Browse Datasets

MAFAND-MT

Machine Translation for African Languages

machine-translation sequence-to-sequence 100MB CC-BY-4.0
Access Dataset Read Paper

Description

MAFAND-MT is a machine translation dataset covering 21 African languages paired with English and French. It provides parallel corpora for training and evaluating translation systems for African languages.

Quick Start

Python 
from datasets import load_dataset

# Load English-Swahili translation data
dataset = load_dataset("masakhane/mafand", "en-swa")

# View a sample
print(dataset['train'][0])
# Output: {'translation': {'en': '...', 'swa': '...'}}

Project Ideas

Bidirectional Translation App

Build a mobile-friendly translation app between English and an African language

Intermediate

Document Translation Pipeline

Create a system for translating documents while preserving formatting

Advanced

Ethical Considerations

  • Translation quality varies by language pair
  • May not handle code-switching well
  • Domain-specific terminology may be missing

Additional Information

MAFAND-MT was created by the Masakhane community to address the lack of machine translation resources for African languages.

Citation

@inproceedings{adelani-etal-2022-mafand,
    title = "A Few Thousand Translations Go a Long Way!",
    author = "Adelani, David and others",
    booktitle = "NAACL",
    year = "2022"
}