Text & NLP data · ready-made dataset
Pakistani safety and toxicity lexicon
A safety classifier trained on English data scores a code-switched Urdu slur as neutral, because it has never seen one. This lexicon exists so that moderation and safety systems can recognise the abuse that actually circulates in Pakistani languages, including the coded and sectarian forms that no translated list contains.
Datasheet summary
What is in Pakistani Safety & Toxicity Lexicon
Culturally specific slurs, hate speech, harassment, sectarian language and coded terms, severity-graded by 3 native raters each.
| Dataset ID | TDA-TX-004 |
|---|---|
| Modality | Text |
| Languages / coverage | Urdu, Punjabi, Pashto, Sindhi |
| Typical first delivery | 10k–50k items |
| Free sample | 1,000 items with full metadata, within one business day |
| Provenance | Collected in Pakistan; collection date, location, contributor pseudonym, device profile and consent version on every record |
| Licence | Perpetual, worldwide, sublicensable; derivative-model rights stated explicitly. Exclusivity optional. Licensing detail |
| Delivery | Your S3, GCS or Azure bucket with manifests, checksums, QA report and datasheet |
| Guarantee | 14-day acceptance window; anything outside signed tolerance re-collected free |
Included in every delivery
- Slurs, hate speech, harassment, sectarian and coded terms
- Severity grade from three native raters per item
- Urdu, Punjabi, Pashto and Sindhi
- Rater agreement metadata
Built for
- Content moderation and safety classifiers for South Asian platforms
- Red-teaming and safety evaluation of multilingual LLMs
- Toxicity detection in code-switched Urdu–English text
- Policy enforcement research on sectarian and coded language
Evaluate before you buy
Start with the free sample, then a pilot
Request 1,000 items with full metadata from Pakistani Safety & Toxicity Lexicon — identical in format and quality to the paid corpus, no watermark, no sales call. Pilots are creditable against production and start within five business days of a signed SOW.
Average first response under 6 business hours. NDA signed same day.
- Consent form, licence text and DPA available before you commit
- Datasheet lists composition, QA method and known limitations
- Subsets by language, region or condition on request
- Custom collection to your spec using the same protocol
Answers
Questions about this dataset
Each item is graded independently by three native raters, and the per-item agreement is delivered with the grade so you can see where raters disagreed.
Yes. Much real abuse is code-switched Urdu–English, and the lexicon includes those forms rather than only single-language items.
Safety, trust-and-safety and research teams. Because of the content, we discuss intended use before delivery and the licence restricts use to safety and moderation purposes.