Harmful and Harmless Text Examples in Portuguese and English

Examples of harmful and harmless language. It aggregates samples from seven source datasets, including Anthropic/hh-rlhf and allenai/real-toxicity-prompts. The data is available in both Portuguese and English.

Use Cases

• Train a multilingual classifier to distinguish harmful from harmless language using Portuguese and English examples.
• Fine-tune a safety-aligned language model using text samples from the Anthropic/hh-rlhf and allenai/prosocial-dialog sources.
• Benchmark toxicity detection models on aggregated data from sources like dirtycomputer/Toxic_Comment_Classification_Challenge and told-br.
• Analyze cross-lingual patterns in hate speech using the Portuguese data from Paul/hatecheck-portuguese and English data from skg/toxigen-data.

Strengths

• Aggregates samples from seven established source datasets, including Anthropic/hh-rlhf and allenai/real-toxicity-prompts.
• Provides multilingual coverage with text in both Portuguese and English.
• Last updated on June 9, 2025, indicating recent maintenance.

Limitations

• Specific dataset size, row count, and column structure are unknown.
• Potential for class imbalance or label inconsistencies due to aggregation from multiple sources with different annotation schemes.
• Geographic and demographic coverage of the Portuguese and English samples is unspecified.

Provenance

Source

Aggregated from seven datasets: Anthropic/hh-rlhf, allenai/prosocial-dialog, allenai/real-toxicity-prompts, dirtycomputer/Toxic_Comment_Classification_Challenge, Paul/hatecheck-portuguese, told-br, and skg/toxigen-data.

Collection Method

Samples were collected from the listed source datasets.

Freshness

Last updated on 2025-06-09.