Task
Automatic Speech Recognition
W
Ephraimmm/WaxalNLPr
Automatic Speech Recognition · Ephraimmm· 425
["cc-by-sa-4.0","cc-by-4.0"]
719 GB
This repository hosts a large multilingual speech corpus for 27 African languages, split into two task collections: Automatic Speech Recognition (ASR) — natural speech paired with human transcriptions — and Text-to-Speech (TTS) — single-speaker studio recordings paired with the scripted text that was read aloud. The dataset card and data itself indicate the underlying recordings were gathered thro
# download instantly
mlforge datasets pull Ephraimmm/WaxalNLPr
mlforge datasets pull Ephraimmm/WaxalNLPr
Dataset details
Language
ach
License
["cc-by-sa-4.0","cc-by-4.0"]
Size
719 GB
Rows / images
2.6M
Creator
Ephraimmm
Downloads
425
Source
huggingface_datasets
Updated
2026-02-04
About Ephraimmm/WaxalNLPr
This repository hosts a large multilingual speech corpus for 27 African languages, split into two task collections: Automatic Speech Recognition (ASR) — natural speech paired with human transcriptions — and Text-to-Speech (TTS) — single-speaker studio recordings paired with the scripted text that was read aloud. The dataset card and data itself indicate the underlying recordings were gathered through partnerships with Makerere University, the University of Ghana, Digital Umuganda, and Media Trust, and are documented in the paper "WAXAL: A Large-Scale Multilingual African Language Speech Corpus" (arXiv:2602.02734). This repo makes that corpus available as ready-to-load Hugging Face datasets configs.