Sign in to view source links and access this dataset
Description
AfriDoc-MT is a document-level multi-parallel translation dataset covering English and five African languages: Amharic, Hausa, Swahili, Yorùbá, and Zulu. The dataset was created by Masakhane and was last updated on Hugging Face in October 2025. Its structure suggests it contains separate training, development, and test splits for both document and sentence-level data across Health and Tech domains.
Use Cases
Train document-level neural machine translation models based on the parallel text data.
Benchmark translation quality for African languages using the provided test splits.
Conduct cross-lingual transfer learning experiments based on the multi-parallel structure.
Analyze domain adaptation between Health and Tech topics based on the separate domain folders.
Strengths
Covers five African languages (Amharic, Hausa, Swahili, Yorùbá, Zulu) alongside English.
Provides both document-level and sentence-level data splits.
Includes separate data for Health and Tech domains.
Limitations
Column-level documentation is absent; field semantics must be inferred after download.
Row count, file formats, and license information are unknown.
Freshness should be verified; last updated 2025-10-14 12:10:37.
Provenance
Source
Masakhane
Freshness
Last updated 2025-10-14 12:10:37.
Geography
Africa (languages: Amharic, Hausa, Swahili, Yorùbá, Zulu)
License is unknown; users must verify permissions before use.