Nigerian language AI data

Commercial-safe Nigerian language data for better African AI.

Collect ASR, TTS, translation, RLHF preferences, red-team safety, and cultural quality data from bank-account-verified Nigerian contributors with explicit AI-data consent.

Chat on WhatsApp Explore AI data

Human contact: Chat on WhatsApp · [email protected]

Launch tracks + Tier-2 on demand

Consent

Linked to data reuse grant

Gold

Quality checks excluded from corpus

Tier-1 and Tier-2 languages

Launch tracks cover Pidgin, Yoruba, Igbo, Hausa, and Nigerian English. Fulfulde, Kanuri, Tiv, Edo, Efik, and other Tier-2 languages are scoped on demand after supply checks.

Language and region tagging
Native-speaker collection
Train and held-out splits at collection

RLHF preference data

A/B preference votes for Nigerian prompts, exported in the NGPT-ready `task_id,worker_id,choice` contract.

A/B preference UI
Tie/both-bad stored but dropped from binary DPO export
Majority-vote collection support

ASR and TTS voice

Speech prompts, voice-likeness consent, transcript review, and provenance-ready corpus manifests.

Voice recordings on real devices
Transcript-backed gold checks
Consent-gated export

Translation and instruction data

Natural Nigerian-language translations and useful Nigerian instruction-response tasks, not literal machine-translation filler.

English ↔ Nigerian language pairs
Held-out benchmark split
Domain-specific prompts

Consent and provenance

Approved exports require linked consent records and exclude gold probes plus testers flagged by gold-check failure.

Append-only consent records
Anonymised contributor ids
Tamper-evident provenance hashes

Buyer deliverables

JSONL/CSV manifests, datasheet, language splits, contributor demographics where consented, and delivery notes.

HuggingFace-friendly metadata
Open speech schema compatibility
Custom schemas for labs

Collect the data your model actually needs.

Tell us the language, modality, volume, domain, consent requirements, and delivery format. We will scope a pilot before a full run.

Chat on WhatsApp Start an order