
• Includes multiple annotation dimensions such as severity, hate-speech flags, tone, and etymology.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Last updated 2026-05-30 11:41:21; freshness should be verified.

Provenance

Source

BibbyResearch on Hugging Face.

Freshness

Last updated 2026-05-30 11:41:21.

Geography

Global, covering 459 countries or dialects.