Datasets:

suleiman2003
/

afri-temp-data2

language:
  - ha
license: cc-by-4.0
task_categories:
  - automatic-speech-recognition
tags:
  - audio
  - speech
  - hausa
  - african-voices
  - asr
pretty_name: AfricanVoices Hausa Train
size_categories:
  - 10K<n<100K

AfricanVoices Hausa — Train Split

Hausa speech dataset from AfricanVoices.io, uploaded in Hugging Face audiofolder format.

Usage

from datasets import load_dataset

ds = load_dataset("suleiman2003/afri-temp-data2", split="train")
print(ds[0])
# {{'audio': Audio(...), 'transcript': '...', 'gender': '...', ...}}

Dataset Structure

Each sample contains:

Field	Description
`audio`	FLAC audio file (auto-loaded by HF)
`transcript`	Hausa text transcription
`speaker_id`	Speaker identifier
`gender`	Speaker gender (m/f)
`age_group`	Speaker age bracket
`education`	Education level
`duration`	Audio duration (seconds)
`language`	Language code
`snr`	Signal-to-noise ratio
`domain`	Topic domain (ev/ag/he/bu)

Source

Platform: AfricanVoices.io
Organization: Data Science Nigeria
Supported by: Gates Foundation