Browse Datasets

NaijaVoices

Culturally Rich Nigerian Language Dataset

language-modeling asr speech-recognition 1,800 hours of speech CC BY-NC-SA 4.0
Access Dataset

Description

Expert-curated text and speech in Nigerian languages. Offers authentic, culturally rich sentences distinct from religious or news texts often found online. Uses a community-benefiting data farming approach.

Quick Start

Python 
from datasets import load_dataset

# Requires approval on HuggingFace
dataset = load_dataset("naijavoices/naijavoices-dataset", split="train", streaming=True)
print(next(iter(dataset))['transcript'])

Project Ideas

Conversational Text Generator

Build a conversational text generator for Nigerian languages

Intermediate

Code-Switching Analysis

Analyze code-switching patterns in Nigerian text

Advanced

Ethical Considerations

  • Non-commercial license; users must register
  • Uses data farming approach that benefits the community