Sign in to view source links and access this dataset
Description
MAFAND-MT is the largest machine translation benchmark for African languages in the news domain, covering 21 languages. The dataset was created by Masakhane and published in 2022. Train, validation, and test splits are available for 16 languages, with validation and test sets for an additional 5 languages.
Use Cases
Train machine translation models based on the provided parallel text data.
Benchmark translation system performance across 21 African languages.
Analyze linguistic patterns and translation quality in the news domain.
Strengths
Covers 21 distinct African languages, making it the largest benchmark of its kind.
Provides structured train/validation/test splits for 16 languages.
Limitations
Column-level documentation is absent; field semantics must be inferred after download.
Row count is unknown, which may limit suitability assessment.
Provenance
Source
Masakhane
Freshness
Last updated 2023-09-11 18:01:53; freshness should be verified.
Geography
Africa
License is unknown; terms of use should be verified.