ATTP

Rolling programme in design

The Translation Engine

A rolling contribution challenge where ATTP participants build high-quality open datasets for Yoruba, Igbo, Hausa, and Nigerian Pidgin: recorded speech, translated sentence pairs, transcribed conversation. Monthly prizes for the best contributions. Every contributor credited by name.

The idea

AI systems serve Nigerians badly for a simple reason: Yoruba, Igbo, Hausa, and Nigerian Pidgin are severely under-represented in the data those systems learn from. This is a known, documented gap, and research labs around the world actively want it closed.

The Translation Engine turns closing it into structured, credited, rewarded work. Participants record speech, translate sentence pairs, and transcribe real conversations in their own languages. Contributions are quality-checked, and the resulting datasets are published openly under ATTP's name with every contributor listed.

A teenager in Bariga becomes, verifiably, a named contributor to African AI infrastructure. And unlike a competition, this never ends. It compounds, month after month, into an asset the whole field can use.

At a glance

Format

Rolling, monthly prize cycles

Languages

Yoruba, Igbo, Hausa, Pidgin

Ages

13 to 35

Equipment needed

A phone

Cost to join

Free

Output

Open datasets, every contributor named

How it works

1

Pick a language and a task

Speech recording, sentence translation, or conversation transcription, in Yoruba, Igbo, Hausa, or Pidgin.

2

Contribute through the platform

Guided tasks with clear quality standards. You work at your own pace, from a phone.

3

Pass quality review

Contributions are checked by reviewers fluent in the language. Quality counts for more than volume.

4

Get credited and win monthly

Every accepted contribution carries your name in the published dataset. The strongest contributors each month win cash prizes.

Who it is for

Anyone fluent in a Nigerian language who wants their name on real AI infrastructure. Students, teachers, market traders, grandparents' grandchildren doing the recording at home. Fluency is the qualification, not a CV.

Questions

Who owns the data?

Datasets are published under an open licence with ATTP as steward and every contributor credited. The licence terms are public before you contribute anything, so you know exactly what you are giving and how it can be used.

Why would anyone pay for open data?

Labs and companies sponsor the creation of high-quality open datasets because they need them and cannot easily produce them. Sponsorship funds the prizes and the review process. Participants never pay and are never charged.

What counts as quality?

Clear audio, accurate translation, natural language rather than textbook stiffness, and correct metadata. The standards are published per task, and reviewers give feedback rather than silent rejections.

Can I contribute in more than one language?

Yes, if you are genuinely fluent in more than one. Quality review applies per language.

Register interest

Be there when this launches.

Tell us who you are and how you want to be involved. We will write to you before anything goes public.